1 2018-10-04 Martin Liska <mliska@suse.cz>
4 * tree-profile.c (init_ic_make_global_vars):
5 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
6 Come up with new ic_tuple* variables. Emit
7 __gcov_indirect_call{,_topn} variables.
8 (gimple_gen_ic_profiler): Access the variable
10 (gimple_gen_ic_func_profiler): Access
11 __gcov_indirect_call.callee field.
12 (gimple_init_gcov_profiler): Use ptr_type_node.
13 * value-prof.c (gimple_ic): Use ptr_type_node.
15 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17 PR tree-optimization/85787
18 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
19 into this function and add support for detecting multiple phis.
20 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
22 2018-10-04 Martin Liska <mliska@suse.cz>
25 * ipa-inline.c (inline_to_all_callers_1):
26 Call ultimate_alias_target for node being inlined.
28 2018-10-04 Thomas Preud'homme <thomas.preudhomme@linaro.org>
30 * lra-constraints.c (process_address_1): Bail out for all
31 satisfied fixed constraints.
33 2018-10-03 Jeff Law <law@redhat.com>
35 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
37 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
38 * tree.h (get_typenode_from_name): Prototype.
40 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
42 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
43 Change operand 2 predicate to nonimmediate_operand.
44 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
46 2018-10-03 Martin Sebor <msebor@redhat.com>
47 Jeff Law <law@redhat.com>
49 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
51 (get_string_length): Detect unterminated arrays.
52 (format_string): Same.
53 (format_directive): Warn about unterminated arrays.
54 (handle_gimple_call): Mark statements with no_warning as needed.
56 2018-10-03 Jim Wilson <jimw@sifive.com>
58 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
59 also define __riscv_abi_rve. Delete trailing white space.
61 2018-10-03 Paul Koning <ni1d@arrl.net>
63 Enable LRA register allocator for PDP11.
64 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
67 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
68 * config/pdp11/pdp11.opt (-mlra): New option.
69 * doc/invoke.texi (PDP-11 Options): Document -mlra.
71 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
73 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
74 (*<absneg:code>extend<mode>xf2): Ditto.
76 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
78 PR tree-optimization/87415
79 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
82 2018-10-02 Jeff Law <law@redhat.com>
84 * gimple-fold.c (get_range_strlen): Only set *nonstr when
85 an unterminated string is discovered. Bubble up range
86 even for unterminated strings.
87 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
88 indicates the string was not terminated via NONSTR.
90 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
92 * tree-vrp.c (extract_range_from_unary_expr): Special case all
94 Do not do anything special for anti-ranges.
96 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
98 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
99 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
101 2018-10-03 Martin Liska <mliska@suse.cz>
103 PR gcov-profile/86109
104 * coverage.c (coverage_begin_function): Do not
105 mark lambdas as artificial.
106 * tree-core.h (struct GTY): Remove tm_clone_flag
107 and introduce new lambda_function.
108 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
110 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
113 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
114 P8_VECTOR and VSX are enabled.
116 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
118 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
119 0x3907 as CPU model number.
121 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
123 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
124 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
125 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
126 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
127 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
128 * config/s390/s390.md: Likewise. Rename also the cpu attribute
129 value from arch12 to z14.
131 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
133 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
135 (isinf<mode>2): Ditto.
137 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
139 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
140 before emitting fxam. Perform calculations in XFmode.
142 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
144 * match.pd (((X /[ex] A) +- B) * A): New transformation.
146 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
149 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
150 * tree.c (signed_or_unsigned_type_for): Handle complex.
152 2018-10-02 Jeff Law <law@redhat.com>
154 * gimple-fold.c (get_range_strlen): Remove dead code.
156 2018-10-02 Martin Sebor <msebor@redhat.com>
157 Jeff Law <law@redhat.com>
159 * builtins.c (unterminated_array): Add new arguments.
160 If argument is not terminated, bubble up size and exact
162 (expand_builtin_strnlen): Detect, avoid expanding
163 and diagnose unterminated arrays.
164 (c_strlen): Fill in offset of start of unterminated strings.
165 * builtins.h (unterminated_array): Update prototype.
167 2018-10-02 Richard Biener <rguenther@suse.de>
169 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
170 of haddv4df, first reduce to SSE width and exploit the fact
171 that we only need element zero with the reduction result.
172 (reduc_plus_scal_v2df): Likewise.
174 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
176 * dojump.h (do_jump): Delete.
177 (do_jump_1): Likewise.
178 (split_comparison): Move around.
179 * dojump.c (do_jump): Make static.
180 (do_jump_1): Likewise.
181 (jumpifnot): Move around.
182 (jumpifnot_1): Likewise.
184 (jumpif_1): Likewise.
185 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
187 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
189 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
190 insns in the delay slot and add_insn_after for the jump insn.
192 2018-10-02 Richard Biener <rguenther@suse.de>
194 * tree-inline.c (expand_call_inline): Use the location of
195 the callee declaration for the inline-entry marker.
196 * final.c (notice_source_line): Remove special-casing of
197 NOTE_INSN_INLINE_ENTRY.
199 2018-10-01 Carl Love <cel@us.ibm.com>
202 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
203 (__builtin_mtfsb0): New.
204 (__builtin_mtfsb1): New.
205 ( __builtin_set_fpscr_rn): New.
206 (__builtin_set_fpscr_drn): New.
207 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
208 (rs6000_expand_set_fpscr_rn_builtin): Add.
209 (rs6000_expand_set_fpscr_drn_builtin): Add.
210 (rs6000_expand_builtin): Add case statement entries for
211 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
212 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
213 RS6000_BUILTIN_MFFSL.
214 (rs6000_init_builtins): Add ftype initialization and def_builtin
215 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
216 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
217 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
218 rs6000_mffscdrn): Add define_insn.
219 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
220 * doc/extend.texi: Add documentation for the builtins.
222 2018-10-01 Richard Biener <rguenther@suse.de>
224 PR tree-optimization/87465
225 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
226 causing branch miscounts.
228 2018-10-01 Tamar Christina <tamar.christina@arm.com>
230 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
231 aarch64_option_default_param): New.
233 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
234 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
235 stack-clash protection validation code.
237 2018-10-01 Tamar Christina <tamar.christina@arm.com>
239 * params.c (validate_param): New.
240 (add_params): Use it.
241 (set_param_value): Refactor param validation into validate_param.
242 (diagnostic.h): Include.
243 * diagnostic.h (diagnostic_ready_p): New.
245 2018-10-01 Tamar Christina <tamar.christina@arm.com>
247 * params.c (set_param_value):
248 Add index of parameter being validated.
249 * common/common-target.def (option_validate_param): New.
250 * common/common-targhooks.h (default_option_validate_param): New.
251 * common/common-targhooks.c (default_option_validate_param): New.
252 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
253 * doc/tm.texi: Regenerate.
255 2018-10-01 Tamar Christina <tamar.christina@arm.com>
258 * config/aarch64/aarch64.c (aarch64_override_options_internal):
259 Add validation for stack-clash parameters and set defaults.
261 2018-10-01 Tamar Christina <tamar.christina@arm.com>
264 * configure.ac: Add stack-clash-protection-guard-size.
265 * doc/install.texi: Document it.
266 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
267 * params.def: Update comment for guard-size.
268 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
269 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
270 * configure: Regenerate.
272 2018-10-01 Tamar Christina <tamar.christina@arm.com>
275 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
276 STACK_DYNAMIC_OFFSET): New.
277 * config/aarch64/aarch64.c (aarch64_layout_frame):
278 Update outgoing args size.
279 (aarch64_stack_clash_protection_alloca_probe_range,
280 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
282 2018-10-01 Tamar Christina <tamar.christina@arm.com>
285 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
287 * target.def (stack_clash_protection_alloca_probe_range): New.
288 (stack_clash_protection_final_dynamic_probe): Remove.
289 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
290 (default_stack_clash_protection_final_dynamic_probe): Remove.
291 * targhooks.c: Likewise.
292 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
293 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
294 * doc/tm.texi: Regenerate.
296 2018-10-01 Tamar Christina <tamar.christina@arm.com>
299 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
300 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
301 aarch64_clamp_to_uimm12_shift): New.
302 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
303 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
305 2018-10-01 Tamar Christina <tamar.christina@arm.com>
308 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
310 2018-10-01 Jeff Law <law@redhat.com>
311 Richard Sandiford <richard.sandiford@linaro.org>
312 Tamar Christina <tamar.christina@arm.com>
315 * config/aarch64/aarch64.md
316 (probe_stack_range): Add k (SP) constraint.
317 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
318 STACK_CLASH_MAX_UNROLL_PAGES): New.
319 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
320 stack probes for stack clash.
321 (aarch64_allocate_and_probe_stack_space): New.
322 (aarch64_expand_prologue): Use it.
323 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
324 (aarch64_sub_sp): Add emit_move_imm optional param.
326 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
328 PR tree-optimization/87261
329 * match.pd: Remove trailing whitespace.
330 Add (x & y) | ~(x | y) -> ~(x ^ y),
331 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
333 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
335 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
337 (ashlsi3_insn): Update instruction constraints.
338 (ashrsi3_insn): Likewise.
340 (add_shift): Likewise.
341 * config/arc/constraints.md (Csz): New 32 bit constraint. It
342 avoids placing in the limm field small constants which, otherwise,
343 could end into a small instruction.
345 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
347 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
348 destination register is not odd-even.
349 (umaddsidi4_split): Likewise.
351 2018-10-01 Richard Biener <rguenther@suse.de>
353 * tree-inline.c (expand_call_inline): Store origin of fn
354 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
355 * tree.c (block_ultimate_origin): Simplify and do some
358 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
360 * config/i386/mmx.md (EMMS): New int iterator.
361 (emms): New int attribute.
362 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
363 EMMS int iterator. Explicitly declare clobbers.
364 (mmx_emms): Remove expander.
366 * config/i386/predicates.md (emms_operation): Remove predicate.
367 (vzeroall_pattern): New predicate.
368 (vzeroupper_pattern): Rename from vzeroupper_operation.
369 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
370 vzeroupper_pattern and vzeroall_pattern predicates.
372 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
374 PR rtl-optimization/86939
375 * ira-lives.c (make_hard_regno_born): Rename from this...
376 (make_hard_regno_live): ... to this. Remove update to conflict
377 information. Update function comment.
378 (make_hard_regno_dead): Add conflict information update. Update
380 (make_object_born): Rename from this...
381 (make_object_live): ... to this. Remove update to conflict information.
382 Update function comment.
383 (make_object_dead): Add conflict information update. Update function
385 (mark_pseudo_regno_live): Call make_object_live.
386 (mark_pseudo_regno_subword_live): Likewise.
387 (mark_hard_reg_dead): Update function comment.
388 (mark_hard_reg_live): Call make_hard_regno_live.
389 (process_bb_node_lives): Likewise.
390 * lra-lives.c (make_hard_regno_born): Rename from this...
391 (make_hard_regno_live): ... to this. Remove update to conflict
392 information. Remove now uneeded check_pic_pseudo_p argument.
393 Update function comment.
394 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
395 to conflict information. Update function comment.
396 (mark_pseudo_live): Remove update to conflict information. Update
398 (mark_pseudo_dead): Add conflict information update.
399 (mark_regno_live): Call make_hard_regno_live.
400 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
401 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
403 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
406 * config/i386/i386.c (construct_container): Use TImode for
407 BLKmode values in 2 integer registers.
409 2018-09-29 Jeff Law <law@redhat.com>
411 * builtins.c (unterminated_array): Pass in c_strlen_data * to
412 c_strlen rather than just a tree *.
413 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
414 Update recursive calls appropriately. If caller did not provide a
415 suitable data pointer, create a local one. When a non-terminated
416 string is discovered, bubble up information about the string via the
417 c_strlen_data object.
418 * builtins.h (c_strlen): Update prototype.
419 (c_strlen_data): New structure.
420 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
421 For a type 2 call, if c_strlen indicates a non-terminated string
422 use the length of the non-terminated string.
423 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
425 2018-09-29 Jakub Jelinek <jakub@redhat.com>
428 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
429 __m512d type for __A argument rather than __m512.
431 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
433 * match.pd (simple_comparison): Don't optimize if either operand is
434 a function pointer when target needs function pointer canonicalization.
436 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
438 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
439 power5 .. power9 to remove indirection.
440 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
441 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
442 ASM_CPU_476_SPEC): Delete.
443 (ASM_CPU_SPEC): Adjust.
444 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
445 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
447 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
449 * config.in: Delete HAVE_AS_DCI.
450 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
451 * config/rs6000/rs6000.h: Ditto.
452 * configure.ac: Delete HAVE_AS_DCI.
453 * configure: Regenerate.
455 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
457 * config.in (HAVE_AS_LWSYNC): Delete.
458 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
459 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
461 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
462 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
464 * configure.ac: Delete HAVE_AS_LWSYNC.
465 * configure: Regenerate.
467 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
468 Pierre-Marie de Rodat <derodat@adacore.com>
470 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
471 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
472 (cgraph_node::create_thunk): Add indirect_offset parameter.
473 (thunk_adjust): Likewise.
474 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
475 and initialize the corresponding field with it.
476 (cgraph_node::dump): Dump indirect_offset field.
477 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
478 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
479 (thunk_adjust): Add indirect_offset parameter and deal with it.
480 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
481 pass it to thunk_adjust. Do not call the target hook if it's non-zero
482 or if the thunk is external or local. Fix formatting. Do not chain
483 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
484 if any, in the GIMPLE representation.
485 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
486 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
487 (input_node): Read indirect_offset field.
488 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
489 call to thunk_adjust.
490 * tree-nested.c (struct nesting_info): Add thunk_p field.
491 (create_nesting_tree): Set it.
492 (convert_all_function_calls): Copy static chain from targets to thunks.
493 (finalize_nesting_tree_1): Return early for thunks.
494 (unnest_nesting_tree_1): Do not finalize thunks.
495 (gimplify_all_functions): Do not gimplify thunks.
497 2018-09-28 David Malcolm <dmalcolm@redhat.com>
499 * opt-suggestions.c (option_proposer::build_option_suggestions):
500 Release "option_values".
502 2018-09-28 David Malcolm <dmalcolm@redhat.com>
504 * coverage.c (get_coverage_counts): Convert problem-reporting dump
505 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
506 * dumpfile.c (kind_as_string): New function.
507 (dump_loc): Rather than a hardcoded prefix of "note: ", use
508 kind_as_string to vary the prefix based on dump_kind.
509 (selftest::test_capture_of_dump_calls): Update for above.
511 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
513 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
514 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
516 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
518 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
519 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
520 INVALID_REGNUM instead of FPSR_REG.
521 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
522 * config/i386/i386.md: Update comment of FP compares.
523 (fldenv): Do not clobber FPSR_REG.
525 2018-09-28 Richard Biener <rguenther@suse.de>
527 * tree.h (BLOCK_ORIGIN): New.
528 * omp-expand.c (grid_expand_target_grid_body): Assign
529 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
530 * tree-inline.c (remap_block): Likewise.
531 * auto-profile.c (get_function_decl_from_block): Simplify
532 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
533 * langhooks.c (lhd_print_error_function): Likewise.
534 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
536 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
537 * tree.c (block_nonartificial_location): Likewise.
538 (block_ultimate_origin): Likewise.
539 * tree-pretty-print.c (percent_K_format): Likewise. Remove
540 no longer needed LTO case.
542 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
543 Jan Hubicka <jh@suse.cz>
544 Martin Jambor <mjambor@suse.cz>
546 * simplify-rtx.c (simplify_merge_mask): New function.
547 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
548 same masks are used in op1 or op2.
549 (test_vec_merge): New function.
550 (test_vector_ops): Call test_vec_merge.
552 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
554 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
555 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
556 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
557 (cypress_costs): Set it.
558 (supersparc_costs): Likewise.
559 (hypersparc_costs): Likewise.
560 (leon_cost): Likewise.
561 (leon3_costs): Likewise.
562 (sparclet_costs): Likewise.
563 (ultrasparc_costs): Likewise.
564 (ultrasparc_costs): Likewise.
565 (niagara_costs): Likewise.
566 (niagara2_costs): Likewise.
567 (niagara3_costs): Likewise.
568 (niagara4_costs): Likewise.
569 (niagara7_costs): Likewise.
570 (m8_costs): Likewise.
571 (TARGET_CAN_FOLLOW_JUMP): Define.
572 (pass_work_around_errata::gate): Minor tweak.
573 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
574 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
575 Automaitcally clear MASK_FSMULD mask for V7 processors.
576 (sparc_can_follow_jump): New static function.
577 (output_ubranch): Deal with CROSSING_JUMP_P.
578 (sparc_use_sched_lookahead): Rewrite using switch statement.
579 (sparc_issue_rate): Reorder.
580 (sparc_branch_cost): New function.
582 2018-09-27 Martin Sebor <msebor@redhat.com>
584 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
585 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
586 (int_fits_type_p): Same.
588 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
590 * config/i386/i386.md (FPCR_REG): Remove.
591 (UNSPEC_FLDCW): Remove.
592 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
593 (x86_fldcw_1): Remove insn pattern.
594 (fnstenv): Do not clobber FPCR_REG.
596 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
597 (CALL_USED_REGISTERS): Ditto.
598 (REG_ALLOC_ORDER): Ditto.
599 (REG_CLASS_CONTENTS): Ditto.
600 (HI_REGISTER_NAMES): Ditto.
601 (ADDITIONAL_REGISTER_NAMES): Use defines instead
602 of numerical constants.
603 * config/i386/i386.c (regclass_map): Remove fpsr register.
604 (dbx_register_map): Ditto.
605 (dbx64_register_map): Ditto.
606 (svr4_dbx_register_map): Ditto.
607 (print_reg): Do not handle FPCR_REG.
609 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
612 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
613 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
614 Delete, always treat as true.
615 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
616 Ditto. Simplify remaining code.
617 * config/powerpcspe/powerpcspe.h: Ditto.
618 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
619 Simplify remaining code.
620 (rs6000_expand_builtin): Ditto.
621 * config/rs6000/rs6000.h: Ditto.
622 * configure.ac: Ditto.
623 * configure: Regenerate.
625 2018-09-27 Martin Liska <mliska@suse.cz>
627 * coverage.c (get_coverage_counts): Revert the formatting
628 of missing profile opt info.
630 2018-09-27 Richard Biener <rguenther@suse.de>
634 * dwarf2out.c (set_block_origin_self): Do not mark outermost
635 block as we do not output that.
636 (gen_inlined_subroutine_die): Elide the originally outermost
637 block, matching what we do for concrete instances.
638 (decls_for_scope): Add parameter specifying whether to recurse
641 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
642 Tom de Vries <tom@codesourcery.com>
646 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
647 STORE_FLAG_VALUE == 1.
649 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
651 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
652 constant definitions.
653 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
654 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
655 ("speculation_barrier"): New expander definition.
657 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
659 PR gcov-profile/86957
660 * common.opt: New warning option -Wmissing-profile.
661 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
662 * doc/invoke.texi: Document -Wmissing-profile.
664 2018-09-26 Jim Wilson <jimw@sifive.com>
666 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
667 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
669 2018-09-26 Martin Sebor <msebor@redhat.com>
671 * tree.c (zerop): Change return type to bool.
672 (integer_zerop, integer_onep, integer_each_onep): Same.
673 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
674 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
675 (real_onep, real_minus_onep, chain_index): Same.
676 (print_type_hash_statistics, type_list_equal): Same.
677 * tree.h (zerop): Same.
678 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
679 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
680 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
681 (real_onep, real_minus_onep, chain_index): Same.
682 (print_type_hash_statistics, type_list_equal): Same.
684 2018-09-26 Jim Wilson <jimw@sifive.com>
686 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
688 2018-09-26 Jakub Jelinek <jakub@redhat.com>
691 * config/i386/i386.c: Include debug.h and dwarf2out.h.
692 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
695 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
697 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
699 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
701 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
702 and FP_SECOND_SSE_REGS.
703 (REG_CLASS_NAMES): Ditto.
704 (REG_CLASS_CONTENTS): Ditto.
705 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
706 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
707 (ix86_preferred_output_reload_class): Ditto.
708 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
709 clobber constraint to "=&f".
710 (fix_truncdi_i387): Ditto.
712 (fistdi2_<rounding>): Ditto.
713 (fpremxf4_i387): Change "=u" constraint to "=f".
714 (fprem1xf4_i387): Ditto.
716 (fptanxf4_i387): Ditto.
717 (fxtractxf3_i387): Ditto.
718 (fscalexf4_i387): Ditto.
719 (atan2xf3): Change "u" constraint to "f".
720 (fyl2xxf3_i387): Ditto.
721 (fyl2xp1xf3_i387): Ditto.
723 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
726 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
727 for removed I387_MASK_PM entity.
730 2018-09-26 Jeff Law <law@redhat.com>
732 2018-09-26 Alexey Neyman <stilor@att.net>
734 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
735 headers are no longer pulled in by <isl/val.h>.
737 2018-09-26 Richard Biener <rguenther@suse.de>
740 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
741 or concrete instance DIE to the tree. Create abstract origin
742 attributes also for concrete instances.
744 2018-09-26 Alexey Neyman <stilor@att.net>
746 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
747 headers are no longer pulled in by <isl/val.h>.
749 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
751 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
752 Use new helper functions.
753 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
754 Use new helper functions.
755 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
756 aarch_mm_needs_release): New declarations.
757 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
758 aarch_mm_needs_release): New.
760 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
762 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
763 (arm32_output_mi_thunk): Deal with long calls.
765 2018-09-26 Richard Biener <rguenther@suse.de>
769 * tree-inline.c (expand_call_inline): When the location
770 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
771 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
772 the inserted BLOCK to make inlined_function_outer_scope_p
774 * dwarf2out.c (add_call_src_coords_attributes): Do not add
775 coords for reserved locations.
777 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
779 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
780 (*call_indirect_nonlocal_sysv<mode>): Ditto.
781 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
782 (*sibcall_nonlocal_sysv<mode>): Ditto.
783 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
784 (<bd>_<mode>): Ditto.
785 (<bd>tf_<mode>): Ditto.
787 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
789 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
790 control string as a list of templates instead of as C code.
791 (*altivec_movti): Ditto.
792 * config/rs6000/darwin.md (movdf_low_di): Ditto.
794 2018-09-25 Jim Wilson <jimw@sifive.com>
796 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
797 when target symbol is weak.
799 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
802 * builtins.c (unterminated_array): Simplify.
803 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
804 where pointer arithmetic is safe.
806 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
809 * config/rs6000/altivec.md (altivec_vspltb): Use
810 const_0_to_15_operand instead of u5bit_cint_operand.
811 (*altivec_vspltb_internal): Ditto.
812 (altivec_vspltb_direct): Ditto.
813 (altivec_vsplth): Use const_0_to_7_operand instead of
815 (*altivec_vsplth_internal): Ditto.
816 (altivec_vsplth_direct): Ditto.
817 (altivec_vspltw): Use const_0_to_3_operand instead of
819 (*altivec_vspltw_internal): Ditto.
820 (altivec_vspltw_direct): Ditto.
821 (altivec_vspltsf): Ditto.
822 (*altivec_vspltsf_internal): Ditto.
823 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
824 various splats with the proper size immediate. Reorder the various
825 cases by ascending size of immediate, and put all such together.
827 2018-09-25 Richard Biener <rguenther@suse.de>
830 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
832 (maybe_create_die_with_external_ref): Do not create
833 DW_TAG_imported_unit here.
834 (add_abstract_origin_attribute): Handle external BLOCK refs.
835 (dwarf2out_abstract_function): Simplify LTO case.
836 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
837 rather than using maybe_create_die_with_external_ref.
839 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
842 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
843 (frndintxf2_mask_pm_i387): Ditto.
844 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
845 Enable for !flag_trapping_math.
846 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
847 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
848 Change operand 1 predicate to nonimmediate_operand.
849 (attr "i387_cw"): Remove mask_pm.
850 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
851 (enum ix86_entity): Remove I387_MASK_PM.
852 * config/i386/i386.c (ix86_i387_mode_needed): Do not
854 (ix86_mode_needed): Ditto.
855 (ix86_mode_after): Ditto.
856 (ix86_mode_entry): Ditto.
857 (ix86_mode_exit): Ditto.
858 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
860 2018-09-25 Jakub Jelinek <jakub@redhat.com>
862 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
863 to_update_switch_stmts to vNULL instead of calling create on them
866 2018-09-25 Richard Biener <rguenther@suse.de>
868 PR tree-optimization/87402
869 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
870 (visit_phi): Re-instantiate handling of supposed to be VARYING
871 but non-VARYING backedge value.
873 2018-09-25 Richard Biener <rguenther@suse.de>
876 * dwarf2out.c (struct sym_off_pair): New.
877 (external_die_map): New global.
878 (lookup_decl_die): When in LTO create DIEs lazily from the
880 (lookup_block_die): New function, create DIEs lazily in LTO.
881 (equate_block_to_die): New function.
882 (dwarf2out_die_ref_for_decl): During WPA get the association
883 from the external DIE map.
884 (dwarf2out_register_external_die): Record mapping into the
886 (maybe_create_die_with_external_ref): New function split out from
887 DIE generation part of old dwarf2out_register_external_die.
888 (add_abstract_origin_attribute): Do not return the DIE. When
889 in LTO reference externals directly.
890 (dwarf2out_abstract_function): When in LTO ignore calls for
891 decls with external DIEs (already present abstract instances).
892 (gen_call_site_die): Adjust.
893 (add_high_low_attributes): Likewise.
894 (gen_lexical_block_die): Likewise.
895 (gen_inlined_subroutine_die): Likewie.
896 (gen_block_die): Likewise.
897 (dwarf2out_inline_entry): Likewise.
898 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
901 2018-09-25 Martin Liska <mliska@suse.cz>
903 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
904 integers and not by a float value.
906 2018-09-25 Martin Liska <mliska@suse.cz>
909 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
911 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
912 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
913 * godump.c (go_format_type): Remove extra parenthesis.
915 2018-09-25 Martin Liska <mliska@suse.cz>
917 * alias.c (set_dest_equal_p): Remove unused function.
918 * config/i386/i386.c (def_builtin_pure2): Likewise.
919 * diagnostic-show-locus.c (class layout): Remove
921 (layout::layout): Likewise here.
922 * dump-context.h (class temp_dump_context): Likewise.
923 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
924 (add_AT_loclistsptr): Likewise.
925 (add_AT_offset): Likewise.
926 (get_AT_hi_pc): Likewise.
927 (is_comdat_die): Likewise.
928 (type_is_enum): Likewise.
930 (add_AT_vms_delta): Likewise.
931 (is_class_die): Likewise.
932 * edit-context.c (class line_event): Remove unused field.
933 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
935 * ipa-cp.c (ipa_get_vr_lat): Likewise.
936 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
937 (ok_for_base_p_nonstrict): Likewise.
938 * tree-chrec.c (is_not_constant_evolution): Likewise.
939 (chrec_fold_poly_cst): Likewise.
940 * tree-if-conv.c (has_pred_critical_p): Likewise.
941 * tree-ssa-coalesce.c (print_exprs): Likewise.
942 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
943 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
944 * tree-vrp.c (value_ranges_intersect_p): Likewise.
945 (value_range_nonnegative_p): Likewise.
947 2018-09-25 Martin Liska <mliska@suse.cz>
949 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
950 Do not handle "GNU Pascal".
951 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
953 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
954 from documentation. Likewise.
955 * dbxout.c (dbxout_range_type): Likewise.
956 * doc/cpp.texi: Likewise.
957 * doc/extend.texi: Likewise.
958 * doc/frontends.texi: Likewise.
959 * doc/invoke.texi: Remove Pascal entry.
960 * tree.def (CLEANUP_POINT_EXPR): Likewise.
961 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
963 2018-09-25 Martin Liska <mliska@suse.cz>
966 * doc/invoke.texi: Document all parameters and remove default
969 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
972 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
973 contains HOST_WIDE_INTs when computing its size.
975 2018-09-24 Jim Wilson <jimw@sifive.com>
978 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
980 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
982 2018-09-24 Andrew Pinski <apinski@marvell.com>
984 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
985 access prev before checking it for NULLness in the
986 AARCH64_FUSE_CMP_BRANCH case.
988 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
991 * config/i386/i386.c (rest_of_insert_endbranch): Set
992 endbr_queued_at_entrance to true and don't insert ENDBR if
993 x86_function_profiler will be called.
994 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
996 * config/i386/i386.h (machine_function): Add
997 endbr_queued_at_entrance.
999 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1001 * genattrtab.c (mk_attr_alt): Use alternative_mask.
1002 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1004 (check_attr_test): Use alternative_mask.
1005 (get_attr_value): Likewise.
1006 (compute_alternative_mask): Use alternative_mask and XWINT.
1007 (make_alternative_compare): Use alternative_mask.
1008 (attr_alt_subset_p): Use XWINT.
1009 (attr_alt_subset_of_compl_p): Likewise.
1010 (attr_alt_intersection): Use alternative_mask and XWINT.
1011 (attr_alt_union): Likewise.
1012 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1013 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1014 (simplify_test_exp): Use alternative_mask and XWINT.
1015 (write_test_expr): Use alternative_mask and XWINT, adjust bit
1016 number calculation to support 64 bits. Generate code that
1017 checks 64-bit masks.
1018 (main): Use alternative_mask.
1019 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1021 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1024 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1025 RETURN+USE when returning via %r14.
1027 2018-09-24 Martin Liska <mliska@suse.cz>
1029 * gcov.c (output_lines): Print colorization legend
1030 for both flag_use_colors and flag_use_hotness_colors.
1033 2018-09-24 Martin Liska <mliska@suse.cz>
1035 * coverage.c (get_coverage_counts): Use warning_at
1036 with current_function_decl location. Use %qD in warning
1039 2018-09-24 Martin Liska <mliska@suse.cz>
1041 * memory-block.h (memory_block_pool::release): Annotate with
1042 valgrind that the memory is not accessible.
1044 2018-09-24 Martin Liska <mliska@suse.cz>
1047 * asan.c: Make asan_handled_variables extern.
1049 * cfgexpand.c (expand_stack_vars): Make sure
1050 a representative is unpoison if another
1051 variable in the partition is handled by
1052 use-after-scope sanitization.
1054 2018-09-24 Richard Biener <rguenther@suse.de>
1056 PR tree-optimization/63155
1057 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1058 the worklist when the edge of the respective argument isn't
1061 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1063 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1064 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1065 (MASK_CLASS_P): Update for rename.
1066 (MAYBE_MASK_CLASS_P): Ditto.
1067 (REG_CLASS_NAMES): Update.
1068 (REG_CLASS_CONTENT): Update.
1069 * config/i386/i386.c (regclass_map): Update for MASK_REG
1070 and ALL_MASK_REGS rename.
1071 * config/i386/constraints.md (Yk): Update for rename.
1074 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1076 * config/i386/i386.h (enum reg_class): Remove
1077 EVEX_SSE_REGS and MOD4_SSE_REGS.
1078 (REG_CLASS_NAMES): Update.
1079 (REG_CLASS_CONTENT): Update.
1080 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1081 registers as ALL_SSE_REGS.
1082 (ix86_additional_allocno_class_p): Remove.
1083 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1084 (ix86_register_priority): Lower priority of EVEX SSE registers.
1085 Use IN_RANGE macro where appropriate.
1086 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1087 AVX-5124VNNIW checks.
1088 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1089 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1090 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1091 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1092 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1093 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1094 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1095 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1096 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1097 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1098 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1099 * config/i386/constraints.md (Yh): Remove.
1101 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1103 * config/i386/i386.c (regclass_map): Declare integer REX registers
1106 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
1108 * doc/service.texi (Service): Switch the fsf.org link to https.
1110 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
1113 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1114 Define to speculation_safe_value_not_needed.
1116 2018-09-21 Florian Weimer <fweimer@redhat.com>
1119 * doc/extend.texi (Common Function Attributes): Mention that
1120 noreturn suppresses tail call optimization.
1122 2018-09-21 Jeff Law <law@redhat.com>
1124 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1125 vr_values::cleanup_edges_and_switches.
1126 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1128 (identify_jump_threads): Remove EDGE_IGNORE handling.
1129 (execute_vrp): Move handling of to_remove_edges and
1130 to_update_switch_stmts into vr_values class member functions.
1131 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1132 (to_update_switch_stmts): Likewise.
1133 * vr-values.c: Include cfghooks.h.
1134 (vr_values::vr_values): Initialize to_remove_edges and
1135 to_update_switch_stmts.
1136 (vr_values::~vr_values): Verify to_remove_edges and
1137 to_update_switch_stmts are empty.
1138 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1139 (vr_values::cleanup_edges_and_switches): New member function.
1140 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1141 function. Add new data members.
1143 2018-09-21 David Malcolm <dmalcolm@redhat.com>
1145 PR tree-optimization/87309
1146 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1147 calls with pflags and alt_flags.
1148 (selftest::test_capture_of_dump_calls): Add test of interaction of
1149 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1151 2018-09-21 Olivier Hainque <hainque@adacore.com>
1153 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1155 2018-09-21 Olivier Hainque <hainque@adacore.com>
1157 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1159 2018-09-21 Olivier Hainque <hainque@adacore.com>
1161 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1162 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1163 (VXWORKS_LIBS_RTP): Minor reordering.
1165 2018-09-21 Olivier Hainque <hainque@adacore.com>
1167 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1168 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1170 2018-09-21 Olivier Hainque <hainque@adacore.com>
1172 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1173 (PTRDIFF_TYPE): Likewise.
1175 2018-09-21 Olivier Hainque <hainque@adacore.com>
1177 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1178 triplet, similar to support for VxWorks7.
1179 * config/vxworks-dummy.h: Provide a default definition
1180 of TARGET_VXWORKS64 to 0.
1182 2018-09-21 Olivier Hainque <hainque@adacore.com>
1184 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1185 * config/vxworks-dummy.h: here.
1187 2018-09-21 Olivier Hainque <hainque@adacore.com>
1189 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1191 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
1192 Bo Zhou <zbo.zhou@hisilicon.com>
1194 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1195 * config/aarch64/aarch64-tune.md: Regenerated.
1196 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1197 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1198 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1200 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
1201 Julian Brown <julian@codesourcery.com>
1203 * builtins.c (get_builtin_sync_mem): Handle address spaces.
1205 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1207 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1208 if the call takes a static chain.
1210 2018-09-21 Martin Liska <mliska@suse.cz>
1212 * auto-profile.c (autofdo_source_profile::read): Do not
1214 (read_profile): Do not add working sets.
1215 (read_autofdo_file): Remove sum_all.
1216 (afdo_callsite_hot_enough_for_early_inline): Remove const
1218 * coverage.c (struct counts_entry): Remove gcov_summary.
1219 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1220 do not support GCOV_TAG_PROGRAM_SUMMARY.
1221 (get_coverage_counts): Remove summary and expected
1223 * coverage.h (get_coverage_counts): Likewise.
1224 * doc/gcov-dump.texi: Remove -w option.
1225 * gcov-dump.c (dump_working_sets): Remove.
1226 (main): Do not support '-w' option.
1227 (print_usage): Likewise.
1228 (tag_summary): Likewise.
1229 * gcov-io.c (gcov_write_summary): Do not dump
1231 (gcov_read_summary): Likewise.
1232 (gcov_histo_index): Remove.
1233 (gcov_histogram_merge): Likewise.
1234 (compute_working_sets): Likewise.
1235 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1237 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1238 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1239 (GCOV_HISTOGRAM_SIZE): Remove.
1240 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1241 (struct gcov_summary): Simplify rapidly just
1242 to runs and sum_max fields.
1243 (gcov_histo_index): Remove.
1244 (NUM_GCOV_WORKING_SETS): Likewise.
1245 (compute_working_sets): Likewise.
1246 * gcov-tool.c (print_overlap_usage_message): Remove
1247 trailing empty line.
1248 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1249 (output_lines): Remove program related line.
1250 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1251 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1253 (input_profile_summary): Do not read it.
1254 (merge_profile_summaries): And do not merge it.
1255 (input_symtab): Do not call removed function.
1256 * modulo-sched.c (sms_schedule): Do not print sum_max.
1257 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1258 removed when histogram method was invented.
1259 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1261 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1263 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1264 and dump selected value.
1265 * profile.c (add_working_set): Remove.
1266 (get_working_sets): Likewise.
1267 (find_working_set): Likewise.
1268 (get_exec_counts): Do not work with working sets.
1269 (read_profile_edge_counts): Do not inform as sum_max is removed.
1270 (compute_branch_probabilities): Likewise.
1271 (compute_value_histograms): Remove argument for call of
1272 get_coverage_counts.
1273 * profile.h: Do not make gcov_summary const.
1275 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
1277 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1279 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1281 PR tree-optimization/86990
1282 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1283 Check that the entire merged store group is made of constants only for
1286 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
1288 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1289 (VTABLE_VERIFICATION_SPEC): Likewise.
1290 (SANITIZER_EARLY_SPEC): Likewise.
1291 (SANITIZER_SPEC): Likewise.
1292 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1293 * doc/invoke.texi (Link Options): Document -r.
1295 2018-09-20 Richard Biener <rguenther@suse.de>
1298 * gimplify.c (gimplify_expr): Retain alignment of
1299 addressable lvalue in dereference.
1301 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
1304 * configure.ac: Check for .loc is_stmt support.
1305 * configure, config.in: Rebuilt.
1306 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1309 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1311 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1312 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1315 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1317 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1319 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1322 2018-09-20 Marek Polacek <polacek@redhat.com>
1324 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1326 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1328 PR tree-optimization/87288
1329 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1330 into account when determining PEELING_FOR_NITERS.
1332 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1334 PR tree-optimization/86877
1335 * tree-vect-loop.c (vect_analyze_loop_2): Call
1336 vect_verify_datarefs_alignment.
1338 2018-09-19 Marek Polacek <polacek@redhat.com>
1340 * doc/invoke.texi: Document -Wclass-conversion.
1342 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
1344 * config/pa/pa.c (pa_adjust_priority): Delete.
1345 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
1347 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
1348 (atomic_storehi): Likewise.
1349 (atomic_storesi): Likewise.
1350 (atomic_loaddi): Restore compare and swap exchange loop code.
1352 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
1354 PR rtl-optimization/86902
1355 * combine.c (try_combine): When changing the CC mode used, don't change
1356 an unrelated mode in other_insn to that new CC mode.
1358 2018-09-19 David Malcolm <dmalcolm@redhat.com>
1360 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
1361 with %T in place of calls to dump_generic_expr.
1362 (prune_runtime_alias_test_list): Likewise.
1363 (create_runtime_alias_checks): Likewise.
1364 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
1365 (vect_analyze_data_ref_dependence): Likewise.
1366 (vect_slp_analyze_data_ref_dependence): Likewise.
1367 (vect_record_base_alignment): Likewise. Use %G in place of call
1368 to dump_gimple_stmt.
1369 (vect_compute_data_ref_alignment): Likewise.
1370 (verify_data_ref_alignment): Likewise.
1371 (vect_find_same_alignment_drs): Likewise.
1372 (vect_analyze_group_access_1): Likewise.
1373 (vect_analyze_data_ref_accesses): Likewise.
1374 (dependence_distance_ge_vf): Likewise.
1375 (dump_lower_bound): Likewise.
1376 (vect_prune_runtime_alias_test_list): Likewise.
1377 (vect_find_stmt_data_reference): Likewise.
1378 (vect_analyze_data_refs): Likewise.
1379 (vect_create_addr_base_for_vector_ref): Likewise.
1380 (vect_create_data_ref_ptr): Likewise.
1381 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
1382 (vect_can_advance_ivs_p): Likewise.
1383 (vect_update_ivs_after_vectorizer): Likewise.
1384 (vect_gen_prolog_loop_niters): Likewise.
1385 (vect_prepare_for_masked_peels): Likewise.
1386 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1387 (vect_determine_vectorization_factor): Likewise.
1388 (vect_is_simple_iv_evolution): Likewise.
1389 (vect_analyze_scalar_cycles_1): Likewise.
1390 (vect_analyze_loop_operations): Likewise.
1391 (report_vect_op): Likewise.
1392 (vect_is_slp_reduction): Likewise.
1393 (check_reduction_path): Likewise.
1394 (vect_is_simple_reduction): Likewise.
1395 (vect_create_epilog_for_reduction): Likewise.
1396 (vect_finalize_reduction:): Likewise.
1397 (vectorizable_induction): Likewise.
1398 (vect_transform_loop_stmt): Likewise.
1399 (vect_transform_loop): Likewise.
1400 (optimize_mask_stores): Likewise.
1401 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
1402 (vect_split_statement): Likewise.
1403 (vect_recog_over_widening_pattern): Likewise.
1404 (vect_recog_average_pattern): Likewise.
1405 (vect_determine_min_output_precision_1): Likewise.
1406 (vect_determine_precisions_from_range): Likewise.
1407 (vect_determine_precisions_from_users): Likewise.
1408 (vect_mark_pattern_stmts): Likewise.
1409 (vect_pattern_recog_1): Likewise.
1410 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1411 (vect_record_max_nunits): Likewise.
1412 (vect_build_slp_tree_1): Likewise.
1413 (vect_build_slp_tree_2): Likewise.
1414 (vect_print_slp_tree): Likewise.
1415 (vect_analyze_slp_instance): Likewise.
1416 (vect_detect_hybrid_slp_stmts): Likewise.
1417 (vect_detect_hybrid_slp_1): Likewise.
1418 (vect_slp_analyze_operations): Likewise.
1419 (vect_slp_analyze_bb_1): Likewise.
1420 (vect_transform_slp_perm_load): Likewise.
1421 (vect_schedule_slp_instance): Likewise.
1422 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1423 (vect_mark_stmts_to_be_vectorized): Likewise.
1424 (vect_init_vector_1): Likewise.
1425 (vect_get_vec_def_for_operand): Likewise.
1426 (vect_finish_stmt_generation_1): Likewise.
1427 (vect_check_load_store_mask): Likewise.
1428 (vectorizable_call): Likewise.
1429 (vectorizable_conversion): Likewise.
1430 (vectorizable_operation): Likewise.
1431 (vectorizable_load): Likewise.
1432 (vect_analyze_stmt): Likewise.
1433 (vect_is_simple_use): Likewise.
1434 (vect_get_vector_types_for_stmt): Likewise.
1435 (vect_get_mask_type_for_stmt): Likewise.
1436 * tree-vectorizer.c (increase_alignment): Likewise.
1438 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
1440 * doc/rtl.texi: Adjust vec_select description.
1441 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
1442 non-constant selectors.
1444 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
1446 * config/aarch64/aarch64-protos.h
1447 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
1448 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
1449 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
1450 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
1451 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
1452 (AARCH64_FL_PROFILE): Move index so flags are ordered.
1453 (AARCH64_ISA_RCPC8_4): New flag.
1454 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
1455 to aarch64_offset_9bit_signed_unscaled_p.
1456 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
1458 * config/aarch64/constraints.md (Ust): New constraint.
1459 * config/aarch64/predicates.md.
1460 (aarch64_9bit_offset_memory_operand): New predicate.
1461 (aarch64_rcpc_memory_operand): New predicate.
1463 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
1465 PR rtl-optimization/87361
1466 * rtlanal.c (nonzero_bits1): Revert accidental change.
1468 2018-09-19 Richard Biener <rguenther@suse.de>
1470 PR tree-optimization/87349
1471 PR tree-optimization/87342
1472 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
1474 2018-09-18 Marek Polacek <polacek@redhat.com>
1476 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
1477 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
1479 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
1481 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
1483 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
1485 PR rtl-optimization/86882
1486 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
1488 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
1490 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
1491 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
1493 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
1496 * doc/invoke.texi (Link Options): Fix formatting and grammar.
1498 2018-09-18 Richard Biener <rguenther@suse.de>
1501 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
1502 (compute_samebase_partition_bases): Likewise.
1503 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
1504 (gimple_can_coalesce_p): Simplify.
1506 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
1508 Handle a library implementation of ffs calling __builtin_ffs.
1509 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
1510 (mmix_init_libfuncs): New function: make __builtin_ffs expand
1513 2018-09-17 David Malcolm <dmalcolm@redhat.com>
1515 * diagnostic-show-locus.c (class layout_range): Add field
1517 (layout_range::layout_range): Add "original_idx" param and use it
1518 to initialize new field.
1519 (make_range): Use 0 for original_idx.
1520 (layout::layout): Pass in index to calls to
1521 maybe_add_location_range.
1522 (layout::maybe_add_location_range): Add param "original_idx" and
1523 pass it on to layout_range.
1524 (layout::print_any_labels): Pass on range->m_original_idx to
1526 (gcc_rich_location::add_location_if_nearby): Use 0 for
1528 * gcc-rich-location.h (text_range_label::get_text): Update for new
1530 (range_label_for_type_mismatch::get_text): Likewise.
1532 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
1534 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
1536 2018-09-17 David Malcolm <dmalcolm@redhat.com>
1538 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
1539 format_string_diagnostic_t.
1540 (fmtwarn_n): Likewise.
1541 * substring-locations.c
1542 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
1543 (format_warning_n_va): Convert to...
1544 (format_string_diagnostic_t::emit_warning_n_va): ...this.
1545 (format_warning_va): Convert to...
1546 (format_string_diagnostic_t::emit_warning_va): ...this.
1547 (format_warning_at_substring): Convert to...
1548 (format_string_diagnostic_t::emit_warning): ...this.
1549 (format_warning_at_substring_n): Convert to...
1550 (format_string_diagnostic_t::emit_warning_n): ...this.
1551 * substring-locations.h (class format_string_diagnostic_t): New
1553 (format_warning_va): Convert to
1554 format_string_diagnostic_t::emit_warning_va.
1555 (format_warning_n_va): Convert to
1556 format_string_diagnostic_t::emit_warning_n_va.
1557 (format_warning_at_substring): Convert to
1558 format_string_diagnostic_t::emit_warning.
1559 (format_warning_at_substring_n): Convert to
1560 format_string_diagnostic_t::emit_warning_n.
1562 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
1563 Bernd Schmidt <bernds_cb1@t-online.de>
1565 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
1568 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
1570 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
1571 operand 0 predicate to nonimmediate operand.
1572 (rint<mode>2_frndint): Remove insn pattern.
1573 (rint<mode>2): Change operand 1 predicate to general_operand.
1574 Extend operand 1 to XFmode and generate rintxf2 insn.
1575 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
1576 Do not use X87MODEF mode macro.
1577 (frndintxf2_<rounding>_i387): Rename from
1578 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
1579 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
1580 to XFmode and generate significandxf3 insn.
1582 2018-09-17 Richard Biener <rguenther@suse.de>
1584 PR tree-optimization/87328
1585 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
1586 visiting unexecutable backedges when not iterating.
1587 (do_rpo_vn): Mark all edges not executable even when not
1590 2018-09-17 Martin Jambor <mjambor@suse.cz>
1593 * doc/invoke.texi (Warning Options): Likewise.
1595 2018-09-17 Richard Biener <rguenther@suse.de>
1597 PR tree-optimization/87301
1598 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
1599 clean EH info from leftover copy assignments.
1601 2018-09-17 Martin Liska <mliska@suse.cz>
1603 PR gcov-profile/85871
1604 * gcov.c (output_intermediate_file): Fix out of bounds
1607 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
1609 * config/arc/arc.c: Object attributes for core4 not reflected
1611 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
1614 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
1616 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
1618 2018-09-17 Martin Liska <mliska@suse.cz>
1620 * doc/gcov.texi: Document new option --use-hotness-colors.
1621 * gcov.c (struct source_info): Declare new field.
1622 (source_info::source_info): Set default for maximum_count.
1623 (print_usage): Add new -q option.
1624 (process_args): Process it.
1625 (accumulate_line_info): Save src->maximum_count.
1626 (output_line_beginning): Make color line number if
1627 flag_use_hotness_colors is set.
1628 (output_line_details): Pass default argument value.
1629 (output_lines): Pass src->maximum_count.
1631 2018-09-17 Martin Liska <mliska@suse.cz>
1633 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
1634 Use processor_names table.
1635 * config/i386/i386.c (ix86_default_align): Use
1636 processor_cost_table for alignment values.
1637 (ix86_option_override_internal): Use processor_names.
1638 (ix86_function_specific_print): Likewise.
1639 * config/i386/i386.h (struct processor_costs):
1640 Add alignment values.
1641 (struct ptt): Remove and replace with const char *.
1642 * config/i386/x86-tune-costs.h (struct processor_costs):
1643 Declare default alignments for all costs.
1645 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
1647 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
1648 symbolics or VR_VARYING ranges for ABS_EXPR.
1649 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
1650 when range will wrap.
1652 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
1655 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
1656 before and after a JUMP_TABLE_DATA.
1658 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
1661 * dojump.c (do_compare_and_jump): Canonicalize function pointers
1662 when one operand is a function pointer. Use POINTER_TYPE_P and
1663 FUNC_OR_METHOD_TYPE_P.
1664 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
1665 * fold-const.c (build_range_check): Likewise.
1666 * match.pd (simple_comparison): Likewise.
1668 2018-09-14 David Malcolm <dmalcolm@redhat.com>
1671 * spellcheck.c (get_edit_distance_cutoff): New function.
1672 (selftest::test_edit_distance_unit_test_oneway): Rename to...
1673 (selftest::test_get_edit_distance_one_way): ...this.
1674 (selftest::test_get_edit_distance_unit): Rename to...
1675 (selftest::test_get_edit_distance_both_ways): ...this.
1676 (selftest::test_edit_distances): Move tests to this new function,
1677 and test some more pairs of strings. Update for above renaming.
1678 (selftest::get_old_cutoff): New function.
1679 (selftest::test_get_edit_distance_cutoff): New function.
1680 (selftest::assert_suggested_for): New function.
1681 (ASSERT_SUGGESTED_FOR): New macro.
1682 (selftest::assert_not_suggested_for): New function.
1683 (ASSERT_NOT_SUGGESTED_FOR): New macro.
1684 (selftest::test_suggestions): New function.
1685 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
1686 tests to selftest::test_edit_distances and call it. Add calls to
1687 selftest::test_get_edit_distance_cutoff and
1688 selftest::test_suggestions.
1689 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
1690 (best_match::consider): Replace hard-coded cutoff calculation with
1692 (best_match::get_cutoff): New declaration.
1693 (best_match::get_best_meaningful_candidate): Likewise.
1695 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
1697 * builtins.c (fold_builtin_strlen): Remove TODO comment.
1699 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
1702 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1704 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1705 terminated string literal.
1707 2018-09-14 Martin Sebor <msebor@redhat.com>
1709 * builtins.c (unterminated_array): Handle ARRAY_REF.
1710 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
1711 * builtins.h (unterminated_array): Declare extern.
1712 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
1714 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
1717 2018-09-14 Martin Sebor <msebor@redhat.com>
1718 Jeff Law <law@redhat.com>
1720 * builtins.c (unterminated_array): New.
1721 (expand_builtin_strcpy): Adjust.
1722 (expand_builtin_strcpy_args): Detect unterminated arrays.
1723 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
1724 unterminated arrays.
1725 * gimple-fold.h (get_maxval_strlen): Add argument.
1726 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
1728 * gimple-fold.c (get_range_strlen): Add argument.
1729 (get_maxval_strlen): Adjust.
1730 * gimple-fold.h (get_range_strlen): Add argument.
1732 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
1734 * config/i386/movdirintrin.h: Fix copyright year.
1736 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
1738 * reg-stack.c: Include regs.h.
1739 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
1740 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
1741 FIRST_STACK_REG, not DFmode.
1742 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
1743 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
1744 (change stack): Default register mode to the reg_raw_mode of
1745 FIRST_STACK_REG, not DFmode.
1746 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
1747 (*swapxf): Rename from swapxf.
1749 2018-09-14 Carl Love <cel@us.ibm.com>
1751 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
1752 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
1754 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
1757 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
1760 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
1763 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
1765 2018-09-14 Jason Merrill <jason@redhat.com>
1767 Fix --enable-gather-detailed-mem-stats.
1768 * hash-table.c (hash_table_usage): Change from variable to function.
1769 * hash-table.h: Adjust.
1770 * Makefile.in: Add missing dependencies on hash-table.h.
1772 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1774 PR tree-optimization/87259
1776 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
1777 execute_cse_reciprocals_1 has tried transforming.
1779 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
1781 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
1782 VR_VARYING for PLUS/MINUS_EXPR.
1784 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
1786 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
1789 2018-09-14 Richard Biener <rguenther@suse.de>
1792 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
1793 bits for the merged partition.
1795 2018-09-13 Martin Sebor <msebor@redhat.com>
1796 Bernd Edlinger <bernd.edlinger@hotmail.de>
1798 * builtins.h (c_srlen): Add argument.
1799 * builtins.c (warn_string_no_nul): New function.
1800 (c_strlen): Add argument and use it. Update recursive calls.
1801 Pass DECL argument to string_constant to get info on non
1802 terminated strings. Update *NONSTR as needed.
1803 (fold_builtin_strlen): Add argument to calls to c_strlen.
1804 Warn for unterminated arrays.
1805 (warn_string_no_null): Add prototype.
1806 * expr.c (string_constant): Update arguments. Update recursive
1807 calls appropriately. Detect missing NUL terminator and outermost
1808 declaration its missing in.
1809 Improve checks for arrays with nonzero lower bound or elements
1810 that are not a single byte. Simplify offset computation.
1811 Simplify checks for non-NUL terminated strings.
1812 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
1813 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
1815 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
1817 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
1819 * fold-const.c (c_getstr): Fix function comment. Remove unused third
1820 argument. Fix range checks.
1821 * fold-const.h (c_getstr): Adjust protoype.
1822 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
1823 string is constant but contains no NUL byte.
1825 * expr.c (string_constant): Adjust function comment.
1826 Remove bogus check for zero termination.
1828 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
1830 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
1831 (get_constant_size): Don't make STRING_CSTs larger than they are.
1832 (check_string_literal): New check function for STRING_CSTs.
1833 (output_constant): Use it.
1835 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
1838 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
1840 2018-09-13 Richard Biener <rguenther@suse.de>
1842 PR tree-optimization/87263
1843 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
1844 (struct unwind_state): Add max_rpo field.
1845 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
1846 Compute max_rpo, the max RPO number a block can be backwards reached
1847 from. Re-write non-iterating mode to a RPO ordered worklist approach,
1848 separating it from the iterating mode.
1850 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1852 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
1853 (rfs_decision): New scheduling decision.
1855 2018-09-13 Richard Biener <rguenther@suse.de>
1858 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
1859 (vn_nary_op_insert_pieces_predicated): Do not write useless
1860 valid_dominated_by_p entry outside of the allocated storage.
1862 2018-09-13 Omar Sandoval <osandov@osandov.com>
1863 Tom de Vries <tdevries@suse.de>
1866 * dwarf2out.c (is_c): New function.
1867 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
1869 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
1872 * config/aarch64/aarch64.md (*aarch64_bfxil):
1874 * config/aarch64/constraints.md (Ulc): Define.
1875 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
1877 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
1880 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1882 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
1883 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
1884 aarch64_layout_frame call.
1885 (aarch64_expand_epilogue): Likewise.
1886 (aarch64_initial_elimination_offset): Likewise.
1887 (aarch64_get_separate_components): Likewise.
1888 (aarch64_use_return_insn_p): Likewise.
1889 (aarch64_layout_frame): Remove unneeded check.
1891 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1893 * configure.ac: Only append
1894 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
1895 gcc_config_arguments if it was never reconfigured or last reconfigure
1896 was with different arguments.
1897 * configure: Regenerated.
1899 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1900 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1903 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
1904 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
1906 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1908 PR tree-optimization/87287
1909 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
1910 X % C == 0 to X % (unsigned) C == 0 optimization to ...
1911 * match.pd (X % C == 0): ... here. New optimization.
1913 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1916 * expr.h (maybe_optimize_mod_cmp): Declare.
1917 * expr.c (mod_inv): New function.
1918 (maybe_optimize_mod_cmp): New function.
1919 (do_store_flag): Use it.
1920 * cfgexpand.c (expand_gimple_cond): Likewise.
1922 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
1923 Julian Brown <julian@codesourcery.com>
1926 * gimplify.c (gimplify_scan_omp_clauses): Set
1927 target_firstprivatize_array_bases in OpenACC parallel and kernels
1928 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
1929 OpenACC data regions.
1931 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
1933 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
1934 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
1936 2018-09-12 Richard Biener <rguenther@suse.de>
1938 PR tree-optimization/87280
1939 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
1940 edge but unreachable target.
1941 (do_rpo_vn): For conservatively handling a PHI only mark
1942 the backedge executable but not the block reachable.
1944 2018-09-12 Richard Biener <rguenther@suse.de>
1946 PR tree-optimization/87266
1947 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
1950 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
1952 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
1954 ("trunc<BFP:mode><DFP_ALL:mode>2")
1955 ("trunc<DFP_ALL:mode><BFP:mode>2")
1956 ("extend<BFP:mode><DFP_ALL:mode>2")
1957 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
1958 according to the target operand type.
1960 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1961 Andreas Krebbel <krebbel@linux.ibm.com>
1963 PR tree-optimization/86844
1964 * gimple-ssa-store-merging.c
1965 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
1966 there are any overlapping stores in between them, make sure they are
1967 also coalesced or we give up completely.
1969 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1972 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
1973 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
1975 2018-09-12 Tom de Vries <tdevries@suse.de>
1977 * common.opt (gdescribe-dies): Add option.
1978 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
1979 attribute for artifical and nameless decls.
1980 (dwarf2out_register_external_die): Add description attribute to
1981 external reference die.
1982 (add_desc_attribute): New functions.
1983 (gen_subprogram_die): Add description attribute to
1984 DW_TAG_call_site_parameter.
1985 * tree-pretty-print.c (print_generic_expr_to_str): New function.
1986 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
1987 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
1989 (@item -gdescribe-dies): Add.
1991 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1993 * tree-vrp.c (vrp_shift_undefined_p): Remove.
1994 (extract_range_from_binary_expr_1: Call
1995 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
1996 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
1999 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2001 * gimple-ssa-warn-alloca.c
2002 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2003 field for ALLOCA_BOUND_*_LARGE.
2005 2018-09-11 Nathan Sidwell <nathan@acm.org>
2007 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2009 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2011 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2012 for clobbers. Remove obsolete comment.
2014 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2016 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2017 mpxchk, mpxld and mpxst types.
2018 (define_attr length_immediate): Remove all processing of mpx types.
2019 (define_attr prefix_0f): Ditto.
2020 (define_attr memory): Ditto.
2022 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2024 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2025 (log<mode>2): Change operand 1 predicate to general_operand.
2026 Extend operand 1 to XFmode and generate logxf3 insn.
2027 (log10<mode>2): Change operand 1 predicate to general_operand.
2028 Extend operand 1 to XFmode and generate log10xf3 insn.
2029 (log2<mode>2): Change operand 1 predicate to general_operand.
2030 Extend operand 1 to XFmode and generate log2xf3 insn.
2031 (fyl2xp1_extend<mode>xf3_i387): Remove.
2032 (log1p<mode>2): Change operand 1 predicate to general_operand.
2033 Extend operand 1 to XFmode and generate log1pxf3 insn.
2034 (fxtract_extend<mode>xf3_i387): Remove.
2035 (logb<mode>2): Change operand 1 predicate to general_operand.
2036 Extend operand 1 to XFmode and generate logbxf3 insn.
2037 (ilogb<mode>2): Change operand 1 predicate to general_operand.
2038 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2039 (significand<mode>2): Change operand 1 predicate to general_operand.
2040 Extend operand 1 to XFmode and generate significandxf3 insn.
2042 2018-09-11 Nathan Sidwell <nathan@acm.org>
2044 * gcc.c (perror_with_name, pfatal_with_name): Delete.
2045 (load_specs): Use fatal_error.
2046 (DELETE_IF_ORDINARY, process_command): Use error.
2047 (execute, run_attempt): Use fatal_error.
2049 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
2051 * diagnostic-core.h (sorry_at): New prototype.
2052 * diagnostic.c (sorry_at): New function.
2054 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
2056 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2057 by zero as VR_UNDEFINED.
2059 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2061 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2062 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2063 (<sincos>mode2): New expander.
2064 (sincos_extend<mode>xf3_i387): Remove insn pattern.
2065 (sincos -> sin, cos splitters): Remove splitter patterns.
2066 (sincos<mode>3): Change operand 2 predicate to general_operand.
2067 Extend operand 2 to XFmode and generate sincosxf3 insn.
2068 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2069 Change operand 3 predicate to const1_operand.
2070 (fptan_extend<mode>xf4_i387): Remove insn pattern.
2071 (tanxf2): Update operands in the call to fptanxf4_i387.
2072 (tan<mode>2): Change operand 1 predicate to general_operand.
2073 Extend operand 1 to XFmode and generate tanxf3 insn.
2074 (atan2xf3): Rename from *fpatanxf3_i387.
2075 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2076 (atan2xf3): Remove expander.
2077 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
2078 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2079 (atan<mode>2): Change operand 1 predicate to general_operand.
2080 Extend operand 1 to XFmode and generate atanxf3 insn.
2082 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2084 * config/i386/i386.md (x87/SSE constant load splitter): Use
2085 memory_operand instead of nonimmediate_operand for input operand
2088 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
2090 * config/i386/i386.md (float partial SSE register stall splitter): Move
2091 splitter near its instruction pattern.
2092 (float_extend partial SSE register stall splitter): Ditto.
2093 (float_truncate partial SSE register stall splitter): Ditto.
2095 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
2098 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2099 to speculation_safe_value_not_needed.
2102 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2103 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2104 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2105 leaf_function_p, instead use has_hard_reg_initial_val.
2107 2018-09-09 Nathan Sidwell <nathan@acm.org>
2109 * gcc.h (pfatal_with_name): Don't declare here.
2110 * gcc.c (pfatal_with_name): Make static.
2112 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
2114 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2117 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
2119 PR rtl-optimization/85458
2120 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2121 priority hook to reduce the priority of EXPR.
2123 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
2125 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2126 DImode for x87 on 32bit targets. Conditionally disable x87 modes
2127 with X87_ENABLE_FLOAT. Remove preparation code.
2128 (*float<SWI48:mode><MODEF:mode>2): Rename from
2129 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
2130 math using "enabled" attribute.
2131 (*floatdi<MODEF:mode>2_i387): Rename from
2132 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
2133 enable for 32bit targets only.
2134 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2136 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2137 as operand 1 predicate. Rewrite as define_insn_and_split.
2138 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2140 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2142 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2143 to fallthru to FLOAT case.
2145 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2148 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2149 around folding of vec_sl to handle out of range shift values.
2151 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2153 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2154 Update callers to gen_fix_trunc<mode>_i387_fisttp
2155 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2156 nonimmediate_operand.
2157 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2158 and corresponding splitters.
2159 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2160 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2161 (fix_truncdi_i387_with_temp): Remove insn pattern
2162 and corresponding splitters.
2163 (fix_trunc<mode>_i387): Change operand 0 predicate to
2164 nonimmediate_operand.
2165 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2166 and corresponding splitters.
2167 (*fistdi2_1): Remove.
2169 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2170 (lrintxfdi2): Remove expander. Reimplement as define_insn.
2171 (*fist<mode>2_1): Remove.
2172 (fist<mode>2): Ditto.
2173 (fist<mode>2_with_temp): Remove insn pattern and corresponding
2175 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
2176 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2177 (fistdi2_<rounding>): Change operand 0 predicate to
2178 nonimmediate_operand.
2179 (fistdi2_<rounding>_with_temp): Remove insn pattern
2180 and corresponding splitters.
2181 (fist<mode>2_<rounding>): Change operand 0 predicate to
2182 nonimmediate_operand.
2183 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2184 and corresponding splitters.
2186 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2188 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
2190 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2193 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2195 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2196 early gimple folding of vec_splat().
2197 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2198 * gimple-fold.h: Add an extern define for tree_vec_extract().
2200 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2202 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2203 wrappers around TREE_TYPE comparisons.
2205 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2208 * config/s390/predicates.md: Add nonsym_memory_operand.
2209 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2210 contains a SYMBOL_REF, load it into an intermediate pseudo.
2211 (s390_emit_compare_and_swap): Legitimize operand.
2212 * config/s390/s390.md: Use the new nonsym_memory_operand
2213 with UNSPECV_CAS patterns.
2215 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2218 * config/s390/s390-passes.def: New file.
2219 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2221 (make_pass_s390_early_mach): Add declaration.
2222 * config/s390/s390.c (make_pass_s390_early_mach):
2223 (s390_option_override): Remove dynamic registration.
2224 * config/s390/t-s390: Add s390-passes.def.
2226 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2228 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2229 Remove UNSPEC_LTREL_BASE check.
2230 (annotate_constant_pool_refs): Likewise.
2231 (find_constant_pool_ref): Likewise.
2232 (find_ltrel_base): Removed.
2233 (replace_ltrel_base): Removed.
2234 (s390_mainpool_finish): Remove replace_ltrel_base call.
2235 (s390_chunkify_start): Remove pending LTREL_BASE logic.
2236 (s390_chunkify_finish): Remove replace_ltrel_base call.
2237 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2239 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
2242 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2243 to speculation_safe_value_not_needed.
2245 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
2246 Bernd Schmidt <bernds_cb1@t-online.de>
2248 * config/nvptx/nvptx-opts.h: New file.
2249 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2250 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2252 (TARGET_SM35): New macro.
2253 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2255 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2257 (misa=): New option.
2258 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2260 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
2262 * config/i386/i386.md (truncdfsf2): Remove expander.
2263 (truncdfsf2_with_temp): Ditto.
2264 (truncxf<mode>2): Ditto.
2265 (*truncdfsf_fast_mixed): Remove insn pattern.
2266 (*truncdfsf_fast_i387): Ditto.
2267 (*truncdfsf_mixed): Ditto.
2268 (*truncdfsf_i387): Ditto.
2269 (*truncdfsf2_i387_1): Ditto.
2270 (*truncxfsf2_mixed): Ditto.
2271 (*truncxfdf2_mixed): Ditto.
2272 (*truncxf<mode>2_i387_noop): Ditto. Update callers
2273 to call gen_truncxf<mode>2 instead.
2274 (*truncxf<mode>2_i387): Remove.
2275 (reg->reg splitters): Remove splitter pattern.
2276 (reg->mem splitters): Ditto.
2278 (truncdfsf2): New insn pattern.
2279 (truncxf<mode>2): Ditto.
2281 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2283 * tree-ssa-math-opts.c (is_mult_by): New function.
2284 (is_square_of): Use the above.
2285 (optimize_recip_sqrt): New function.
2286 (pass_cse_reciprocals::execute): Use the above.
2288 2018-09-05 Richard Biener <rguenther@suse.de>
2291 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2292 to zero-init the emplaced vec.
2294 2018-09-05 Martin Liska <mliska@suse.cz>
2296 PR tree-optimization/87205
2297 * tree-switch-conversion.c (pass_lower_switch::execute):
2298 Group cases for switch statements.
2300 2018-09-05 Richard Biener <rguenther@suse.de>
2302 PR tree-optimization/87217
2303 * tree-ssa-sccvn.c (vuse_valueize): New.
2304 (vn_reference_lookup_pieces): Use it.
2305 (vn_reference_lookup): Likewise.
2307 2018-09-05 Nathan Sidwell <nathan@acm.org>
2310 * stor-layout.c (place_field): Scan forwards to check last
2311 bitfield when ms_bitfield_placement is in effect.
2313 2018-09-05 Richard Biener <rguenther@suse.de>
2316 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2319 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
2320 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2322 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2323 * config.gcc (extra_objs): Build it.
2324 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2326 * config/aarch64/aarch64-passes.def
2327 (pass_tag_collision_avoidance): New pass.
2328 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2329 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2330 (aarch64_classify_address): Remove static qualifier.
2331 (aarch64_address_info, aarch64_address_type): Move to...
2332 * config/aarch64/aarch64-protos.h: ... here.
2333 (make_pass_tag_collision_avoidance): New function.
2334 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
2337 2018-09-05 Martin Liska <mliska@suse.cz>
2339 * doc/gcov.texi: Update documentation of humar
2341 * gcov.c (format_count): Print one decimal place, it provides
2342 more fine number of situations like '1G' vs. '1.4G'.
2344 2018-09-05 Martin Liska <mliska@suse.cz>
2347 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
2348 * optc-gen.awk: Allow 'Var' for Deprecated options in order
2349 to generate a MASK value.
2351 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
2354 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
2355 if frame pointer isn't used.
2356 (compute_frame_pointer_to_fb_displacement): Likewise.
2358 2018-09-04 Jakub Jelinek <jakub@redhat.com>
2361 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2362 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2363 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2364 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2365 and OPTION_MASK_ISA_XSAVEC_UNSET.
2367 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2369 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2370 XOR operations in NAND case.
2372 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
2374 * wide-int-range.cc (wide_int_range_convert): New.
2375 * wide-int-range.h (wide_int_range_convert): New.
2376 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
2377 code into wide_int_range_convert.
2378 (extract_range_into_wide_ints): Do not munge anti range constants
2379 into the entire domain. Just return the range back.
2381 2018-09-04 Martin Liska <mliska@suse.cz>
2383 * genmatch.c (output_line_directive): Add new argument
2385 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
2387 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2389 * doc/invoke.texi (Option Summary): Add whitespace.
2391 * doc/invoke.texi (Option Summary): Add -Waligned-new.
2393 2018-09-04 Richard Biener <rguenther@suse.de>
2395 PR tree-optimization/87211
2396 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
2397 backedge value we're supposed to treat as VARYING also number
2398 the PHI to VARYING in case it got a different value-number already.
2400 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
2402 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
2403 (extract_range_from_binary_expr_1): Do not call
2404 vrp_can_optimize_bit_op.
2405 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
2407 (wide_int_range_get_mask_and_bounds): New.
2408 (wide_int_range_optimize_bit_op): New.
2409 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
2410 (wide_int_range_bit_and): Same.
2411 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
2412 (wide_int_range_optimize_bit_op): New.
2413 (wide_int_range_get_mask_and_bounds): New.
2415 2018-09-04 Richard Biener <rguenther@suse.de>
2417 PR tree-optimization/87176
2418 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
2419 variable. When value-numbering a virtual PHI node make sure
2420 to not value-number to the backedge value.
2422 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2424 * doc/extend.texi (Long Long, Hex Floats): Document support for
2425 long long and hex floats in more recent versions of ISO C++.
2427 2018-09-03 Richard Biener <rguenther@suse.de>
2429 PR tree-optimization/87177
2430 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
2433 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2435 * bb-reorder.c (edge_order): Convert to C-qsort-style
2436 tri-state comparator.
2437 (reorder_basic_blocks_simple): Change std::stable_sort to
2440 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2442 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
2443 tri-state comparator.
2444 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
2446 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2448 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
2449 (mergesort): ... here as maximum count for using netsort.
2450 (gcc_qsort): Set nlim to 3 if stable sort is requested.
2451 (gcc_stablesort): New.
2452 * system.h (gcc_stablesort): Declare.
2454 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2456 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
2457 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
2458 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
2460 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
2462 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
2463 lxsdx and stxsdx alternatives.
2464 (*mov<mode>_hardfloat64): Ditto.
2465 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
2467 2018-09-03 Richard Biener <rguenther@suse.de>
2469 PR tree-optimization/87200
2470 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
2473 2018-09-03 Martin Liska <mliska@suse.cz>
2475 PR tree-optimization/87201
2476 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2477 Fix parenthesis in an expression.
2479 2018-09-03 Richard Biener <rguenther@suse.de>
2481 PR tree-optimization/87197
2482 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
2483 visited. CSE the VN_INFO hashtable lookup.
2485 PR tree-optimization/87169
2486 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
2487 iterating make sure there's no extra backedges from irreducible
2488 regions feeding the header. Mark the destination block
2491 2018-09-03 Martin Liska <mliska@suse.cz>
2494 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
2495 * common/common-targhooks.c (default_get_valid_option_values):
2497 * common/common-targhooks.h (default_get_valid_option_values):
2499 * common/config/i386/i386-common.c: Move processor_target_table
2501 (ix86_get_valid_option_values): New function.
2502 (TARGET_GET_VALID_OPTION_VALUES): New macro.
2503 * config/i386/i386.c (struct ptt): Move to i386-common.c.
2504 (PTA_*): Move all defined masks into i386-common.c.
2505 (ix86_function_specific_restore): Use new processor_cost_table.
2506 * config/i386/i386.h (struct ptt): Moved from i386.c.
2507 (struct pta): Likewise.
2508 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
2509 * doc/tm.texi.in: Likewise.
2510 * opt-suggestions.c (option_proposer::suggest_option):
2511 Pass prefix to build_option_suggestions.
2512 (option_proposer::get_completions): Likewise.
2513 (option_proposer::build_option_suggestions): Use the new target
2515 * opts.c (struct option_help_tuple): New struct.
2516 (print_filtered_help): Use the new target hook.
2518 2018-09-03 Martin Liska <mliska@suse.cz>
2521 * predict.c (set_even_probabilities): Add likely_edges
2522 argument and handle cases where we have precisely one
2524 (combine_predictions_for_bb): Catch also likely_edges.
2525 (tree_predict_by_opcode): Handle gswitch statements.
2526 * tree-cfg.h (find_case_label_for_value): New declaration.
2527 (find_taken_edge_switch_expr): Likewise.
2528 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2529 Find pivot in decision tree based on probabily, not by number of
2532 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
2534 * doc/standards.texi (Standards): Update Objective-C reference.
2536 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2538 * doc/install.texi (Prerequisites): Update link for MPC.
2540 2018-09-01 Michael Matz <matz@suse.de>
2542 PR tree-optimization/87074
2543 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
2544 PHIs for outer-loop uses.
2546 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2548 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
2549 * doc/invoke.texi (C Dialect Options): Ditto.
2551 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2553 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
2555 2018-08-31 Richard Biener <rguenther@suse.de>
2557 PR tree-optimization/87168
2558 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
2559 (rpo_elim::eliminate_avail): When OP was not visited it must
2562 2018-08-31 David Malcolm <dmalcolm@redhat.com>
2564 * tree-vrp.c (copy_value_range): Convert param "from" from
2565 "value_range *" to "const value_range *".
2566 (range_is_null): Likewise for param "vr".
2567 (range_int_cst_p): Likewise.
2568 (range_int_cst_singleton_p): Likewise.
2569 (symbolic_range_p): Likewise.
2570 (value_ranges_intersect_p): Likewise for both params.
2571 (value_range_nonnegative_p): Likewise for param "vr".
2572 (value_range_constant_singleton): Likewise.
2573 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
2574 (extract_range_into_wide_ints): Likewise for param "vr".
2575 (extract_range_from_multiplicative_op): Likewise for params "vr0"
2577 (vrp_can_optimize_bit_op): Likewise.
2578 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
2580 (extract_range_from_unary_expr): Likewise.
2581 (debug_value_range): Likewise for param "vr".
2582 (value_range::dump): Add "const" qualifier.
2583 (vrp_prop::check_array_ref): Convert local "vr" from
2584 "value_range *" to "const value_range *".
2585 (vrp_prop::check_mem_ref): Likewise.
2586 (vrp_prop::visit_stmt): Likewise for local "old_vr".
2587 (vrp_intersect_ranges_1): Likewise for param "vr_1".
2588 (vrp_intersect_ranges): Likewise.
2589 (simplify_stmt_for_jump_threading): Likewise for local "vr".
2590 (vrp_prop::vrp_finalize): Likewise.
2591 * tree-vrp.h (value_range::dump): Add "const" qualifier.
2592 (vrp_intersect_ranges): Add "const" qualifier to params as above.
2593 (extract_range_from_unary_expr): Likewise.
2594 (value_range_constant_singleton): Likewise.
2595 (symbolic_range_p): Likewise.
2596 (copy_value_range): Likewise.
2597 (extract_range_from_binary_expr_1): Likewise.
2598 (range_int_cst_p): Likewise.
2599 (vrp_set_zero_nonzero_bits): Likewise.
2600 (range_int_cst_singleton_p): Likewise.
2602 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
2604 * config/aarch64/arm_neon.h (vabsd_s64): New.
2605 (vnegd_s64): Likewise.
2607 2018-08-31 Martin Jambor <mjambor@suse.cz>
2609 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
2611 2018-08-31 Martin Liska <mliska@suse.cz>
2613 * ipa-icf.c (sem_item::add_type): Use
2614 sem_item::m_type_hash_cache.
2615 * ipa-icf.h: Move the cache from sem_item_optimizer
2618 2018-08-31 Nathan Sidwell <nathan@acm.org>
2620 * doc/extend.texi (Backwards Compatibility): Remove implicit
2621 extern C leeway of () being (...).
2623 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2625 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
2627 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
2631 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
2633 2018-08-31 Jakub Jelinek <jakub@redhat.com>
2636 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
2637 gen_int_mode. Formatting fixes.
2639 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
2641 * target.def (custom_function_descriptors): Improve documentation.
2642 * doc/tm.texi.in (Trampolines): Expand discussion of function
2643 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
2644 beginning of the section.
2645 * doc/tm.texi: Regenerated.
2647 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
2649 * cfg.h (class auto_edge_flag): Spell out the template-id of the
2650 base class in the initializer list. This is a workaround for
2651 building with older GCC.
2652 (class auto_bb_flag): Likewise.
2654 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
2656 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
2657 (altivec_vcmpequ<VI_char>_p): Remove star.
2658 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
2660 (expand_strncmp_vec_sequence): New function.
2661 (emit_final_str_compare_vec): New function.
2662 (expand_strn_compare): Add support for vector strncmp.
2663 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
2664 length specification to bytes.
2665 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
2666 (vcmpnezb_p): New pattern.
2667 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
2668 for option -mstring-compare-inline-limit.
2670 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
2672 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
2673 (PTA_SKYLAKE): Add PTA_AES.
2674 (PTA_GOLDMONT): Likewise.
2676 2018-08-29 Jan Hubicka <jh@suse.cz>
2679 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
2680 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
2682 2018-08-29 Jan Hubicka <jh@suse.cz>
2684 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
2686 (hash_tree): Do not visit TYPE_STUB_DECL.
2687 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
2688 stream TYPE_STUB_DECL.
2689 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
2690 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
2691 after free_lang_data.
2692 (type_in_anonymous_namespace_p): Likewise.
2694 2018-08-29 Jan Hubicka <jh@suse.cz>
2696 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2697 comment that it has to be even number.
2698 (class sreal): Change m_sig type to int32_t.
2699 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2700 int64_t for temporary calculations.
2701 (sreal_verify_basics): Drop one bit from minimum and maximum.
2703 2018-08-30 Richard Biener <rguenther@suse.de>
2705 PR tree-optimization/87147
2706 * tree-ssa-sccvn.c (SSA_VISITED): New function.
2707 (visit_phi): When the degenerate result is from the backedge and
2708 we didn't visit its definition yet drop to VARYING.
2709 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
2711 2018-08-29 Jan Hubicka <jh@suse.cz>
2713 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
2715 (hash_tree): Likewise.
2717 2018-08-29 Jan Hubicka <jh@suse.cz>
2719 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
2720 and TYPE_NEXT_REF_TO.
2722 2018-08-29 Jan Hubicka <jh@suse.cz>
2724 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2725 comment that it has to be even number.
2726 (class sreal): Change m_sig type to int32_t.
2727 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2728 int64_t for temporary calculations.
2729 (sreal_verify_basics): Drop one bit from minimum and maximum.
2731 2018-08-30 Tamar Christina <tamar.christina@arm.com>
2733 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
2735 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
2738 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
2739 if to_add is negative.
2741 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
2744 * builtins.c (c_strlen): Improve range checks.
2746 2018-08-29 Martin Sebor <msebor@redhat.com>
2747 Jeff Law <law@redhat.com>
2749 PR tree-optimization/86714
2750 PR tree-optimization/86711
2751 * builtins.c (c_strlen): Add arguments to call to string_constant.
2752 * expr.c (string_constant): Add argument. Detect missing nul
2753 terminator and outermost declaration it's missing in.
2754 * expr.h (string_constant): Add argument.
2755 * fold-const.c (read_from_constant_string): Add arguments to call to
2757 (c_getstr): Likewise.
2758 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2760 * tree-ssa-strlen.c (get_stridx): Likewise.
2762 2018-08-29 Jan Hubicka <jh@suse.cz>
2764 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
2765 Do not stream DECL_VINDEX.
2766 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
2767 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
2768 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
2770 2018-08-29 Richard Biener <rguenther@suse.de>
2772 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
2773 virtual operands that are not default defs to honor region
2775 (rpo_vn_valueize): Remove ineffective code here.
2777 2018-08-29 Richard Biener <rguenther@suse.de>
2779 PR tree-optimization/87132
2780 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
2781 when skipping defs reachable over backedges.
2783 2018-08-29 Richard Biener <rguenther@suse.de>
2785 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
2786 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
2787 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
2788 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
2789 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
2790 (vn_reference_lookup_3): Remove use of const_parms.
2791 (free_rpo_vn): Do not free const_parms.
2792 (do_rpo_vn): Do not call init_const_parms.
2793 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
2794 SSA_NAME_POINTS_TO_READONLY_MEMORY.
2795 (call_may_clobber_ref_p_1): Likewise.
2797 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
2800 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
2802 (-ftree-scev-cprop): Document.
2804 2018-08-29 Jan Hubicka <jh@suse.cz>
2806 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
2808 (sreal constructor): Update.
2809 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
2810 sreal:operator/): Update.
2812 2018-08-29 Martin Liska <mliska@suse.cz>
2814 * tree-switch-conversion.c (switch_conversion::expand):
2815 Strenghten assumption about gswitch statements.
2817 2018-08-29 Richard Biener <rguenther@suse.de>
2819 PR tree-optimization/87117
2820 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
2821 re-value-number released SSA VDEFs.
2823 2018-08-29 Richard Biener <rguenther@suse.de>
2825 PR tree-optimization/87126
2826 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
2828 2018-08-28 Jim Wilson <jimw@sifive.com>
2830 * config/riscv/pic.md: Rewrite.
2831 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
2833 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
2834 (SOFTF, default_load, softload, softstore): New.
2836 2018-08-28 Jeff Law <law@redhat.com>
2838 * fold-const.c (fold_binary_loc): Remove recently added assert.
2840 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
2842 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
2843 to OP parmeter of generated function.
2845 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
2847 PR tree-optimization/87009
2848 * match.pd: Add boolean optimizations.
2850 2018-08-28 Martin Sebor <msebor@redhat.com>
2853 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
2854 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
2855 (pass_walloca::gate): Use it.
2856 (alloca_call_type): Same.
2857 (pass_walloca::execute): Same.
2858 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
2860 2018-08-28 David Malcolm <dmalcolm@redhat.com>
2862 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
2863 GCC_VERSION for usage of "__gcc_dump_printf__" format from
2866 2018-08-28 Richard Biener <rguenther@suse.de>
2868 PR tree-optimization/87124
2869 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
2870 constants before looking up avail.
2872 2018-08-28 Jakub Jelinek <jakub@redhat.com>
2875 * calls.c (maybe_warn_nonstring_arg): Punt early if
2876 warn_stringop_overflow is zero. Don't call get_range_strlen
2877 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2878 Swap comparison operands to have constants on rhs. Only use
2879 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
2880 increment lenrng[0].
2882 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
2884 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
2885 use of tree_to_shwi. Remove duplicated test for the size being
2886 a whole number of bytes.
2888 2018-08-28 Richard Biener <rguenther@suse.de>
2890 PR tree-optimization/87117
2891 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
2892 Handle removed stmt without LHS (GIMPLE_NOP).
2894 2018-08-28 Richard Biener <rguenther@suse.de>
2896 PR tree-optimization/87117
2897 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
2898 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
2900 2018-08-28 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/87117
2903 * tree-ssa-pre.c (compute_avail): Do not make expressions
2904 with predicated values available.
2905 (get_expr_value_id): Assert we do not run into predicated value
2908 2018-08-28 Richard Biener <rguenther@suse.de>
2910 PR tree-optimization/87117
2911 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
2912 get virtual operands.
2913 (get_expr_operands): Handle STRING_CST like other decls.
2915 2018-08-28 Martin Liska <mliska@suse.cz>
2917 * tree.h: Update documentation of fndecl_built_in_p
2921 2018-08-27 Jeff Law <law@redhat.com>
2922 PR tree-optimization/87110
2923 * tree-ssa-dse.c (compute_trims): Handle non-constant
2926 2018-08-27 Martin Sebor <msebor@redhat.com>
2928 PR tree-optimization/86914
2929 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2931 2018-08-27 Martin Sebor <msebor@redhat.com>
2933 PR tree-optimization/87112
2934 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
2935 the type of the bound argument.
2937 2018-08-27 Jeff Law <law@redhat.com>
2939 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
2940 type does not have a TYPE_SIZE_UNIT.
2942 2018-08-27 Steve Ellcey <sellcey@cavium.com>
2944 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
2945 with include of backend.h.
2947 2018-08-27 Richard Biener <rguenther@suse.de>
2949 PR tree-optimization/86927
2950 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2951 use const cond reduction code.
2953 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
2955 PR tree-optimization/85758
2956 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
2958 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2961 * diagnostic-show-locus.c (class layout_range): Update for
2962 conversion of show_caret_p to a tri-state.
2963 (layout_range::layout_range): Likewise.
2964 (make_range): Likewise.
2965 (layout::maybe_add_location_range): Likewise.
2966 (layout::should_print_annotation_line_p): Don't show annotation
2967 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
2968 (layout::get_state_at_point): Update for conversion of
2969 show_caret_p to a tri-state. Bail out early for
2970 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
2971 underlining or source colorization.
2972 (gcc_rich_location::add_location_if_nearby): Update for conversion
2973 of show_caret_p to a tri-state.
2974 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
2975 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
2977 (selftest::test_one_liner_labels): Likewise.
2978 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
2979 conversion of show_caret_p to a tri-state.
2980 * pretty-print.c (text_info::set_location): Likewise.
2981 * pretty-print.h (text_info::set_location): Likewise.
2982 * substring-locations.c (format_warning_n_va): Likewise.
2983 * tree-diagnostic.c (default_tree_printer): Likewise.
2984 * tree-pretty-print.c (newline_and_indent): Likewise.
2986 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2989 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
2990 line above for line-insertion fix-it hints.
2991 (selftest::test_fixit_insert_containing_newline): Update the
2992 expected results, and add a test with line-numbering enabled.
2994 2018-08-27 Martin Liska <mliska@suse.cz>
2997 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
2998 params with DECL_HAS_VALUE_EXPR_P.
3000 2018-08-27 Martin Liska <mliska@suse.cz>
3002 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3003 selected expansion strategy.
3005 2018-08-27 Martin Liska <mliska@suse.cz>
3007 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3008 * builtins.c (is_builtin_fn): Likewise.
3009 * attribs.c (diag_attr_exclusions): Use new function
3010 fndecl_built_in_p and remove check for FUNCTION_DECL if
3012 (builtin_mathfn_code): Likewise.
3013 (fold_builtin_expect): Likewise.
3014 (fold_call_expr): Likewise.
3015 (fold_builtin_call_array): Likewise.
3016 (fold_call_stmt): Likewise.
3017 (set_builtin_user_assembler_name): Likewise.
3018 (is_simple_builtin): Likewise.
3019 * calls.c (gimple_alloca_call_p): Likewise.
3020 (maybe_warn_nonstring_arg): Likewise.
3021 * cfgexpand.c (expand_call_stmt): Likewise.
3022 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3023 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3024 (cgraph_node::verify_node): Likewise.
3025 * cgraphclones.c (build_function_decl_skip_args): Likewise.
3026 (cgraph_node::create_clone): Likewise.
3027 * config/arm/arm.c (arm_insert_attributes): Likewise.
3028 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3029 * dse.c (scan_insn): Likewise.
3030 * expr.c (expand_expr_real_1): Likewise.
3031 * fold-const.c (operand_equal_p): Likewise.
3032 (fold_binary_loc): Likewise.
3033 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3034 * gimple-low.c (lower_stmt): Likewise.
3035 * gimple-pretty-print.c (dump_gimple_call): Likewise.
3036 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3038 * gimple.c (gimple_build_call_from_tree): Likewise.
3039 (gimple_call_builtin_p): Likewise.
3040 (gimple_call_combined_fn): Likewise.
3041 * gimplify.c (gimplify_call_expr): Likewise.
3042 (gimple_boolify): Likewise.
3043 (gimplify_modify_expr): Likewise.
3044 (gimplify_addr_expr): Likewise.
3045 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3046 * ipa-cp.c (determine_versionability): Likewise.
3047 * ipa-fnsummary.c (compute_fn_summary): Likewise.
3048 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3049 * ipa-split.c (visit_bb): Likewise.
3050 (split_function): Likewise.
3051 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3052 * lto-cgraph.c (input_node): Likewise.
3053 * lto-streamer-out.c (write_symbol): Likewise.
3054 * omp-low.c (setjmp_or_longjmp_p): Likewise.
3055 (lower_omp_1): Likewise.
3056 * predict.c (strip_predict_hints): Likewise.
3057 * print-tree.c (print_node): Likewise.
3058 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3059 * trans-mem.c (is_tm_irrevocable): Likewise.
3060 (is_tm_load): Likewise.
3061 (is_tm_simple_load): Likewise.
3062 (is_tm_store): Likewise.
3063 (is_tm_simple_store): Likewise.
3064 (is_tm_abort): Likewise.
3065 (tm_region_init_1): Likewise.
3066 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3067 * tree-cfg.c (verify_gimple_call): Likewise.
3068 (move_stmt_r): Likewise.
3069 (stmt_can_terminate_bb_p): Likewise.
3070 * tree-eh.c (lower_eh_constructs_2): Likewise.
3071 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3072 * tree-inline.c (remap_gimple_stmt): Likewise.
3073 (copy_bb): Likewise.
3074 (estimate_num_insns): Likewise.
3075 (fold_marked_statements): Likewise.
3076 * tree-sra.c (scan_function): Likewise.
3077 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3078 (optimize_stack_restore): Likewise.
3079 (pass_fold_builtins::execute): Likewise.
3080 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3081 (mark_all_reaching_defs_necessary_1): Likewise.
3082 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3083 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3084 (pass_forwprop::execute): Likewise.
3085 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3086 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3087 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3088 * tree-ssa-strlen.c (get_string_length): Likewise.
3089 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3090 (find_func_aliases_for_call): Likewise.
3091 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3092 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3093 * tree-tailcall.c (find_tail_calls): Likewise.
3094 * tree.c (need_assembler_name_p): Likewise.
3095 (free_lang_data_in_decl): Likewise.
3096 (get_call_combined_fn): Likewise.
3097 * ubsan.c (is_ubsan_builtin_p): Likewise.
3098 * varasm.c (incorporeal_function_p): Likewise.
3099 * tree.h (DECL_BUILT_IN): Remove and replace with
3101 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3102 (fndecl_built_in_p): New.
3104 2018-08-27 Martin Liska <mliska@suse.cz>
3106 PR tree-optimization/86847
3107 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3108 Dump also subtree probability.
3109 (switch_decision_tree::do_jump_if_equal): New function.
3110 (switch_decision_tree::emit_case_nodes): Handle special
3111 situations in balanced tree that can be emitted much simpler.
3112 Fix calculation of probabilities that happen in tree expansion.
3113 * tree-switch-conversion.h (struct cluster): Add
3115 (struct simple_cluster): Likewise.
3116 (struct case_tree_node): Add new function has_child.
3117 (do_jump_if_equal): New.
3119 2018-08-27 Martin Liska <mliska@suse.cz>
3121 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3122 Add new argument to bit_test_cluster constructor.
3123 (bit_test_cluster::emit): Set bits really number of values
3125 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3126 probability argument.
3127 * tree-switch-conversion.h (struct bit_test_cluster):
3128 Add m_handles_entire_switch.
3130 2018-08-27 Martin Liska <mliska@suse.cz>
3132 PR tree-optimization/86702
3133 * tree-switch-conversion.c (jump_table_cluster::emit):
3134 Make probabilities even for values in jump table
3135 according to number of cases handled.
3136 (switch_decision_tree::compute_cases_per_edge): Pass
3137 argument to reset_out_edges_aux function.
3138 (switch_decision_tree::analyze_switch_statement): Likewise.
3139 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3142 2018-08-27 Martin Liska <mliska@suse.cz>
3144 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3145 cfun argument explicitly.
3146 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3147 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3148 function gimple_switch_default_bb.
3149 (convert_switch_statements):
3151 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3152 * stmt.c (label_to_block_fn): Use label_to_block and pass
3153 cfun argument explicitly and use gimple_switch_label_bb.
3154 (expand_case): Likewise.
3155 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3156 cfun argument explicitly. Likewise.
3157 (make_edges_bb): Likewise.
3158 (make_cond_expr_edges): Likewise.
3159 (get_cases_for_edge): Likewise.
3160 (make_gimple_switch_edges): Likewise.
3161 (label_to_block_fn): Likewise.
3162 (label_to_block): Likewise.
3163 (make_goto_expr_edges): Likewise.
3164 (make_gimple_asm_edges): Likewise.
3165 (main_block_label): Likewise.
3166 (group_case_labels_stmt): Likewise.
3167 (find_taken_edge_computed_goto): Likewise.
3168 (find_taken_edge_switch_expr): Likewise.
3169 (gimple_verify_flow_info): Likewise.
3170 (gimple_redirect_edge_and_branch): Likewise.
3171 (gimple_switch_label_bb): New function.
3172 (gimple_switch_default_bb): Likewise.
3173 (gimple_switch_edge): Likewise.
3174 (gimple_switch_default_edge): Likewise.
3175 * tree-cfg.h (label_to_block_fn): Remove and replace ...
3176 (label_to_block): ... with this.
3177 (gimple_switch_label_bb): New.
3178 (gimple_switch_default_bb): Likewise.
3179 (gimple_switch_edge): Likewise.
3180 (gimple_switch_default_edge): Likewise.
3181 * tree-cfgcleanup.c (convert_single_case_switch): Use
3182 new gimple functions and pass new argument to label_to_block.
3183 (cleanup_control_flow_bb):
3184 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3185 cfun argument explicitly.
3186 (make_eh_edges): Likewise.
3187 (redirect_eh_dispatch_edge): Likewise.
3188 (lower_resx): Likewise.
3189 (lower_eh_dispatch): Likewise.
3190 (maybe_remove_unreachable_handlers): Likewise.
3191 (unsplit_eh): Likewise.
3192 (cleanup_empty_eh): Likewise.
3193 (verify_eh_edges): Likewise.
3194 (verify_eh_dispatch_edge): Likewise.
3195 * tree-ssa-dom.c (record_edge_info): Likewise.
3196 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3197 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3198 (thread_through_normal_block): Likewise.
3199 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3200 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3201 * tree-switch-conversion.c (switch_conversion::collect): Use new
3203 (switch_conversion::check_final_bb): Likewise.
3204 (switch_conversion::gather_default_values): Pass new argument
3206 (switch_conversion::build_constructors): Likewise.
3207 (switch_decision_tree::compute_cases_per_edge): Use new
3208 gimple_switch_edge function.
3209 (switch_decision_tree::analyze_switch_statement): Pass new argument
3211 (switch_decision_tree::try_switch_expansion): Use
3212 gimple_switch_default_edge.
3213 * tree-vrp.c (find_switch_asserts): Pass new argument
3215 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3216 (vr_values::simplify_switch_using_ranges): Likewise.
3218 2018-08-27 Richard Biener <rguenther@suse.de>
3220 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3221 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3223 * tree-ssa-sccvn.h (struct vn_pval): New structure.
3224 (struct vn_nary_op_s): Add unwind_to member. Add
3225 predicated_values flag and put result into a union together
3226 with a linked list of vn_pval.
3227 (struct vn_ssa_aux): Add name member to make maintaining
3228 a map of SSA name to vn_ssa_aux possible. Remove no longer
3229 needed info, dfsnum, low, visited, on_sccstack, use_processed
3230 and range_info_anti_range_p members.
3231 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3232 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3234 (vn_valueize): New global.
3235 (vn_context_bb): Likewise.
3236 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3237 VN_INFO_PTR_INFO): Remove.
3238 * tree-ssa-sccvn.c: ... (rewrite)
3239 (pass_fre::execute): For -O2+ initialize loops and run
3240 RPO VN in optimistic mode (iterating). For -O1 and -Og
3241 run RPO VN in non-optimistic mode.
3242 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3243 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3244 * doc/invoke.texi (sccvn-max-scc-size): Remove.
3245 (rpo-vn-max-loop-depth): Document.
3246 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3247 when valuezing the VUSE signals we walked out of the region.
3248 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3249 (phi_translate): Set VN context block to use for availability
3251 (compute_avail): Likewise.
3252 (pre_valueize): New function.
3253 (pass_pre::execute): Adjust to the RPO VN API.
3255 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3256 (propagate_constants_for_unrolling): Remove.
3257 (tree_unroll_loops_completely): Perform value-numbering
3258 on the unrolled bodies loop parent.
3260 2018-08-27 Richard Biener <rguenther@suse.de>
3262 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3263 for partial antic compute.
3265 2018-08-27 Jakub Jelinek <jakub@redhat.com>
3267 PR rtl-optimization/87065
3268 * combine.c (simplify_if_then_else): Formatting fix.
3269 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3271 (known_cond): Don't return const_true_rtx for vector modes. Use
3272 CONST0_RTX instead of const0_rtx. Formatting fixes.
3274 2018-08-27 Martin Liska <mliska@suse.cz>
3276 PR gcov-profile/87069
3277 * gcov.c (process_file): Record files already processed
3278 and warn about a file being processed multiple times.
3280 2018-08-27 Martin Liska <mliska@suse.cz>
3283 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3284 Set default values for x_aarch64_*_string strings.
3285 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3286 prefix. For -mabi do not print '=ABI' in help and use
3287 <option_value> format for -msve-vector-bits and -moverride
3290 2018-08-26 Jeff Law <law@redhat.com>
3292 * config/mips/frame-header-opt.c: Include "backend.h" rather than
3295 2018-08-26 Marek Polacek <polacek@redhat.com>
3297 PR c++/87029, Implement -Wredundant-move.
3298 * doc/invoke.texi: Document -Wredundant-move.
3300 2018-08-25 Martin Sebor <msebor@redhat.com>
3302 PR tree-optimization/87059
3303 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3304 to the same type as the other.
3305 * fold-const.c (fold_binary_loc): Assert expectation.
3307 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
3309 * config/darwin.c (machopic_legitimize_pic_address): Clean up
3310 extraneous parentheses, dead code section and formatting.
3312 2018-08-24 David Malcolm <dmalcolm@redhat.com>
3315 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3316 wide enough for jumps in the line-numbering to be visible.
3317 (layout::print_gap_in_line_numbering): New member function.
3318 (layout::calculate_line_spans): When using line numbering, merge
3319 line spans that are only 1 line apart.
3320 (diagnostic_show_locus): When printing line numbers, show gaps in
3321 line numbering directly, rather than printing headers.
3322 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3323 line-numbering with multiple line spans.
3324 (selftest::test_fixit_insert_containing_newline_2): Add test of
3325 line-numbering, in which the spans are close enough to be merged.
3327 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3329 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3330 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
3331 * tree-vrp.c (range_is_nonnull): Remove.
3332 (range_includes_zero_p): Accept value_range instead of min/max.
3333 (extract_range_from_binary_expr_1): Do not early bail on
3335 Use range_includes_zero_p instead of range_is_nonnull.
3336 (extract_range_from_unary_expr): Use range_includes_zero_p instead
3337 of range_is_nonnull.
3338 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
3339 special case VR_ANTI_RANGE.
3340 (vrp_finalize): Same.
3341 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
3343 (range_is_nonnull): Remove.
3344 * vr-values.c (vrp_stmt_computes_nonzero): Use
3345 range_includes_zero_p instead of range_is_nonnull.
3346 (extract_range_basic): Pass value_range to range_includes_zero_p
3347 instead of range_is_nonnull.
3349 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
3351 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
3352 * emit-rtl.h (rtl_data): Remove return_bnd.
3353 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
3354 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
3355 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
3356 (POINTER_BOUNDS_MODE): Remove definition.
3357 (make_pointer_bounds_mode): Remove.
3358 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
3359 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
3360 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
3361 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
3362 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
3363 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
3364 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
3366 * config/i386/i386-modes.def (BND32, BND64): Remove.
3367 * config/i386/i386.c (dbx_register_map): Remove bound registers.
3368 (dbx64_register_map): Ditto.
3369 (svr4_dbx_register_map): Ditto.
3370 (indirect_thunk_bnd_needed): Remove.
3371 (indirect_thunks_bnd_used): Ditto.
3372 (indirect_return_bnd_needed): Ditto.
3373 (indirect_return_via_cx_bnd): Ditto.
3374 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
3375 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
3376 (output_indirect_thunk): Ditto. Remove need_prefix argument.
3377 (output_indirect_thunk_function): Remove handling of
3378 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
3379 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
3380 (ix86_save_reg): Remove handling of crtl->return_bnd.
3381 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
3382 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
3383 and UNSPEC_BNDLX_ADDR.
3384 (ix86_output_indirect_branch_via_reg): Remove handling of
3385 indirect_thunk_prefix_bnd.
3386 (ix86_output_indirect_branch_via_push): Ditto.
3387 (ix86_output_function_return): Ditto.
3388 (ix86_output_indirect_function_return): Ditto.
3389 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
3390 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
3391 (CALL_USED_REGISTERS): Ditto.
3392 (REG_ALLOC_ORDER): Update for removal of bound registers.
3393 (HI_REGISTER_NAMES): Ditto.
3394 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
3395 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
3396 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
3397 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
3398 (FIRST_PSEUDO_REG): Update.
3399 (BND): Remove mode iterator.
3400 * config/i386/predicates.md (bnd_mem_operator): Remove.
3402 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
3404 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
3407 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
3409 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
3410 the case in which the permute needs only a single element and
3411 repeats for every vector of the result. Extend that case to
3412 handle variable-length vectors.
3413 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3415 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
3418 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
3419 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
3421 2018-08-24 Richard Biener <rguenther@suse.de>
3423 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
3424 bb_flags_allocated members.
3425 (auto_flag): New RAII class for allocating flags.
3426 (auto_edge_flag): New RAII class for allocating edge flags.
3427 (auto_bb_flag): New RAII class for allocating bb flags.
3428 * cfgloop.c (verify_loop_structure): Allocate temporary edge
3430 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
3431 in favor of temporarily allocated BB flag.
3432 * hsa-brig.c: Re-order includes.
3433 * hsa-dump.c: Likewise.
3434 * hsa-regalloc.c: Likewise.
3435 * print-rtl.c: Likewise.
3436 * profile-count.c: Likewise.
3438 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
3441 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
3444 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3447 * wide-int-range.cc (wide_int_range_div): Do not ignore result
3448 from wide_int_range_multiplicative_op.
3450 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3452 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
3455 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
3457 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
3460 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
3462 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
3463 full GENERIC expression used for replacement.
3465 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
3467 * tree-vrp.c (abs_extent_range): Remove.
3468 (extract_range_into_wide_ints): Pass wide ints by reference.
3469 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
3470 Pass wide ints by reference in all calls to
3471 extract_range_into_wide_ints.
3472 * wide-int-range.cc (wide_int_range_div): New.
3473 * wide-int-range.h (wide_int_range_div): New.
3474 (wide_int_range_includes_zero_p): New.
3475 (wide_int_range_zero_p): New.
3477 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
3479 * config/aarch64/aarch64.md (arches): New enum.
3480 (arch): New enum attr.
3481 (arch_enabled): New attr.
3482 (enabled): Now uses arch_enabled only.
3483 (simd, sve, fp16): Removed attribute.
3484 (fp): Attr now defined in terms of 'arch'.
3485 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
3486 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
3487 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
3488 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
3489 attributes into 'arch'.
3490 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
3491 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
3492 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
3493 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
3494 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
3495 'simd' attribute into 'arch'.
3496 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
3497 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
3498 Convert use of 'fp' attribute to 'arch'.
3499 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
3500 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
3502 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
3503 (different modes) Merge 'fp' and 'simd' into 'arch'.
3504 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
3507 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
3509 PR rtl-optimization/87026
3510 * expmed.c (canonicalize_comparison): If we can no longer create
3511 pseudoregisters, don't.
3513 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
3516 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
3518 * config/arm/arm.c (speculation_barrier_libfunc): New static
3520 (arm_init_libfuncs): Initialize it.
3521 (arm_emit_speculation_barrier): New function.
3522 * config/arm/arm.md (speculation_barrier): Call
3523 arm_emit_speculation_barrier for architectures that do not have
3525 (speculation_barrier_insn): Only match on Armv7 or later.
3527 2018-08-23 Richard Biener <rguenther@suse.de>
3530 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
3533 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
3535 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
3536 of single-vector TBLs.
3537 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
3540 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
3543 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
3544 aarch64_evpc_tbl guard.
3546 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
3548 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
3551 2018-08-22 Martin Sebor <msebor@redhat.com>
3554 * tree-pretty-print.c (pretty_print_string): Add argument.
3555 (dump_generic_node): Call to pretty_print_string with string size.
3557 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
3559 PR rtl-optimization/86771
3560 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
3561 of two SETs into those two SETs, one to be placed at i2, if that SETs
3562 destination is modified between i2 and i3.
3564 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3566 PR tree-optimization/86725
3567 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
3569 (vect_analyze_scalar_cycles_1): Check it.
3571 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3573 PR tree-optimization/86725
3574 * tree-vect-loop.c (vect_is_simple_reduction): When treating
3575 an outer loop phi as a double reduction, make sure that the
3576 single user of the phi result is an inner loop phi.
3578 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3580 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
3581 grouped stores with gaps to a strided group.
3583 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3585 * tree-vect-stmts.c (get_group_load_store_type)
3586 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
3587 first statement in a group.
3589 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3591 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
3592 the sequence used in gcc/gcc.c.
3594 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3597 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
3600 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3602 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
3603 Darwin10-specific unwinder-shim.
3604 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
3605 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
3606 New to cater for Darwin10 Rosetta.
3608 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3610 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
3611 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
3613 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3618 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
3619 (dwarf2out_switch_text_section): Generate a local label for the
3620 second function sub-section and apply it as the second FDE start
3622 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
3623 second sub-section start.
3625 2018-08-22 Richard Biener <rguenther@suse.de>
3627 PR tree-optimization/86988
3628 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
3630 2018-08-22 Richard Biener <rguenther@suse.de>
3632 PR tree-optimization/86945
3633 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
3635 2018-08-22 Alexandre Oliva <oliva@adacore.com>
3637 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
3638 a comment about how uses of r2 for .sdata2 come about.
3640 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
3642 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
3644 2018-08-21 Marek Polacek <polacek@redhat.com>
3646 PR c++/86981, Implement -Wpessimizing-move.
3647 * doc/invoke.texi: Document -Wpessimizing-move.
3649 2018-08-21 Jan Hubicka <jh@suse.cz>
3651 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
3652 * tree.h (is_redundant_typedef): Remove.
3653 * dwarf2out.c (is_redundant_typedef): Turn into static function.
3655 2018-08-21 Jan Hubicka <jh@suse.cz>
3657 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
3660 2018-08-21 Tamar Christina <tamar.christina@arm.com>
3662 * expmed.c (extract_low_bits): Reject invalid subregs early.
3664 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
3667 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
3670 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3672 * config/vxworks.h: Guard vxworks_asm_out_constructor and
3673 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
3674 * config/vxworks.c: Likewise.
3676 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3678 * config/vxworks.c: Set targetm.have_ctors_dtors
3679 if HAVE_INITFINI_ARRAY_SUPPORT.
3680 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
3681 if HAVE_INITFINI_ARRAY_SUPPORT.
3683 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3685 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
3686 default search path for VxWorks < 7.
3688 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
3690 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
3691 (get_format_string): Refer to c_getstr.
3693 2018-08-21 Tom de Vries <tdevries@suse.de>
3695 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3696 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3697 (debuginfo_early_stop): Declare.
3698 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3699 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3700 (debuginfo_early_stop): New function.
3701 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
3702 and debuginfo_early_stop.
3703 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
3704 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
3705 (general_init): Call debuginfo_early_init.
3706 (finalize): Call debuginfo_fini.
3707 (do_compile): Call debuginfo_init.
3708 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
3710 (@item -fdump-debug, @item -fdump-earlydebug): Add.
3712 2018-08-21 Tom de Vries <tdevries@suse.de>
3714 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
3715 flag_dump_noaddr and flag_dump_unnumbered.
3717 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
3719 * wide-int-range.cc (wide_int_range_abs): New.
3720 (wide_int_range_order_set): Rename from wide_int_range_min_max.
3721 * wide-int-range.h (wide_int_range_abs): New.
3722 (wide_int_range_min_max): New.
3723 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
3724 case to call wide_int_range_abs.
3725 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
3726 (extract_range_from_abs_expr): Delete.
3728 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
3731 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
3732 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
3735 2018-08-20 Nathan Sidwell <nathan@acm.org>
3736 Jeff Law <law@redhat.com>
3738 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
3739 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
3741 2018-08-20 David Malcolm <dmalcolm@redhat.com>
3744 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
3745 (decl_attributes): Likewise.
3746 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
3748 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3749 * diagnostic-core.h (class auto_diagnostic_group): New class.
3750 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
3751 (diagnostic_report_diagnostic): Handle the first diagnostics within
3753 (emit_diagnostic): Add auto_diagnostic_group instance.
3755 (inform_n): Likewise.
3756 (warning): Likewise.
3757 (warning_at): Likewise.
3758 (warning_n): Likewise.
3759 (pedwarn): Likewise.
3760 (permerror): Likewise.
3762 (error_n): Likewise.
3763 (error_at): Likewise.
3765 (fatal_error): Likewise.
3766 (internal_error): Likewise.
3767 (internal_error_no_backtrace): Likewise.
3768 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
3769 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
3770 * diagnostic.h (struct diagnostic_context): Add fields
3771 "diagnostic_group_nesting_depth",
3772 "diagnostic_group_emission_count", "begin_group_cb",
3774 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
3775 Add auto_diagnostic_group instance(s).
3776 (find_explicit_erroneous_behavior): Likewise.
3777 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
3778 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
3779 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
3780 (gimplify_va_arg_expr): Likewise.
3781 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
3782 (HSA_SORRY_AT): Likewise.
3783 * ipa-devirt.c (compare_virtual_tables): Likewise.
3784 (warn_odr): Likewise.
3785 * multiple_target.c (expand_target_clones): Likewise.
3786 * opts-common.c (cmdline_handle_error): Likewise.
3787 * reginfo.c (globalize_reg): Likewise.
3788 * substring-locations.c (format_warning_n_va): Likewise.
3789 * tree-inline.c (expand_call_inline): Likewise.
3790 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
3791 * tree-ssa-loop-niter.c
3792 (do_warn_aggressive_loop_optimizations): Likewise.
3793 * tree-ssa-uninit.c (warn_uninit): Likewise.
3794 * tree.c (warn_deprecated_use): Likewise.
3796 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
3799 * config/i386/i386.md (eh_return): Always update EH return
3800 address in word_mode.
3802 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
3804 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
3805 TARGET_SPLIT_COMPLEX_ARG is defined.
3807 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3809 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
3811 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3814 * expr.c (expand_assignment): Assert that bitpos is positive.
3815 (store_field): Likewise
3816 (expand_expr_real_1): Make sure that bitpos is positive.
3817 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
3820 2018-08-20 Nathan Sidwell <nathan@acm.org>
3822 * Makefile.in (CPP_ID_DATA_H): Delete.
3823 (CPP_INTERNAL_H): Don't add it.
3824 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
3825 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
3827 2018-08-20 Richard Biener <rguenther@suse.de>
3829 PR tree-optimization/78655
3830 * tree-vrp.c (extract_range_from_binary_expr_1): Make
3831 pointer + offset nonnull if either operand is nonnull work.
3833 2018-08-20 Tom de Vries <tdevries@suse.de>
3835 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
3836 unless the referenced die describes the added property using
3837 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
3838 Otherwise, add a DW_AT_location to the referenced die.
3840 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
3843 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
3844 register_operand when calling ix86_set_reg_reg_cost.
3845 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
3846 Set *total to 0 for operands that satisfy x86_64_immediate_operand
3847 predicate and to 1 otherwise.
3849 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
3851 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
3852 emit a diagnostic that it is not supported and reset the option.
3853 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
3854 supported and consume the option. (ASM_FINAL_SPEC): New.
3856 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3858 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
3861 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
3863 C-SKY port: Documentation
3865 * doc/extend.texi (C-SKY Function Attributes): New section.
3866 * doc/invoke.texi (Option Summary): Add C-SKY options.
3867 (C-SKY Options): New section.
3868 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
3870 2018-08-17 Jojo <jijie_rong@c-sky.com>
3871 Huibin Wang <huibin_wang@c-sky.com>
3872 Sandra Loosemore <sandra@codesourcery.com>
3873 Chung-Lin Tang <cltang@codesourcery.com>
3875 C-SKY port: Backend implementation
3877 * config/csky/*: New.
3878 * common/config/csky/*: New.
3880 2018-08-17 Jojo <jijie_rong@c-sky.com>
3881 Huibin Wang <huibin_wang@c-sky.com>
3882 Sandra Loosemore <sandra@codesourcery.com>
3883 Chung-Lin Tang <cltang@codesourcery.com>
3884 Andrew Jenner <andrew@codesourcery.com>
3886 C-SKY port: Configury
3888 * config.gcc (csky-*-*): New.
3889 * configure.ac: Add csky to targets for dwarf2 debug_line support.
3890 * configure: Regenerated.
3892 2018-08-17 David Malcolm <dmalcolm@redhat.com>
3894 * dump-context.h: Include "dumpfile.h".
3895 (dump_context::dump_printf_va): Convert final param from va_list
3896 to va_list *. Convert from ATTRIBUTE_PRINTF to
3897 ATTRIBUTE_GCC_DUMP_PRINTF.
3898 (dump_context::dump_printf_loc_va): Likewise.
3899 * dumpfile.c: Include "stringpool.h".
3900 (make_item_for_dump_printf_va): Delete.
3901 (make_item_for_dump_printf): Delete.
3902 (class dump_pretty_printer): New class.
3903 (dump_pretty_printer::dump_pretty_printer): New ctor.
3904 (dump_pretty_printer::emit_items): New member function.
3905 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
3907 (dump_pretty_printer::emit_item): New member function.
3908 (dump_pretty_printer::stash_item): New member function.
3909 (dump_pretty_printer::format_decoder_cb): New member function.
3910 (dump_pretty_printer::decode_format): New member function.
3911 (dump_context::dump_printf_va): Reimplement in terms of
3912 dump_pretty_printer.
3913 (dump_context::dump_printf_loc_va): Convert final param from va_list
3915 (dump_context::begin_scope): Reimplement call to
3916 make_item_for_dump_printf.
3917 (dump_printf): Update for change to dump_printf_va.
3918 (dump_printf_loc): Likewise.
3919 (selftest::test_capture_of_dump_calls): Convert "stmt" from
3920 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
3921 with %T, %E, and %G.
3922 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
3923 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
3924 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
3925 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
3926 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
3927 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
3928 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
3929 within a dump_printf_loc call to "%wu".
3930 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
3931 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
3932 missing space after "=".
3933 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
3934 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
3935 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
3936 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
3937 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
3938 duplicate "vectorized" from message.
3940 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
3942 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
3943 polyNxK_t element's TYPE_STRING_FLAG.
3945 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3947 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
3948 (they were unnamed before). Fix comments.
3950 2018-08-17 Nathan Sidwell <nathan@acm.org>
3952 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
3954 2018-08-17 Richard Biener <rguenther@suse.de>
3956 PR tree-optimization/86841
3957 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
3959 2018-08-17 Martin Liska <mliska@suse.cz>
3961 * common.opt: Remove Warn, Init and Report for options with
3962 Ignore/Deprecated flag. Warning is done automatically for
3964 * config/i386/i386.opt: Likewise.
3965 * config/ia64/ia64.opt: Likewise.
3966 * config/rs6000/rs6000.opt: Likewise.
3967 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
3968 Remove usage of flag_check_pointer_bounds.
3969 * lto-wrapper.c (merge_and_complain): Do not handle
3970 OPT_fcheck_pointer_bounds.
3971 (append_compiler_options): Likewise.
3972 * opt-functions.awk: Do not handle Deprecated.
3973 * optc-gen.awk: Check that Var, Report and Init are not
3974 used for an option with Ignore/Deprecated flag.
3975 * opts-common.c (decode_cmdline_option): Do not report
3977 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
3979 * opts.h (struct cl_option): Remove cl_deprecated flag.
3980 (CL_ERR_DEPRECATED): Remove error enum value.
3982 2018-08-17 Richard Biener <rguenther@suse.de>
3985 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
3986 across a va-arg-pack using call adjust its return value accordingly.
3988 2018-08-16 Martin Sebor <msebor@redhat.com>
3990 PR tree-optimization/86853
3991 * gimple-ssa-sprintf.c (struct format_result): Rename member.
3992 (struct fmtresult): Add member and initialize it in ctors.
3993 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
3994 (format_string): Handle %S the same as %ls. Set MAYFAIL.
3995 (format_directive): Set POSUNDER4K when MAYFAIL is set.
3996 (parse_directive): Handle %C same as %c.
3997 (sprintf_dom_walker::compute_format_length): Adjust.
3998 (is_call_safe): Adjust.
4000 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4002 * builtins.c (c_strlen): Add new parameter eltsize. Use it
4003 for determining how to count the elements.
4004 * builtins.h (c_strlen): Adjust prototype.
4005 * expr.c (string_constant): Add new parameter mem_size.
4006 Set *mem_size appropriately.
4007 * expr.h (string_constant): Adjust protoype.
4008 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4009 * gimple-fold.h (get_range_strlen): Adjust prototype.
4010 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4011 (format_string): Call get_string_length with eltsize.
4013 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4015 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4016 to emit the span, rather than setting it as the prefix.
4018 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4020 * diagnostic-show-locus.c (layout::start_annotation_line): Add
4021 "margin_char" parameter, defaulting to space. Use it in place
4022 of pp_space for the initial part of the margin.
4023 (layout::print_leading_fixits): Use '+' when filling the margin
4024 of line-insertion fix-it hints.
4026 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4028 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4031 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4033 * config/rs6000/altivec.md: Don't set length attribute to the default
4035 * config/rs6000/darwin.md: Ditto.
4036 * config/rs6000/dfp.md: Ditto.
4037 * config/rs6000/htm.md: Ditto.
4038 * config/rs6000/rs6000.md: Ditto.
4039 * config/rs6000/sync.md: Ditto.
4040 * config/rs6000/vsx.md: Ditto.
4042 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4044 * config/rs6000/altivec.md: Don't set length attribute to the default
4045 value, for branch instructions.
4046 * config/rs6000/darwin.md: Ditto.
4047 * config/rs6000/rs6000.md: Ditto.
4049 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4051 * config/rs6000/rs6000.md (length): Always define as const_int 4.
4052 (unnamed conditional branch define_insn): Set length to 4 or 8
4053 depending on offset.
4054 (<bd>_<mode>): Similar, for alternative 0.
4055 (<bd>tf_<mode>): Ditto.
4057 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4059 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4061 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
4063 * doc/rtl.texi: Replace old RTX class names with new names.
4066 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
4068 * expmed.h (canonicalize_comparison): New declaration.
4069 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4070 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4071 * optabs.c (prepare_cmp_insn): Likewise.
4072 * rtl.h (unsigned_condition_p): New function which checks if a
4073 comparison operator is unsigned.
4075 2018-08-16 Nathan Sidwell <nathan@acm.org>
4077 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4078 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4080 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4083 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4084 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4085 (mov<mov>): ..this and enable unconditionally.
4087 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4089 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4091 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
4093 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4094 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4095 "Common" with "Target".
4097 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4099 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4100 * doc/invoke.texi (mmitigate-rop): Remove.
4101 * config/i386/i386.c: Do not include "regrename.h".
4102 (ix86_rop_should_change_byte_p, reg_encoded_number)
4103 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4105 (ix86_reorg): Remove call to ix86_mitigate_rop.
4106 * config/i386/i386.md (attr "modrm_class"): Remove.
4107 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4108 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4109 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4111 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
4113 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4114 allow folding of mergeh() and mergel() for the float and double types.
4115 (fold_mergehl_helper): Rework to handle building a permute tree
4118 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4120 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4123 2018-08-15 David Malcolm <dmalcolm@redhat.com>
4125 * common.opt (fdiagnostics-show-labels): New option.
4126 * diagnostic-show-locus.c (class layout_range): Add field
4128 (class layout): Add field "m_show_labels_p".
4129 (layout_range::layout_range): Add param "label" and use it to
4131 (make_range): Pass in NULL for new "label" param of layout_range's
4133 (layout::layout): Initialize m_show_labels_p.
4134 (layout::maybe_add_location_range): Pass in loc_range->m_label
4135 when constructing layout_range instances.
4136 (struct line_label): New struct.
4137 (layout::print_any_labels): New member function.
4138 (layout::print_line): Call it if label-printing is enabled.
4139 (selftest::test_one_liner_labels): New test.
4140 (selftest::test_diagnostic_show_locus_one_liner): Call it.
4141 * diagnostic.c (diagnostic_initialize): Initialize
4142 context->show_labels_p.
4143 * diagnostic.h (struct diagnostic_context): Add field
4145 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4146 -fno-diagnostics-show-labels.
4147 * dwarf2out.c (gen_producer_string): Add
4148 OPT_fdiagnostics_show_labels to the ignored options.
4149 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4151 (gcc_rich_location::maybe_add_expr): Likewise.
4152 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4153 label" param, defaulting to NULL.
4154 (gcc_rich_location::add_expr): Add "label" param.
4155 (gcc_rich_location::maybe_add_expr): Likewise.
4156 (class text_range_label): New class.
4157 (class range_label_for_type_mismatch): New class.
4158 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4159 of format_warning_va.
4160 (fmtwarn_n): Likewise for new params of format_warning_n_va.
4161 * lto-wrapper.c (merge_and_complain): Add
4162 OPT_fdiagnostics_show_labels to the "pick one setting" options.
4163 (append_compiler_options): Likewise to the dropped options.
4164 (append_diag_options): Likewise to the passed-on options.
4165 * opts.c (common_handle_option): Handle the new option.
4166 * selftest-diagnostic.c
4167 (test_diagnostic_context::test_diagnostic_context): Enable
4169 * substring-locations.c: Include "gcc-rich-location.h".
4170 (format_warning_n_va): Add "fmt_label" and "param_label" params
4171 and use them as appropriate.
4172 (format_warning_va): Add "fmt_label" and "param_label" params,
4173 passing them on to format_warning_n_va.
4174 (format_warning_at_substring): Likewise.
4175 (format_warning_at_substring_n): Likewise.
4176 * substring-locations.h (format_warning_va): Add "fmt_label" and
4177 "param_label" params.
4178 (format_warning_n_va): Likewise.
4179 (format_warning_at_substring): Likewise.
4180 (format_warning_at_substring_n): Likewise.
4181 * toplev.c (general_init): Initialize global_dc->show_labels_p.
4183 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
4186 * builtins.c (expand_builtin_memcmp): Do not expand the call
4187 when overflow is detected.
4189 2018-08-15 Martin Sebor <msebor@redhat.com>
4191 PR tree-optimization/71625
4192 * config/aarch64/aarch64-builtins.c
4193 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4195 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
4197 * config/s390/s390.c (s390_reorg): Remove loop.
4199 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4202 (darwin_function_switched_text_sections): Delete.
4203 * gcc/config/darwin.h
4204 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4206 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4209 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4210 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
4211 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4213 2018-08-15 Martin Liska <mliska@suse.cz>
4215 PR tree-optimization/86925
4216 * predict.c (expr_expected_value_1): When taking
4217 later predictor, assign also probability.
4218 Use fold_build2_initializer_loc in order to fold
4219 the expression in -frounding-math.
4221 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
4223 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4225 (expand_vec_perm_1): Try the new method.
4227 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
4230 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4231 Check whether lra_live_max_point is 0 before dividing.
4233 2018-08-14 Martin Sebor <msebor@redhat.com>
4235 PR tree-optimization/86650
4236 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4237 (vrp_prop::check_mem_ref): Same.
4239 2018-08-13 Liu Hao <lh_mouse@126.com>
4241 * pretty-print.c (eat_esc_sequence): Swap the foreground and
4242 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4243 and clear it thereafter, as it only works for DBCS.
4245 2018-08-13 Liu Hao <lh_mouse@126.com>
4247 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4248 handle returned by _get_osf_handle().
4250 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4252 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4253 for folding vec_perm.
4255 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4257 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4258 Add support for gimple-folding of vec_pack() and vec_unpack()
4261 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4263 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4264 vec_xst variants to the list.
4265 (rs6000_gimple_fold_builtin): Add support for folding unaligned
4266 vector loads and stores.
4268 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
4270 * config.gcc (rs6000-ibm-aix4.x): Delete.
4271 (rs6000-ibm-aix5.1): Delete.
4272 (rs6000-ibm-aix5.2): Delete.
4273 (rs6000-ibm-aix5.3): Delete.
4274 * config/rs6000/aix43.h: Delete.
4275 * config/rs6000/aix51.h: Delete.
4276 * config/rs6000/aix52.h: Delete.
4277 * config/rs6000/t-aix43: Delete.
4279 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
4281 * config/s390/s390.c (s390_decompose_constant_pool_ref):
4283 (s390_decompose_address): Factor out constant pool ref
4286 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4288 * config/nds32/nds32-predicates.c
4289 (nds32_can_use_bclr_p): Change return type as bool.
4290 (nds32_can_use_bset_p): Ditto.
4291 (nds32_can_use_btgl_p): Ditto.
4292 (nds32_can_use_bitci_p): Ditto.
4293 * config/nds32/nds32-protos.h
4294 (nds32_can_use_bclr_p): Change declaration.
4295 (nds32_can_use_bset_p): Ditto.
4296 (nds32_can_use_btgl_p): Ditto.
4297 (nds32_can_use_bitci_p): Ditto.
4299 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4301 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4302 Support -msched-prolog-epilog option.
4303 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4305 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4307 * common/config/nds32/nds32-common.c
4308 (nds32_option_optimization_table): Enalbe -malways-align.
4310 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4312 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4314 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4315 OPT_misr_secure_ case.
4316 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4317 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4318 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4320 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4321 (nds32_isr_info): New field security_level.
4322 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4323 * config/nds32/nds32.md (return_internal): Consider critical attribute.
4324 * config/nds32/nds32.opt (misr-secure): New option.
4325 * config/nds32/nds32_init.inc: New file.
4326 * config/nds32/nds32_isr.h: New file.
4328 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
4330 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4331 Update comment for atomic instructions.
4332 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4333 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
4335 (atomic_loaddi): Revise fence expansion to only emit fence prior to
4336 load for __ATOMIC_SEQ_CST model.
4337 (atomic_loaddi_1): Remove float register target.
4338 (atomic_storedi): Handle CONST_INT values.
4339 (atomic_storedi_1): Remove float register source. Add special case
4341 (memory_barrier): New expander and insn.
4343 2018-08-11 Jakub Jelinek <jakub@redhat.com>
4345 PR tree-optimization/86835
4346 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
4347 new_stmt after def_gsi, make sure to insert new_square_stmt after
4348 that stmt, not 2 stmts before it.
4350 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
4353 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
4356 2018-08-10 Martin Liska <mliska@suse.cz>
4359 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
4361 * builtins.c (expand_builtin_expect_with_probability):
4363 (expand_builtin_expect_with_probability): New function.
4364 (build_builtin_expect_predicate): Add new argumnet probability
4365 for BUILT_IN_EXPECT_WITH_PROBABILITY.
4366 (fold_builtin_expect):
4369 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
4370 * builtins.h (fold_builtin_expect): Set new argument.
4371 * doc/extend.texi: Document __builtin_expect_with_probability.
4372 * doc/invoke.texi: Likewise.
4373 * gimple-fold.c (gimple_fold_call): Pass new argument.
4374 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
4375 also BUILT_IN_EXPECT_WITH_PROBABILITY.
4376 * predict.c (get_predictor_value): New function.
4377 (expr_expected_value): Add new argument probability. Assume
4378 that predictor and probability are always non-null.
4379 (expr_expected_value_1): Likewise. For __builtin_expect and
4380 __builtin_expect_with_probability set probability. Handle
4381 combination in binary expressions.
4382 (tree_predict_by_opcode): Simplify code by simply calling
4383 get_predictor_value.
4384 (pass_strip_predict_hints::execute): Add handling of
4385 BUILT_IN_EXPECT_WITH_PROBABILITY.
4386 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
4388 * tree.h (DECL_BUILT_IN_P): New function.
4390 2018-08-10 Martin Liska <mliska@suse.cz>
4392 PR tree-optimization/85799
4393 * passes.def: Add argument for pass_strip_predict_hints.
4394 * predict.c (class pass_strip_predict_hints): Add new argument
4396 (strip_predictor_early): New function.
4397 (pass_strip_predict_hints::execute): Call the function to
4399 (strip_predict_hints): New function.
4400 * predict.def: Fix comment.
4402 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
4404 * Makefile.in: Clarify which tm.texi to copy over to assert the
4405 right to grant a GFDL license for all.
4407 2018-08-09 Jeff Law <law@redhat.com>
4409 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
4412 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
4414 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
4417 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4419 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
4420 reductions for variable-length vectors.
4422 2018-08-09 David Malcolm <dmalcolm@redhat.com>
4425 * common.opt (fdiagnostics-show-line-numbers): New option.
4426 * diagnostic-show-locus.c (class layout): Add fields
4427 "m_show_line_numbers_p" and "m_linenum_width";
4428 (num_digits): New function.
4429 (test_num_digits): New function.
4430 (layout::layout): Initialize new fields. Update m_x_offset
4431 logic to handle any left margin.
4432 (layout::print_source_line): Print line number when requested.
4433 (layout::start_annotation_line): New member function.
4434 (layout::print_annotation_line): Call it.
4435 (layout::print_leading_fixits): Likewise.
4436 (layout::print_trailing_fixits): Likewise. Update calls to
4437 move_to_column for new parameter.
4438 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
4439 it to potentially call start_annotation_line.
4440 (layout::show_ruler): Call start_annotation_line.
4441 (selftest::test_line_numbers_multiline_range): New selftest.
4442 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
4443 and selftest::test_line_numbers_multiline_range.
4444 * diagnostic.c (diagnostic_initialize): Initialize
4445 show_line_numbers_p.
4446 * diagnostic.h (struct diagnostic_context): Add field
4447 "show_line_numbers_p".
4448 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4449 -fno-diagnostics-show-line-numbers.
4450 * dwarf2out.c (gen_producer_string): Add
4451 OPT_fdiagnostics_show_line_numbers to the ignored options.
4452 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
4453 one setting" options.
4454 (append_compiler_options): Likewise to the dropped options.
4455 (append_diag_options): Likewise to the passed-on options.
4456 * opts.c (common_handle_option): Handle the new option.
4457 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
4459 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
4461 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
4462 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
4463 third argument of type "const signed char" to descriptions of
4464 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
4465 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
4466 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
4467 __builtin_bcdsub_ov functions.
4469 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4471 PR tree-optimization/86858
4472 * tree-vect-loop.c (vect_is_simple_reduction): Restore
4473 flow_bb_inside_loop_p calls.
4475 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4477 PR tree-optimization/86871
4478 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
4479 instead of gimple_assign_lhs.
4481 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
4484 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
4485 register constraint to operand 0.
4486 (add<mode>3_carryinC): Likewise.
4487 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
4489 2018-08-09 Martin Liska <mliska@suse.cz>
4492 * common.opt: Remove extra line.
4494 2018-08-09 Martin Liska <mliska@suse.cz>
4496 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
4497 at the end of a line, make first letter capital and end up
4498 a sentence with a dot.
4499 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
4500 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
4501 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
4502 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
4503 (PARAM_MAX_ISL_OPERATIONS): Likewise.
4504 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
4505 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
4506 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
4507 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
4508 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
4509 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
4510 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
4511 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
4512 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
4513 (PARAM_TREE_REASSOC_WIDTH): Likewise.
4514 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
4515 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
4516 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
4518 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
4521 * config/s390/s390.c (s390_option_override_internal): Reduce the
4522 stack-clash-protection-probe-interval param if it would be too big
4525 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
4528 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
4529 * config/m68k/m68k.c (handle_move_double): Don't call
4530 m68k_final_prescan_insn.
4531 (m68k_adjust_decorated_operand): Renamed from
4532 m68k_final_prescan_insn, remove first and third operand and
4534 (print_operand): Call it.
4535 (print_operand_address): Call it.
4537 2018-08-08 Nathan Sidwell <nathan@acm.org>
4539 * diagnostic.c (diagnostic_report_current_module): Use
4540 linemap_included_from & linemap_included_from_linemap.
4542 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
4544 * config/aarch64/aarch64-cores.def: Add phecda core.
4545 * config/aarch64/aarch64-tune.md: Regenerate.
4546 * doc/invoke.texi: Add phecda core.
4548 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
4551 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
4553 * config/s390/s390.md ("movti"): Add more alternatives for
4554 constant to GPR copies.
4556 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
4558 * config/s390/s390.c: Fix whitespace damage throughout the file.
4559 * config/s390/s390.h: Likewise.
4560 * config/s390/tpf.h: Likewise.
4562 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
4564 * config/s390/s390.c (s390_loadrelative_operand_p):
4565 Remove TARGET_CPU_ZARCH usages.
4566 (s390_rtx_costs): Likewise.
4567 (s390_legitimate_constant_p): Likewise.
4568 (s390_cannot_force_const_mem): Likewise.
4569 (legitimate_reload_constant_p): Likewise.
4570 (s390_preferred_reload_class): Likewise.
4571 (legitimize_pic_address): Likewise.
4572 (legitimize_tls_address): Likewise.
4573 (s390_split_branches): Removed.
4574 (s390_add_execute): Removed.
4575 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
4576 (s390_mainpool_start): Likewise.
4577 (s390_mainpool_finish): Likewise.
4578 (s390_mainpool_cancel): Removed.
4579 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
4580 (s390_chunkify_cancel): Likewise.
4581 (s390_return_addr_rtx): Likewise.
4582 (s390_register_info): Remove split_branches_pending_p uages.
4583 (s390_optimize_register_info): Likewise.
4584 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
4585 split_branches_pending_p usages.
4586 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
4587 (s390_load_got): Likewise.
4588 (s390_expand_split_stack_prologue): Likewise.
4589 (output_asm_nops): Likewise.
4590 (s390_function_profiler): Likewise.
4591 (s390_emit_call): Likewise.
4592 (s390_conditional_register_usage): Likewise.
4593 (s390_optimize_prologue): Likewise.
4594 (s390_reorg): Remove TARGET_CPU_ZARCH and
4595 split_branches_pending_p usages.
4596 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
4598 (s390_output_indirect_thunk_function): Likewise.
4599 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
4600 (TARGET_CPU_ZARCH_P): Removed.
4601 (struct machine_function): Remove split_branches_pending_p.
4602 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
4604 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
4606 * common/config/s390/s390-common.c (processor_flags_table):
4608 * config.gcc: Remove with_arch/with_tune support.
4609 * config/s390/2064.md: Remove cpu attribute comparisons.
4610 * config/s390/driver-native.c (s390_host_detect_local_cpu):
4612 * config/s390/linux.h (ASM_SPEC):
4613 Remove -march support.
4614 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
4615 Use a table to get an arch level.
4616 * config/s390/s390-opts.h (enum processor_type):
4618 * config/s390/s390.c
4619 (processor_table): Remove entries, add arch_level values.
4620 (s390_issue_rate): Remove cases.
4621 (s390_option_override): Adjust
4622 s390_option_override_internal() call.
4623 (s390_option_override_internal): Remove deprecation warning.
4624 (s390_valid_target_attribute_tree): Adjust
4625 s390_option_override_internal() call.
4626 * config/s390/s390.h (struct s390_processor):
4627 Share with s390-c.c, add arch_level field.
4628 * config/s390/s390.md:
4629 Remove occurrences in cpu attribute.
4630 * config/s390/s390.opt: Remove -march/-mtune support.
4631 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
4632 * doc/invoke.texi: Remove deprecation warning.
4634 2018-08-08 Luis Machado <luis.machado@linaro.org>
4636 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
4638 (qdf24xx_tunings): Set vector cost structure to
4639 qdf24xx_vector_cost.
4641 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
4642 <register_sextend>: Set to 3.
4644 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
4647 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
4648 * config/aarch64/aarch64-simd.md
4649 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
4650 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
4651 (aarch64_frecpx<mode>): ...this new pattern.
4652 * config/aarch64/aarch64-simd-builtins.def: Remove comment
4653 about aarch64_frecp<FRECP:frecp_suffix><mode>.
4655 2018-08-07 Martin Liska <mliska@suse.cz>
4658 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
4659 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
4660 * predict.def (PRED_MALLOC_NONNULL): New predictor.
4661 * doc/extend.texi: Document that malloc attribute adds
4664 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
4667 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4668 Define to speculation_safe_value_not_needed.
4670 2018-08-06 Jeff Law <law@redhat.com>
4672 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
4673 the vr_values instance to cprop_into_stmt.
4674 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
4675 (cprop_operand): Also query EVRP to determine if OP is a constant.
4677 2018-08-06 Nathan Sidwell <nathan@acm.org>
4679 * diagnostic.c (diagnostic_report_current_module): Reroll
4680 included-at loop. Translate text.
4682 2018-08-06 David Malcolm <dmalcolm@redhat.com>
4684 * function-tests.c (selftest::test_expansion_to_rtl): Call
4685 free_after_compilation.
4687 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4689 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
4691 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
4693 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
4694 loops with memory block operations from getting unrolled.
4696 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
4699 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4700 Define to speculation_safe_value_not_needed.
4702 2018-08-06 Jeff Law <law@redhat.com>
4704 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
4707 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4710 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
4711 with all enabled __intN types.
4713 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
4715 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4717 * alias.c (record_set): Check for clobber high.
4718 * cfgexpand.c (expand_gimple_stmt): Likewise.
4719 * combine-stack-adj.c (single_set_for_csa): Likewise.
4720 * combine.c (find_single_use_1): Likewise.
4721 (set_nonzero_bits_and_sign_copies): Likewise.
4722 (get_combine_src_dest): Likewise.
4723 (is_parallel_of_n_reg_sets): Likewise.
4724 (try_combine): Likewise.
4725 (record_dead_and_set_regs_1): Likewise.
4726 (reg_dead_at_p_1): Likewise.
4727 (reg_dead_at_p): Likewise.
4728 * dce.c (deletable_insn_p): Likewise.
4729 (mark_nonreg_stores_1): Likewise.
4730 (mark_nonreg_stores_2): Likewise.
4731 * df-scan.c (df_find_hard_reg_defs): Likewise.
4732 (df_uses_record): Likewise.
4733 (df_get_call_refs): Likewise.
4734 * dwarf2out.c (mem_loc_descriptor): Likewise.
4735 * haifa-sched.c (haifa_classify_rtx): Likewise.
4736 * ira-build.c (create_insn_allocnos): Likewise.
4737 * ira-costs.c (scan_one_insn): Likewise.
4738 * ira.c (equiv_init_movable_p): Likewise.
4739 (rtx_moveable_p): Likewise.
4740 (interesting_dest_for_shprep): Likewise.
4741 * jump.c (mark_jump_label_1): Likewise.
4742 * postreload-gcse.c (record_opr_changes): Likewise.
4743 * postreload.c (reload_cse_simplify): Likewise.
4744 (struct reg_use): Add source expr.
4745 (reload_combine): Check for clobber high.
4746 (reload_combine_note_use): Likewise.
4747 (reload_cse_move2add): Likewise.
4748 (move2add_note_store): Likewise.
4749 * print-rtl.c (print_pattern): Likewise.
4750 * recog.c (decode_asm_operands): Likewise.
4751 (store_data_bypass_p): Likewise.
4752 (if_test_bypass_p): Likewise.
4753 * regcprop.c (kill_clobbered_value): Likewise.
4754 (kill_set_value): Likewise.
4755 * reginfo.c (reg_scan_mark_refs): Likewise.
4756 * reload1.c (maybe_fix_stack_asms): Likewise.
4757 (eliminate_regs_1): Likewise.
4758 (elimination_effects): Likewise.
4759 (mark_not_eliminable): Likewise.
4760 (scan_paradoxical_subregs): Likewise.
4761 (forget_old_reloads_1): Likewise.
4762 * reorg.c (find_end_label): Likewise.
4763 (try_merge_delay_insns): Likewise.
4764 (redundant_insn): Likewise.
4765 (own_thread_p): Likewise.
4766 (fill_simple_delay_slots): Likewise.
4767 (fill_slots_from_thread): Likewise.
4768 (dbr_schedule): Likewise.
4769 * resource.c (update_live_status): Likewise.
4770 (mark_referenced_resources): Likewise.
4771 (mark_set_resources): Likewise.
4772 * rtl.c (copy_rtx): Likewise.
4773 * rtlanal.c (reg_referenced_p): Likewise.
4774 (single_set_2): Likewise.
4775 (noop_move_p): Likewise.
4776 (note_stores): Likewise.
4777 * sched-deps.c (sched_analyze_reg): Likewise.
4778 (sched_analyze_insn): Likewise.
4780 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4782 * cse.c (invalidate_reg): New function extracted from...
4783 (invalidate): ...here.
4784 (canonicalize_insn): Check for clobber high.
4785 (invalidate_from_clobbers): invalidate clobber highs.
4786 (invalidate_from_sets_and_clobbers): Likewise.
4787 (count_reg_usage): Check for clobber high.
4788 (insn_live_p): Likewise.
4789 * cselib.c (cselib_expand_value_rtx_1):Likewise.
4790 (cselib_invalidate_regno): Check for clobber in setter.
4791 (cselib_invalidate_rtx): Pass through setter.
4792 (cselib_invalidate_rtx_note_stores):
4793 (cselib_process_insn): Check for clobber high.
4794 * cselib.h (cselib_invalidate_rtx): Add operand.
4796 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4798 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
4799 (mark_not_eliminable): Likewise.
4800 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
4801 * lra-lives.c (process_bb_lives): Check for clobber high.
4802 * lra.c (new_insn_reg): Remember clobber highs.
4803 (collect_non_operand_hard_regs): Check for clobber high.
4804 (lra_set_insn_recog_data): Likewise.
4805 (add_regs_to_insn_regno_info): Likewise.
4806 (lra_update_insn_regno_info): Likewise.
4808 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4810 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
4811 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
4813 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4815 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
4816 (copy_insn_1): Likewise.
4817 (gen_hard_reg_clobber_high): New gen function.
4818 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
4819 * genemit.c (gen_exp): Likewise.
4820 (gen_emit_seq): Pass through info.
4821 (gen_insn): Check for CLOBBER_HIGH.
4822 (gen_expand): Pass through info.
4823 (gen_split): Likewise.
4824 (output_add_clobbers): Likewise.
4825 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
4826 (remove_clobbers): Likewise.
4827 * rtl.h (gen_hard_reg_clobber_high): New declaration.
4829 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4831 * doc/rtl.texi (clobber_high): Add.
4832 (parallel): Add in clobber high
4833 * rtl.c (rtl_check_failed_code3): Add function.
4834 * rtl.def (CLOBBER_HIGH): Add expression.
4835 * rtl.h (RTL_CHECKC3): Add macro.
4836 (rtl_check_failed_code3): Add declaration.
4837 (XC3EXP): Add macro.
4839 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
4842 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
4843 cfun->machine->max_used_stack_alignment if needed.
4845 2018-08-04 Martin Sebor <msebor@redhat.com>
4847 PR tree-optimization/86571
4848 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
4851 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
4854 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4857 2018-08-03 Jeff Law <law@redhat.com>
4860 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4861 Define to speculation_safe_value_not_needed.
4863 2018-08-03 David Malcolm <dmalcolm@redhat.com>
4865 * doc/gcov.texi (-x): Remove duplicate "to".
4866 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
4867 (-Wif-not-aligned): Remove duplicate "is".
4868 (-flto): Remove duplicate "the".
4869 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
4870 duplicate "v5.00.b".
4871 (MSP430 Options): Remove duplicate "and" from the description
4872 of "-mgprel-sec=regexp".
4873 (x86 Options): Remove duplicate copies of "vmldLog102" and
4874 vmlsLog104 from description of "-mveclibabi=type".
4876 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
4878 * internal-fn.h (first_commutative_argument): Declare.
4879 * internal-fn.c (first_commutative_argument): New function.
4880 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
4881 restrictions for pattern statements. Use first_commutative_argument
4882 to look for commutative operands in calls to internal functions.
4884 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
4886 * Makefile.in (wide-int-range.o): New.
4887 * tree-vrp.c: Move all the wide_int_* functions to...
4888 * wide-int-range.cc: ...here.
4889 * tree-vrp.h: Move all the wide_int_* prototypes to...
4890 * wide-int-range.h: ...here.
4892 2018-08-03 Tom de Vries <tdevries@suse.de>
4894 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
4896 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
4897 * except.c (output_function_exception_table): Do early exit if
4898 targetm_common.except_unwind_info (&global_options) == UI_NONE.
4900 2018-08-03 Martin Liska <mliska@suse.cz>
4902 * predict.c (dump_prediction): Change to 2 digits
4905 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
4907 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
4909 (falkor_am_1_gtov_gtov): ... a new insn reservation.
4911 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
4913 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
4914 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
4915 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
4916 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
4917 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
4919 2018-08-02 David Malcolm <dmalcolm@redhat.com>
4921 * diagnostic-show-locus.c (diagnostic_show_locus): Use
4922 pp_take_prefix when saving the existing prefix.
4923 * diagnostic.c (diagnostic_append_note): Likewise.
4924 * langhooks.c (lhd_print_error_function): Likewise.
4925 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
4926 param's type. Free the existing prefix.
4927 (pp_take_prefix): New function.
4928 (pretty_printer::pretty_printer): Drop the prefix parameter.
4929 Rename the length parameter to match the comment.
4930 (pretty_printer::~pretty_printer): Free the prefix.
4931 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
4933 (struct pretty_printer): Drop the "const" from "prefix" field's
4934 type and clarify memory management.
4935 (pp_set_prefix): Drop the "const" from the 2nd param.
4936 (pp_take_prefix): New decl.
4938 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4940 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
4941 for word_mode_ok here instead of passing as argument.
4942 (expand_block_compare): Change select_block_compare_mode() call.
4943 (expand_strncmp_gpr_sequence): New function.
4944 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
4946 2018-08-02 Jeff Law <law@redhat.com>
4949 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4950 Define to speculation_safe_value_not_needed.
4953 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4954 Define to speculation_safe_value_not_needed.
4956 2018-08-02 Tom de Vries <tdevries@suse.de>
4959 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
4960 function. Return UI_TARGET unconditionally.
4961 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
4962 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
4964 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4966 * genemit.c (print_overload_test): Fix typo.
4968 2018-08-02 Richard Biener <rguenther@suse.de>
4970 PR tree-optimization/86816
4971 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
4972 which checks for value availability before querying it.
4973 (gvn_uses_equal): Use it.
4974 (same_succ_hash): Likewise.
4975 (gimple_equal_p): Likewise.
4977 2018-08-02 Nick Clifton <nickc@redhat.com>
4980 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4981 Define to speculation_safe_value_not_needed.
4984 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4985 Define to speculation_safe_value_not_needed.
4988 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4989 Define to speculation_safe_value_not_needed.
4992 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4993 Define to speculation_safe_value_not_needed.
4996 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4997 Define to speculation_safe_value_not_needed.
5000 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5001 Define to speculation_safe_value_not_needed.
5004 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5005 Define to speculation_safe_value_not_needed.
5008 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5009 Define to speculation_safe_value_not_needed.
5012 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5013 speculation_safe_value_not_needed.
5016 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5017 to speculation_safe_value_not_needed.
5019 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5021 * doc/md.texi: Expand the documentation of instruction names
5022 to mention port-local uses. Document '@' in pattern names.
5023 * read-md.h (overloaded_instance, overloaded_name): New structs.
5025 (md_reader::handle_overloaded_name): New member function.
5026 (md_reader::get_overloads): Likewise.
5027 (md_reader::m_first_overload): New member variable.
5028 (md_reader::m_next_overload_ptr): Likewise.
5029 (md_reader::m_overloads_htab): Likewise.
5030 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5031 m_next_overload_ptr and m_overloads_htab.
5032 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5033 (get_mode_token, get_code_token, get_int_token): New functions.
5034 (map_attr_string): Add an optional argument that passes back
5035 the associated iterator.
5036 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5037 (md_reader::handle_overloaded_name, add_overload_instance): New
5039 (apply_iterators): Handle '@' names. Report an error if '@'
5040 is used without iterators.
5041 (initialize_iterators): Initialize the new iterator_group fields.
5042 * genopinit.c (handle_overloaded_code_for)
5043 (handle_overloaded_gen): New functions.
5044 (main): Use them to print declarations of maybe_code_for_* and
5045 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5046 * genemit.c (print_overload_arguments, print_overload_test)
5047 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5048 (main): Use it to print definitions of maybe_code_for_* and
5049 maybe_gen_* functions.
5050 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5051 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5052 instead of explicit mode checks.
5053 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5054 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5055 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5056 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5057 (aarch64_expand_compare_and_swap): Likewise
5058 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5059 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5060 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5061 (aarch64_constant_pool_reload_icode): Delete.
5062 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5063 instead of aarch64_constant_pool_reload_icode. Use
5064 code_for_aarch64_reload_mov instead of explicit mode checks.
5065 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5066 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5067 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5068 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5069 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5070 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5071 (aarch64_atomic_load_op_code): Delete.
5072 (aarch64_emit_atomic_load_op): Likewise.
5073 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5074 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
5075 instead of aarch64_emit_atomic_load_op.
5076 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5077 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5078 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5079 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5080 character before the pattern name.
5081 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5082 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5083 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5084 (aarch64_frecps<mode>): Likewise.
5085 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5086 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5087 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5088 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5089 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5091 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5093 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5094 Allow HFmode constants if TARGET_FP_F16INST.
5096 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
5099 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5100 No longer check last store for clobber of address register.
5102 2018-08-02 Martin Liska <mliska@suse.cz>
5104 PR gcov-profile/86817
5105 * gcov.c (process_all_functions): New function.
5107 (process_file): Move functions processing to
5108 process_all_functions.
5110 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5112 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5113 "const" to the "gimple *" and "rtx_insn *" parameters.
5114 * dumpfile.h (dump_user_location_t::dump_user_location_t):
5116 (dump_location_t::dump_location_t): Likewise.
5118 2018-08-01 Martin Sebor <msebor@redhat.com>
5120 PR tree-optimization/86650
5121 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5122 rather than a "gcall *". Directly pass the data of interest
5123 to percent_K_format, rather than building a temporary CALL_EXPR
5125 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5126 (gimple_fold_builtin_strncat): Adjust.
5127 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5128 gcall* argument with gimple*.
5129 * gimple-ssa-warn-restrict.c (check_call): Same.
5130 (wrestrict_dom_walker::before_dom_children): Same.
5131 (builtin_access::builtin_access): Same.
5132 (check_bounds_or_overlap): Same
5133 (maybe_diag_overlap): Same.
5134 (maybe_diag_offset_bounds): Same.
5135 * tree-diagnostic.c (default_tree_printer): Move usage of
5136 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5138 * tree-pretty-print.c (percent_K_format): Add argument.
5139 * tree-pretty-print.h: Add argument.
5140 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5141 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5142 (maybe_diag_stxncpy_trunc): Same.
5143 (handle_builtin_stxncpy): Same.
5144 (handle_builtin_strcat): Same.
5146 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5148 * match.pd: Optimise pointer range checks.
5150 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5152 PR tree-optimization/86758
5153 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5154 to remove pattern statements.
5156 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5158 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5159 result of dfs_enumerate_from when constructing stmt_vec_infos,
5160 instead of additionally calling get_loop_body.
5162 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5164 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5166 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5167 When creating an iv, assert that the step is not known to be zero.
5168 (vect_setup_realignment): Update call accordingly.
5169 * tree-vect-stmts.c (vectorizable_store): Likewise.
5170 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
5172 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5174 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5175 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5176 (vectorizable_reduction): Likewise.
5177 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5178 (vect_detect_hybrid_slp_stmts): Likewise.
5179 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5181 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
5183 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5184 (wide_int_set_zero_nonzero_bits): ...this.
5185 (zero_nonzero_bits_from_vr): Rename to...
5186 (vrp_set_zero_nonzero_bits): ...this.
5187 (extract_range_from_multiplicative_op_1): Abstract wide int
5189 (wide_int_range_multiplicative_op): ...here.
5190 (extract_range_from_binary_expr_1): Extract wide int binary
5191 operations into their own functions.
5192 (wide_int_range_lshift): New.
5193 (wide_int_range_can_optimize_bit_op): New.
5194 (wide_int_range_shift_undefined_p): New.
5195 (wide_int_range_bit_xor): New.
5196 (wide_int_range_bit_ior): New.
5197 (wide_int_range_bit_and): New.
5198 (wide_int_range_trunc_mod): New.
5199 (extract_range_into_wide_ints): New.
5200 (vrp_shift_undefined_p): New.
5201 (extract_range_from_multiplicative_op): New.
5202 (vrp_can_optimize_bit_op): New.
5203 * tree-vrp.h (value_range::dump): New.
5204 (wide_int_range_multiplicative_op): New.
5205 (wide_int_range_lshift):New.
5206 (wide_int_range_shift_undefined_p): New.
5207 (wide_int_range_bit_xor): New.
5208 (wide_int_range_bit_ior): New.
5209 (wide_int_range_bit_and): New.
5210 (wide_int_range_trunc_mod): New.
5211 (zero_nonzero_bits_from_bounds): Rename to...
5212 (wide_int_set_zero_nonzero_bits): ...this.
5213 (zero_nonzero_bits_from_vr): Rename to...
5214 (vrp_set_zero_nonzero_bits): ...this.
5215 (range_easy_mask_min_max): Rename to...
5216 (wide_int_range_can_optimize_bit_op): this.
5217 * vr-values.c (simplify_bit_ops_using_ranges): Rename
5218 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5220 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5222 * tree-vectorizer.h (vect_orig_stmt): New function.
5223 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5224 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5225 (vect_create_epilog_for_reduction): Likewise.
5226 (vectorizable_live_operation): Likewise.
5227 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5228 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5229 * tree-vect-stmts.c (vectorizable_call): Likewise.
5230 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5232 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5234 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5236 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5237 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5238 (vect_transform_loop): Likewise.
5239 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5241 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5243 * tree-vectorizer.h (vect_schedule_slp): Return void.
5244 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5245 (vect_schedule_slp): Likewise.
5247 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5249 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5251 (vect_transform_loop): Update calls accordingly. Schedule SLP
5252 instances before the main loop, if any exist.
5254 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5256 PR tree-optimization/86749
5257 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5258 If the lhs is used in a COND_EXPR, check that it is being used
5259 as the "then" or "else" value.
5261 2018-08-01 Tom de Vries <tdevries@suse.de>
5264 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5265 speculation_safe_value_not_needed.
5267 2018-08-01 Richard Biener <rguenther@suse.de>
5269 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5272 2018-08-01 Martin Liska <mliska@suse.cz>
5274 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5275 format how successful transformation is dumped.
5276 (gimple_mod_pow2_value_transform): Likewise.
5277 (gimple_mod_subtract_transform): Likewise.
5278 (gimple_stringops_transform): Likewise.
5280 2018-08-01 Martin Liska <mliska@suse.cz>
5283 * value-prof.c (interesting_stringop_to_profile_p):
5284 Simplify the code and add BUILT_IN_MEMMOVE.
5285 (gimple_stringops_transform): Likewise.
5287 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
5289 * config/aarch64/aarch64-simd.md
5290 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5291 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5292 use GPI iterator instead of SI mode.
5294 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5296 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5297 rs6000_speculation_barrier.
5298 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5299 new barrier pattern name.
5301 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5303 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5304 (speculation_barrier): New insn.
5306 2018-08-01 Richard Biener <rguenther@suse.de>
5309 * graphite.h: Include isl/id.h and isl/space.h to allow build
5312 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
5315 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5316 mode for COFF targets.
5317 * defaults.h (TARGET_COFF): Define.
5318 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5319 TARGET_COFF): Define.
5320 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5321 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5322 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5324 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5326 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5327 * gimple-streamer-out.c (output_bb): Save it.
5328 * lto-streamer-in.c (input_struct_function_base): Restore
5329 instance discriminator if available. Create map on demand.
5330 * lto-streamer-out.c (output_struct_function_base): Save it if
5332 * final.c (decl_to_instance_map): Document LTO strategy.
5334 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5335 Olivier Hainque <hainque@adacore.com>
5337 * debug.h (decl_to_instance_map_t): New type.
5338 (decl_to_instance_map): Declare.
5339 (maybe_create_decl_to_instance_map): New inline function.
5340 * final.c (bb_discriminator, last_bb_discriminator): New statics,
5341 to track basic block discriminators.
5342 (final_start_function_1): Initialize them.
5343 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
5345 (decl_to_instance_map): New variable.
5346 (map_decl_to_instance, maybe_set_discriminator): New functions.
5347 (notice_source_line): Set discriminator.
5349 2018-07-31 Ian Lance Taylor <iant@golang.org>
5351 * targhooks.c (default_have_speculation_safe_value): Add
5354 2018-07-31 David Malcolm <dmalcolm@redhat.com>
5356 * dump-context.h: Include "pretty-print.h".
5357 (dump_context::refresh_dumps_are_enabled): New decl.
5358 (dump_context::emit_item): New decl.
5359 (class dump_context): Add fields "m_test_pp" and
5361 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5362 (temp_dump_context::get_dumped_text): New decl.
5363 (class temp_dump_context): Add field "m_pp".
5364 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
5365 (dump_context::refresh_dumps_are_enabled): ...and add a test for
5367 (set_dump_file): Update for above change.
5368 (set_alt_dump_file): Likewise.
5369 (dump_loc): New overload, taking a pretty_printer *.
5370 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
5371 to any test pretty-printer.
5372 (make_item_for_dump_gimple_stmt): New function, adapted from
5373 optinfo::add_gimple_stmt.
5374 (dump_context::dump_gimple_stmt): Call it, and use the result,
5375 eliminating the direct usage of dump_file and alt_dump_file in
5376 favor of indirectly using them via emit_item.
5377 (make_item_for_dump_gimple_expr): New function, adapted from
5378 optinfo::add_gimple_expr.
5379 (dump_context::dump_gimple_expr): Call it, and use the result,
5380 eliminating the direct usage of dump_file and alt_dump_file in
5381 favor of indirectly using them via emit_item.
5382 (make_item_for_dump_generic_expr): New function, adapted from
5384 (dump_context::dump_generic_expr): Call it, and use the result,
5385 eliminating the direct usage of dump_file and alt_dump_file in
5386 favor of indirectly using them via emit_item.
5387 (make_item_for_dump_printf_va): New function, adapted from
5388 optinfo::add_printf_va.
5389 (make_item_for_dump_printf): New function.
5390 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
5391 and use the result, eliminating the direct usage of dump_file and
5392 alt_dump_file in favor of indirectly using them via emit_item.
5393 (make_item_for_dump_dec): New function.
5394 (dump_context::dump_dec): Call it, and use the result,
5395 eliminating the direct usage of dump_file and alt_dump_file in
5396 favor of indirectly using them via emit_item.
5397 (make_item_for_dump_symtab_node): New function, adapted from
5398 optinfo::add_symtab_node.
5399 (dump_context::dump_symtab_node): Call it, and use the result,
5400 eliminating the direct usage of dump_file and alt_dump_file in
5401 favor of indirectly using them via emit_item.
5402 (dump_context::begin_scope): Reimplement, avoiding direct usage
5403 of dump_file and alt_dump_file in favor of indirectly using them
5405 (dump_context::emit_item): New member function.
5406 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5407 Set up test pretty-printer on the underlying context. Call
5408 refresh_dumps_are_enabled.
5409 (temp_dump_context::~temp_dump_context): Call
5410 refresh_dumps_are_enabled.
5411 (temp_dump_context::get_dumped_text): New member function.
5412 (selftest::verify_dumped_text): New function.
5413 (ASSERT_DUMPED_TEXT_EQ): New macro.
5414 (selftest::test_capture_of_dump_calls): Run all tests twice, with
5415 and then without optinfo enabled. Add uses of
5416 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
5417 * dumpfile.h: Update comment for the dump_* API.
5418 * optinfo-emit-json.cc
5419 (selftest::test_building_json_from_dump_calls): Update for new
5420 param for temp_dump_context ctor.
5421 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
5422 and "m_owned" field.
5423 (optinfo_item::~optinfo_item): Likewise.
5424 (optinfo::add_item): New member function.
5425 (optinfo::emit): Update comment.
5426 (optinfo::add_string): Delete.
5427 (optinfo::add_printf): Delete.
5428 (optinfo::add_printf_va): Delete.
5429 (optinfo::add_gimple_stmt): Delete.
5430 (optinfo::add_gimple_expr): Delete.
5431 (optinfo::add_tree): Delete.
5432 (optinfo::add_symtab_node): Delete.
5433 (optinfo::add_dec): Delete.
5434 * optinfo.h (class dump_context): New forward decl.
5435 (optinfo::add_item): New decl.
5436 (optinfo::add_string): Delete.
5437 (optinfo::add_printf): Delete.
5438 (optinfo::add_printf_va): Delete.
5439 (optinfo::add_gimple_stmt): Delete.
5440 (optinfo::add_gimple_expr): Delete.
5441 (optinfo::add_tree): Delete.
5442 (optinfo::add_symtab_node): Delete.
5443 (optinfo::add_dec): Delete.
5444 (optinfo::add_poly_int): Delete.
5445 (optinfo_item::optinfo_item): Remove "owned" param.
5446 (class optinfo_item): Remove field "m_owned".
5448 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5451 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
5452 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
5453 requested variable alignment.
5454 (expand_one_ssa_partition): Likewise.
5455 (expand_one_var): Likewise.
5457 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5459 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5460 to speculation_safe_value_not_needed.
5462 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5464 * targhooks.h (speculation_safe_value_not_needed): New prototype.
5465 * targhooks.c (speculation_safe_value_not_needed): New function.
5466 * target.def (have_speculation_safe_value): Update documentation.
5467 * doc/tm.texi: Regenerated.
5469 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5471 * config/aarch64/iterators.md (ALLI_TI): New iterator.
5472 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
5474 (despeculate_copy<ALLI:mode>_insn): New insn.
5475 (despeculate_copyti_insn): New insn.
5476 (despeculate_simple<ALLI:mode>): New insn
5477 (despeculate_simpleti): New insn.
5478 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
5480 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
5481 aarch64_speculation_safe_value.
5482 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
5484 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5486 * config/aarch64/aarch64-speculation.cc: New file.
5487 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
5488 before pass_reorder_blocks.
5489 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
5491 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
5492 X14 and X15 when tracking speculation.
5493 * config/aarch64/aarch64.md (register name constants): Add
5494 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
5495 (unspec): Add UNSPEC_SPECULATION_TRACKER.
5496 (speculation_barrier): New insn attribute.
5497 (cmp<mode>): Allow SP in comparisons.
5498 (speculation_tracker): New insn.
5499 (speculation_barrier): Add speculation_barrier attribute.
5500 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
5501 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
5502 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
5504 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5506 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
5507 aarch64_track_speculation is true.
5508 (tb<optab><mode>1): Likewise.
5509 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
5510 generate CB[N]Z when tracking speculation.
5511 (aarch64_split_compare_and_swap): Likewise.
5512 (aarch64_split_atomic_op): Likewise.
5514 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5516 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
5518 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5520 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
5521 (speculation_barrier): New insn.
5523 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5525 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
5526 * config/arm/arm.md (speculation_barrier): New expand.
5527 (speculation_barrier_insn): New pattern.
5529 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5531 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
5532 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
5533 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
5534 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
5536 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
5537 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
5538 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
5539 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
5540 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
5541 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
5542 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
5543 * builtins.c (expand_speculation_safe_value): New function.
5544 (expand_builtin): Call it.
5545 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
5546 * doc/extend.texi: Document __builtin_speculation_safe_value.
5547 * doc/md.texi: Document "speculation_barrier" pattern.
5548 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
5549 TARGET_HAVE_SPECULATION_SAFE_VALUE.
5550 * doc/tm.texi: Regenerated.
5551 * target.def (have_speculation_safe_value, speculation_safe_value): New
5553 * targhooks.c (default_have_speculation_safe_value): New function.
5554 (default_speculation_safe_value): New function.
5555 * targhooks.h (default_have_speculation_safe_value): Add prototype.
5556 (default_speculation_safe_value): Add prototype.
5558 2018-07-31 David Malcolm <dmalcolm@redhat.com>
5560 * dump-context.h (dump_context::dump_loc): New decl.
5561 * dumpfile.c (dump_context::dump_loc): New member function.
5562 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
5563 and dump_gimple_stmt.
5564 (dump_context::dump_gimple_expr_loc): Likewise, using
5566 (dump_context::dump_generic_expr_loc): Likewise, using
5568 (dump_context::dump_printf_loc_va): Likewise, using
5570 (dump_context::begin_scope): Explicitly using the global function
5571 "dump_loc", rather than the member function.
5573 2018-07-31 Martin Sebor <msebor@redhat.com>
5575 PR tree-optimization/86741
5576 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
5578 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
5580 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
5581 depend on whether prefetch instructions will be emitted or not.
5582 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
5583 will be emitted or not.
5584 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
5585 (TARGET_SETMEM_PFD): New macros.
5587 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5589 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
5590 (NULL_STMT_VEC_INFO): Delete.
5591 (stmt_vec_info::operator*): Likewise.
5592 (stmt_vec_info::operator gimple *): Likewise.
5593 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
5594 of NULL_STMT_VEC_INFO.
5595 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5596 (vect_reassociating_reduction_p): Likewise.
5597 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5598 (vectorizable_store): Likewise.
5599 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
5600 (vec_info::free_stmt_vec_infos): Likewise.
5602 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5604 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
5605 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
5607 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5609 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
5610 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5611 (vec_info::free_stmt_vec_info): New private member functions.
5612 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
5613 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
5614 * tree-parloops.c (gather_scalar_reductions): Remove calls to
5615 set_stmt_vec_info_vec and free_stmt_vec_infos.
5616 * tree-vect-loop.c (_loop_vec_info): Remove call to
5617 set_stmt_vec_info_vec.
5618 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
5619 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
5620 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
5621 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5622 (vec_info::free_stmt_vec_info): ...these new functions. Remove
5623 assignments in {vec_info::,}new_stmt_vec_info that are redundant
5624 with the clearing in the xcalloc.
5625 (stmt_vec_info_vec): Delete.
5626 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
5627 (vectorize_loops): Likewise.
5628 (vec_info::~vec_info): Remove argument from call to
5629 free_stmt_vec_infos.
5630 (vec_info::add_stmt): Remove vinfo argument from call to
5633 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5635 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
5636 rather than a gimple stmt.
5637 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
5638 information for pattern statements when passed the original
5639 statement; instead wait to be passed the pattern statement itself.
5640 Don't call set_vinfo_for_stmt here.
5641 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
5642 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
5643 stmt_vec_infos here.
5644 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
5645 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
5646 stmt_vec_infos entry.
5648 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5650 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
5651 * tree-vectorizer.c (vec_info::replace_stmt): New function.
5652 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
5653 * tree-vect-stmts.c (vectorizable_call): Likewise.
5654 (vectorizable_simd_clone_call): Likewise.
5656 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5658 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
5659 * tree-vectorizer.c (vec_info::remove_stmt): New function.
5660 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
5661 * tree-vect-loop.c (vect_transform_loop): Likewise.
5662 * tree-vect-slp.c (vect_schedule_slp): Likewise.
5663 * tree-vect-stmts.c (vect_remove_stores): Likewise.
5665 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5667 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
5668 (vect_dr_stmt): Delete.
5669 * tree-vectorizer.c (vec_info::lookup_dr): New function.
5670 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
5672 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
5673 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
5674 (vect_verify_datarefs_alignment, vect_peeling_supportable)
5675 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5676 (vect_analyze_data_refs): Likewise.
5677 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
5679 (vect_find_same_alignment_drs): Likewise.
5680 (vect_slp_analyze_node_dependences): Update calls accordingly.
5681 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
5682 instead of DR_VECT_AUX.
5683 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
5684 of a vector data references. Use vec_info::lookup_dr instead of
5686 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
5687 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
5688 instead of DR_VECT_AUX.
5690 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5692 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
5694 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5696 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5697 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5698 (vect_gen_prolog_loop_niters): Likewise.
5700 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5702 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
5703 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
5704 (vect_known_alignment_in_bytes, vect_dr_behavior)
5705 (vect_get_scalar_dr_size): Take references as dr_vec_infos
5706 instead of data_references. Update calls to other routines for
5707 which the same change has been made.
5708 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
5709 dr_vec_infos instead of stmt_vec_infos.
5710 (vect_analyze_data_ref_dependence): Update call accordingly.
5711 (vect_slp_analyze_data_ref_dependence)
5712 (vect_record_base_alignments): Use DR_VECT_AUX.
5713 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
5714 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
5715 (vector_alignment_reachable_p, vect_get_data_access_cost)
5716 (vect_peeling_supportable, vect_analyze_group_access_1)
5717 (vect_analyze_group_access, vect_analyze_data_ref_access)
5718 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
5719 (vect_compile_time_alias, vect_small_gap_p)
5720 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
5721 (vect_supportable_dr_alignment): Take references as dr_vec_infos
5722 instead of data_references. Update calls to other routines for
5723 which the same change has been made.
5724 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
5725 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5726 (vect_slp_analyze_and_verify_node_alignment)
5727 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5728 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5729 (vect_setup_realignment): Use dr_vec_infos. Update calls after
5731 (_vect_peel_info::dr): Replace with...
5732 (_vect_peel_info::dr_info): ...this new field.
5733 (vect_peeling_hash_get_most_frequent)
5734 (vect_peeling_hash_choose_best_peeling): Update accordingly.
5735 (vect_peeling_hash_get_lowest_cost):
5736 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
5737 routines for which the same change has been made.
5738 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
5740 * tree-vect-loop-manip.c (get_misalign_in_elems)
5741 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
5743 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5744 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
5745 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
5746 (get_group_load_store_type, get_negative_load_store_type)
5747 (vect_get_data_ptr_increment, vectorizable_store)
5748 (vectorizable_load): Likewise.
5749 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
5750 Update calls to other routines for which the same change has been made.
5752 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5754 * tree-vectorizer.h (vec_info::move_dr): New member function.
5755 (dataref_aux): Rename to...
5756 (dr_vec_info): ...this and add "dr" and "stmt" fields.
5757 (_stmt_vec_info::dr_aux): Update accordingly.
5758 (_stmt_vec_info::data_ref_info): Delete.
5759 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
5760 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
5761 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
5762 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
5764 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
5765 (STMT_VINFO_DR_INFO): New macro.
5766 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
5767 (set_dr_misalignment): Update after rename of dataref_aux.
5768 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
5769 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
5770 initialization of STMT_VINFO_DATA_REF.
5771 * tree-vectorizer.c (vec_info::move_dr): New function.
5772 * tree-vect-patterns.c (vect_recog_bool_pattern)
5773 (vect_recog_mask_conversion_pattern)
5774 (vect_recog_gather_scatter_pattern): Use it.
5775 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
5776 the "dr" and "stmt" fields of dr_vec_info instead of
5777 STMT_VINFO_DATA_REF.
5779 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5781 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
5782 (is_pattern_stmt_p): Use it.
5783 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
5784 on pattern statements.
5786 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5788 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
5789 original stmt as a stmt_vec_info rather than a gimple stmt.
5790 (vect_pattern_recog_1): Take the statement directly as a
5791 stmt_vec_info, rather than via a gimple_stmt_iterator.
5792 Update call to vect_mark_pattern_stmts.
5793 (vect_pattern_recog): Update calls accordingly.
5795 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5797 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
5798 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
5799 a vect_def_type for the first argument.
5800 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
5801 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
5802 operand if it isn't defined by a vectorized statement.
5803 (vect_build_gather_load_calls): Remove the mask_dt argument and
5804 update calls to vect_get_vec_def_for_stmt_copy.
5805 (vectorizable_bswap): Likewise the dt argument.
5806 (vectorizable_call): Update calls to vectorizable_bswap and
5807 vect_get_vec_def_for_stmt_copy.
5808 (vectorizable_simd_clone_call, vectorizable_assignment)
5809 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
5810 (vectorizable_comparison): Update calls to
5811 vect_get_vec_def_for_stmt_copy.
5812 (vectorizable_store): Likewise. Remove now-unnecessary calls to
5814 (vect_get_loop_based_defs): Remove dt argument and update call
5815 to vect_get_vec_def_for_stmt_copy.
5816 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
5817 and vect_get_vec_def_for_stmt_copy.
5818 (vectorizable_load): Update calls to vect_build_gather_load_calls
5819 and vect_get_vec_def_for_stmt_copy.
5820 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5821 (vectorizable_reduction, vectorizable_live_operation): Update calls
5822 to vect_get_vec_def_for_stmt_copy.
5824 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5826 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
5827 and gimple stmt arguments with a stmt_vec_info.
5828 (vect_record_base_alignments): Update calls accordingly.
5829 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
5830 and gimple stmt arguments with a stmt_vec_info.
5831 (vect_build_slp_tree_1): Remove vinfo argument and update call
5832 to vect_record_max_nunits.
5833 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
5834 and vect_record_max_nunits.
5836 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5838 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
5839 file and take a stmt_vec_info instead of a gimple stmt.
5840 (supportable_widening_operation, vect_finish_replace_stmt)
5841 (vect_finish_stmt_generation, vect_get_store_rhs)
5842 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
5843 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
5844 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
5845 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
5846 (vect_create_data_ref_ptr, bump_vector_ptr)
5847 (vect_permute_store_chain, vect_setup_realignment)
5848 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
5849 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
5850 (vectorizable_reduction, vectorizable_induction)
5851 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
5852 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
5853 than gimple stmts as arguments.
5854 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
5855 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
5856 (can_group_stmts_p, vect_check_gather_scatter)
5857 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5858 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
5859 (vect_permute_load_chain, vect_shift_permute_load_chain)
5860 (vect_transform_grouped_load)
5861 (vect_record_grouped_load_vectors): Likewise.
5862 * tree-vect-loop.c (vect_fixup_reduc_chain)
5863 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
5864 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
5865 (vectorizable_reduction, vectorizable_induction)
5866 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
5867 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
5868 (vect_get_load_store_mask): Likewise.
5869 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
5870 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
5871 * tree-vect-stmts.c (vect_mark_relevant)
5872 (is_simple_and_all_uses_invariant)
5873 (exist_non_indexing_operands_for_use_p, process_use)
5874 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
5875 (vect_get_vec_def_for_operand, vect_get_vec_defs)
5876 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
5877 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
5878 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
5879 (get_negative_load_store_type, get_load_store_type)
5880 (vect_check_load_store_mask, vect_check_store_rhs)
5881 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
5882 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5883 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5884 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5885 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
5886 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
5887 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
5888 (supportable_widening_operation): Likewise.
5890 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5892 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
5893 a stmt_vec_info instead of a gcall.
5894 (vect_check_gather_scatter): Update call accordingly.
5895 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
5897 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
5898 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
5899 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
5900 instead of a gimple stmt.
5901 (vect_transform_loop): Update calls accordingly.
5902 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
5903 stmt_vec_infos instead of gimple stmts.
5904 (vect_analyze_slp_instance): Update use accordingly.
5905 * tree-vect-stmts.c (read_vector_array, write_vector_array)
5906 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
5907 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
5908 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
5909 (vect_gen_widened_results_half, vect_get_loop_based_defs)
5910 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
5911 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
5912 down to subroutines.
5914 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5916 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
5917 of the worklist from a vector of gimple stmts to a vector of
5919 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5920 (vect_mark_stmts_to_be_vectorized): Likewise
5922 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5924 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
5925 statement before passing it to vect_analyze_stmt.
5926 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
5927 the chain of phi vector definitions. Track the exit phi via its
5929 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
5930 STMT_VINFO_REDUC_DEF.
5931 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
5932 stmt_vec_infos to handle the statement chains.
5933 (vect_get_slp_defs): Record the first statement in the node
5934 using a stmt_vec_info.
5935 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
5936 statements here and pass their stmt_vec_info down to subroutines.
5937 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
5938 down to vect_finish_stmt_generation.
5939 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
5940 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
5941 stmt_vec_infos to subroutines.
5942 (vect_remove_stores): Use stmt_vec_infos to handle the statement
5945 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5947 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
5948 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
5949 (vect_permute_store_chain, vect_setup_realignment)
5950 (vect_permute_load_chain, vect_shift_permute_load_chain)
5951 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
5952 stmts internally, and when passing values to other vectorizer routines.
5953 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5954 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
5955 (vect_analyze_loop_operations, get_initial_def_for_reduction)
5956 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
5957 (vectorizable_reduction, vectorizable_induction)
5958 (vectorizable_live_operation, vect_transform_loop_stmt)
5959 (vect_transform_loop): Likewise.
5960 * tree-vect-patterns.c (vect_reassociating_reduction_p)
5961 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
5962 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
5963 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5964 (vect_slp_analyze_node_operations_1): Likewise.
5965 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5966 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
5967 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
5968 (vect_finish_stmt_generation_1, get_group_load_store_type)
5969 (get_load_store_type, vect_build_gather_load_calls)
5970 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5971 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5972 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5973 (vectorizable_store, vectorizable_load, vectorizable_condition)
5974 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
5975 (supportable_widening_operation): Likewise.
5976 (vect_get_vector_types_for_stmt): Likewise.
5977 * tree-vectorizer.h (vect_dr_behavior): Likewise.
5979 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5981 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5982 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
5983 (vect_permute_store_chain, vect_permute_load_chain)
5984 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
5985 repeated stmt_vec_info lookups.
5986 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5987 (vect_update_ivs_after_vectorizer): Likewise.
5988 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
5989 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5990 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
5991 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5992 (vect_bb_slp_scalar_cost): Likewise.
5993 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
5995 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5997 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
5998 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
5999 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6000 to get gassigns and gcalls, rather than operating on generc gimple
6002 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6003 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6004 (vectorizable_load, vect_analyze_stmt): Likewise.
6005 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6007 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6009 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6010 return stmt_vec_infos rather than gimple stmts. Do not accept
6012 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6014 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6015 Update use of get_later_stmt.
6016 (vect_get_constant_vectors): Update call accordingly.
6017 (vect_schedule_slp_instance): Likewise
6018 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6019 (vect_slp_analyze_instance_dependence): Likewise.
6020 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6022 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6024 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6025 (stmt_info_for_cost::stmt_info): ...this new field.
6026 (add_stmt_costs): Update accordingly.
6027 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6028 (vect_get_known_peeling_cost): Likewise.
6029 (vect_estimate_min_profitable_iters): Likewise.
6030 * tree-vect-stmts.c (record_stmt_cost): Likewise.
6032 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6034 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6035 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6036 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6038 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6040 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6042 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6043 a gimple stmt to a stmt_vec_info.
6044 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6046 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6048 * tree-vectorizer.h (vec_info::grouped_stores): Change from
6049 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6050 (_loop_vec_info::reduction_chains): Likewise.
6051 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6053 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6055 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6057 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6058 a gimple stmt to a stmt_vec_info.
6059 (_stmt_vec_info::next_element): Likewise.
6060 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6061 (vect_slp_analyze_and_verify_node_alignment)
6062 (vect_analyze_group_access_1, vect_analyze_group_access)
6063 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6064 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6065 (vect_supportable_dr_alignment): Update accordingly.
6066 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6067 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6068 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6069 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6070 * tree-vect-slp.c (vect_build_slp_tree_1)
6071 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6072 (vect_split_slp_store_group, vect_analyze_slp_instance)
6073 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6074 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6075 (get_group_load_store_type, get_load_store_type)
6076 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6077 (vect_transform_stmt, vect_remove_stores): Likewise.
6079 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6081 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6083 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6084 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6085 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6086 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6087 (vector_alignment_reachable_p, vect_get_data_access_cost)
6088 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6089 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6090 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6091 (vect_analyze_group_access_1, vect_analyze_group_access)
6092 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6093 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6094 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6095 result of vect_dr_stmt and use the stmt_vec_info instead of
6096 the associated gimple stmt.
6097 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6098 (vect_gen_prolog_loop_niters): Likewise.
6099 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6101 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6103 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6104 to a vec<stmt_vec_info>.
6105 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6106 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6108 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6109 to a vec<stmt_vec_info>.
6110 (bst_traits::value_type, bst_traits::value_type): Likewise.
6111 (bst_traits::hash): Update accordingly.
6112 (vect_get_and_check_slp_defs): Change the stmts parameter from
6113 a vec<gimple *> to a vec<stmt_vec_info>.
6114 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6115 (vect_build_slp_tree): Likewise.
6116 (vect_build_slp_tree_2): Likewise. Update uses of
6117 SLP_TREE_SCALAR_STMTS.
6118 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6119 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6120 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6121 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6122 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6123 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6124 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6125 (vect_get_constant_vectors, vect_get_slp_defs)
6126 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6127 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6128 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6129 instead of gimple stmts.
6130 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6131 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6132 (vect_slp_analyze_instance_dependence): Update uses of
6133 SLP_TREE_SCALAR_STMTS.
6134 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6135 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6136 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6137 (get_initial_defs_for_reduction): Likewise.
6138 (vect_create_epilog_for_reduction): Likewise.
6139 (vectorize_fold_left_reduction): Likewise.
6140 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6141 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6142 (can_vectorize_live_stmts): Likewise.
6144 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6146 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6147 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6148 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6150 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6152 (gather_scalar_reductions): Update after above interface changes.
6153 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6154 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6156 (vect_force_simple_reduction): Likewise.
6157 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6158 LOOP_VINFO_REDUCTIONS.
6159 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6161 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6163 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6164 a gimple stmt to a stmt_vec_info.
6165 * tree-vect-loop.c (vect_active_double_reduction_p)
6166 (vect_force_simple_reduction, vectorizable_reduction): Update
6169 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6171 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6172 vec<gimple *> to a vec<stmt_vec_info>.
6173 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6174 the reduction_phis argument from a vec<gimple *> to a
6176 (vectorizable_reduction): Likewise the phis local variable that
6177 is passed to vect_create_epilog_for_reduction. Update for new type
6178 of SLP_TREE_VEC_STMTS.
6179 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6180 (vectorizable_live_operation): Likewise.
6181 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6182 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6184 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6186 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6187 a gimple stmt to a stmt_vec_info.
6188 (vectorizable_condition, vectorizable_live_operation)
6189 (vectorizable_reduction, vectorizable_induction): Pass back the
6190 vectorized statement as a stmt_vec_info.
6191 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6192 use of STMT_VINFO_VEC_STMT.
6193 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6194 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6195 as stmt_vec_infos rather than gimple stmts.
6196 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6198 (vectorizable_live_operation): Likewise.
6199 (vectorizable_reduction, vectorizable_induction): Likewise,
6200 updating use of STMT_VINFO_VEC_STMT.
6201 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6202 of STMT_VINFO_VEC_STMT.
6203 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6204 (vectorizable_simd_clone_call, vectorizable_conversion)
6205 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6206 (vectorizable_store, vectorizable_load, vectorizable_condition)
6207 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6208 from a gimple stmt to a stmt_vec_info.
6209 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
6210 pointer to a stmt_vec_info to the vectorizable_* routines.
6212 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6214 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6215 a gimple stmt to a stmt_vec_info.
6216 (is_pattern_stmt_p): Update accordingly.
6217 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6218 (vect_record_grouped_load_vectors): Likewise.
6219 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6220 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6221 (vect_model_reduction_cost): Likewise.
6222 (vect_create_epilog_for_reduction): Likewise.
6223 (vectorizable_reduction, vectorizable_induction): Likewise.
6224 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6225 Return the stmt_vec_info for the pattern statement.
6226 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6227 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6228 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6229 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6230 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6231 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6232 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6233 (free_stmt_vec_info, vect_is_simple_use): Likewise.
6235 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6237 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6238 (vect_finish_stmt_generation): Likewise.
6239 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6240 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6241 (vect_build_gather_load_calls): Use the return value of the above
6242 functions instead of a separate call to vinfo_for_stmt. Use narrow
6243 scopes for the input gimple stmt and wider scopes for the associated
6244 stmt_vec_info. Use vec_info::lookup_def when setting these
6245 stmt_vec_infos from an SSA_NAME definition.
6246 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6247 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6248 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6249 (vectorizable_store, vectorizable_load, vectorizable_condition)
6250 (vectorizable_comparison): Likewise.
6251 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6252 (vectorizable_reduction): Likewise.
6254 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6256 * tree-vectorizer.h (vect_is_simple_use): Add an optional
6257 stmt_vec_info * parameter before the optional gimple **.
6258 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6259 (process_use, vect_get_vec_def_for_operand_1): Update callers.
6260 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6261 * tree-vect-loop.c (vectorizable_reduction): Likewise.
6262 (vectorizable_live_operation): Likewise.
6263 * tree-vect-patterns.c (type_conversion_p): Likewise.
6264 (vect_look_through_possible_promotion): Likewise.
6265 (vect_recog_rotate_pattern): Likewise.
6266 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6268 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6270 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6271 a typedef to a wrapper class.
6272 (NULL_STMT_VEC_INFO): New macro.
6273 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6274 (stmt_vec_info::operator*): New function.
6275 (stmt_vec_info::operator gimple *): Likewise.
6276 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6277 (add_stmt_costs): Likewise.
6278 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6279 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6280 (vect_get_known_peeling_cost): Likewise.
6281 (vect_estimate_min_profitable_iters): Likewise.
6282 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6283 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6284 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6285 (vectorizable_store, free_stmt_vec_infos): Likewise.
6286 (new_stmt_vec_info): Change return type of xcalloc to
6289 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6291 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6292 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6293 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6294 a single_imm_use-based sequence.
6295 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6297 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6299 * tree-vectorizer.h (vec_info::lookup_def): Declare.
6300 * tree-vectorizer.c (vec_info::lookup_def): New function.
6301 * tree-vect-patterns.c (vect_get_internal_def): Use it.
6302 (vect_widened_op_tree): Likewise.
6303 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6304 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6305 (vectorizable_reduction): Likewise.
6306 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6308 (vect_is_slp_reduction): Update calls accordingly. Use
6309 vec_info::lookup_def.
6310 (vect_is_simple_reduction): Likewise
6311 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6313 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6315 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6316 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6317 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6319 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6320 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6321 (vect_update_vf_for_slp, vect_analyze_loop_operations)
6322 (vect_is_slp_reduction, vectorizable_induction)
6323 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6324 * tree-vect-patterns.c (vect_init_pattern_stmt):
6325 (vect_determine_min_output_precision_1, vect_determine_precisions)
6326 (vect_pattern_recog): Likewise.
6327 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6328 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6329 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6330 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6331 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6332 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
6333 info field from a loop to a loop_vec_info.
6335 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6337 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
6338 (vec_info::add_stmt): Declare.
6339 * tree-vectorizer.c (vec_info::add_stmt): New function.
6340 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
6341 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
6342 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6343 (vectorizable_induction): Likewise.
6344 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
6345 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6346 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
6347 (vectorizable_load): Likewise.
6348 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6349 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
6350 (vect_recog_gather_scatter_pattern): Likewise.
6351 (append_pattern_def_seq): Likewise. Remove a check that is
6352 performed by add_stmt itself.
6354 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6356 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
6357 which make_ssa_name was called with new_stmt before new_stmt
6360 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6362 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
6364 (vect_is_slp_reduction): ...here...
6365 (vect_is_simple_reduction): ...and here. Remove repetition of tests
6366 that are already known to be false.
6368 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6370 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
6371 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
6372 STMT_VINFO_NUM_SLP_USES when it's true.
6373 (vect_free_slp_instance): Add a final_p parameter and pass it to
6375 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
6376 (vect_analyze_slp_instance): Likewise.
6377 (vect_slp_analyze_operations): Likewise.
6378 (vect_slp_analyze_bb_1): Likewise.
6379 * tree-vectorizer.c (vec_info): Likewise.
6380 * tree-vect-loop.c (vect_transform_loop): Likewise.
6382 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6384 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6385 function is not called for second and subsequent members of
6388 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6390 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
6391 cases for nested loops from here to ...
6392 (vect_create_epilog_for_reduction): ...here. Only call
6393 vect_is_simple_use for inner-loop reductions.
6395 2018-07-31 Martin Liska <mliska@suse.cz>
6397 PR gcov-profile/85338
6398 PR gcov-profile/85350
6399 PR gcov-profile/85372
6400 * profile.c (struct location_triplet): New.
6401 (struct location_triplet_hash): Likewise.
6402 (output_location): Do not output a BB that
6403 is already recorded for a line.
6404 (branch_prob): Use streamed_locations.
6406 2018-07-31 Martin Liska <mliska@suse.cz>
6408 PR gcov-profile/85370
6409 * coverage.c (coverage_begin_function): Do not mark target
6410 clones as artificial functions.
6412 2018-07-31 Martin Liska <mliska@suse.cz>
6414 PR gcov-profile/83813
6415 PR gcov-profile/84758
6416 PR gcov-profile/85217
6417 PR gcov-profile/85332
6418 * profile.c (branch_prob): Do not record GOTO expressions
6419 for GIMPLE statements which locations are already streamed.
6421 2018-07-31 Olivier Hainque <hainque@adacore.com>
6423 * gcc.c (handle_spec_function): Accept a soft_matched_part
6424 argument, as do_spec_1. Pass it down to ...
6425 (eval_spec_function): Accept a soft_matched_part argument,
6426 and pass it down to ...
6427 (do_spec_2): Accept a soft_matched_part argument, and pass
6428 it down to do_spec_1.
6429 (do_spec_1): Pass soft_matched_part to handle_spec_function.
6430 (handle_braces): Update call to handle_spec_function.
6431 (driver::set_up_specs): Update calls to do_spec_2.
6432 (compare_debug_dump_opt_spec_function): Likewise.
6433 (compare_debug_self_opt_spec_function): Likewise.
6435 2018-07-31 Olivier Hainque <hainque@adacore.com>
6437 * common.opt (nolibc): New option.
6438 * doc/invoke.texi (Link Options): Document it.
6439 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
6440 * config/alpha/linux.h: Likewise.
6441 * config/arc/elf.h: Likewise.
6442 * config/arm/uclinux-elf.h: Likewise.
6443 * config/arm/unknown-elf.h: Likewise.
6444 * config/avr/avrlibc.h: Likewise.
6445 * config/bfin/bfin.h: Likewise.
6446 * config/bfin/linux.h: Likewise.
6447 * config/bfin/uclinux.h: Likewise.
6448 * config/darwin.h: Likewise.
6449 * config/darwin10.h: Likewise.
6450 * config/darwin12.h: Likewise.
6451 * config/gnu-user.h: Likewise.
6452 * config/lm32/uclinux-elf.h: Likewise.
6453 * config/pa/pa-hpux11.h: Likewise.
6454 * config/pa/pa64-hpux.h: Likewise.
6455 * config/sparc/sparc.h: Likewise.
6457 2018-07-31 Olivier Hainque <hainque@adacore.com>
6459 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
6460 undefined variables.
6462 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
6465 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
6468 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6470 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6471 terminated string literal.
6473 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
6475 PR rtl-optimization/85160
6476 * combine.c (is_just_move): New function.
6477 (try_combine): Allow combining two instructions into two if neither of
6478 the original instructions was a move.
6480 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
6483 * doc/extend.texi (Global Register Variables): Discourage use of type
6485 (Local Register Variables): Likewise.
6487 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
6489 PR tree-optimization/86506
6490 * hwint.h (ceil_log2): Resync with hwint.c implementation.
6492 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
6495 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
6496 hard_regno, make sure no insn between `from` and `to` clobbers it.
6498 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
6499 Tom de Vries <tdevries@suse.de>
6501 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
6502 (PTX_DEFAULT_RUNTIME_DIM): ... this.
6503 (nvptx_goacc_validate_dims): Set default worker and gang dims to
6504 PTX_DEFAULT_RUNTIME_DIM.
6505 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
6507 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
6509 * config/pa/pa.c (pa_output_addr_vec): Align address table.
6510 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
6511 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
6513 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
6515 * config/rs6000/constraints.md (wG constraint): Delete, no longer
6517 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
6518 predicate to reflect toc fusion has been deleted.
6519 (toc_fusion_mem_raw): Delete, no longer used.
6520 (toc_fusion_mem_wrapped): Likewise.
6521 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
6523 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
6524 Delete, no longer used.
6525 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
6526 meant to be used for toc fusion.
6527 (rs6000_debug_print_mode): Delete toc fusion debugging.
6528 (rs6000_debug_reg_global): Likewise.
6529 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
6530 fusion and secondary reload support that were never used.
6531 (rs6000_option_override_internal): Delete TOC fusion, that was only
6532 partially defined, and it did not work unless you also used the
6534 (rs6000_legitimate_address_p): Delete TOC fusion support.
6535 (rs6000_opt_masks): Likewise.
6536 (fusion_wrap_memory_address): Delete function, no longer used.
6537 (fusion_split_address); Delete TOC fusion support.
6538 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
6539 longer used with toc fusion being deleted.
6540 (TARGET_TOC_FUSION_FP): Likewise.
6541 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
6543 (toc fusion spliter): Delete TOC fusion support.
6544 (toc_fusionload_<mode>): Likewise.
6545 (toc_fusionload_di): Likewise.
6546 (fusion_gpr_load_<mode>): Delete generator function, this insn no
6547 longer needs to be named. Rename predicate to delete TOC fusion.
6548 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6549 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6550 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6551 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6552 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
6554 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
6556 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
6557 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
6558 __int128 in built-in function prototypes.
6559 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
6560 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
6562 2018-07-27 Martin Sebor <msebor@redhat.com>
6564 PR tree-optimization/86696
6565 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
6566 types, including enums.
6567 (handle_char_store): Be prepared for the above function to fail.
6569 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
6571 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
6572 when optimization level is lower than 2 or optimize for size.
6574 2018-07-26 Martin Sebor <msebor@redhat.com>
6576 PR tree-optimization/86043
6577 PR tree-optimization/86042
6578 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
6579 (get_string_cst_length): Rename...
6580 (get_min_string_length): ...to this. Add argument.
6581 (handle_char_store): Extend to handle multi-character stores by
6583 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
6584 * tree.h (initializer_zerop): Add argument.
6586 2018-07-26 Jakub Jelinek <jakub@redhat.com>
6589 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
6590 declare target to variables if they have always,{to,from,tofrom} map
6593 2018-07-26 Martin Liska <mliska@suse.cz>
6596 * lto-wrapper.c: Add linker_output as prefix
6597 for ltrans_output_file.
6599 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
6601 PR rtl-optimization/85805
6602 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
6603 value for hard registers if that was written in the same mode.
6605 2018-07-26 Martin Liska <mliska@suse.cz>
6607 PR gcov-profile/86536
6608 * gcov.c (format_gcov): Use printf format %.*f directly
6609 and do not handle special values.
6611 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6613 * common/config/arc/arc-common.c (arc_option_optimization_table):
6614 Update default optimizations for size.
6616 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6618 * config/arc/arc.md (movsf_insn): Add short instruction selection.
6619 * config/arc/constraints.md (CfZ): New constraint.
6620 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
6621 (subsf3_fpu): Likewise.
6622 (cmpsf_fpu): Likewise.
6623 (cmpsf_fpu_uneq): Likewise.
6625 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6627 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
6629 (arc_is_uncached_mem_p): uncached applies to both the variable and
6632 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6634 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
6637 2018-07-25 David Malcolm <dmalcolm@redhat.com>
6639 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
6640 field "m_scopes" from vec to auto_vec.
6642 2018-07-25 Martin Liska <mliska@suse.cz>
6644 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
6647 2018-07-25 Richard Biener <rguenther@suse.de>
6650 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
6651 special wrt context_die late.
6652 (gen_subprogram_die): Re-use DIEs in local scope.
6654 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
6656 PR tree-optimization/86644
6657 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
6659 2018-07-25 Martin Liska <mliska@suse.cz>
6662 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
6663 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
6665 2018-07-25 Martin Liska <mliska@suse.cz>
6668 * params.def: Explain ASan abbreviation and provide
6669 a documentation link.
6671 2018-07-24 Martin Sebor <msebor@redhat.com>
6673 PR tree-optimization/86622
6674 PR tree-optimization/86532
6675 * builtins.h (string_length): Declare.
6676 * builtins.c (c_strlen): Correct handling of non-constant offsets.
6677 (check_access): Be prepared for non-constant length ranges.
6678 (string_length): Make extern.
6679 * expr.c (string_constant): Only handle the minor non-constant
6680 array index. Use string_constant to compute the length of
6681 a generic string constant.
6683 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
6685 PR tree-optimization/86618
6686 * tree-vect-stmts.c (vectorizable_call): Don't take the address
6687 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
6689 2018-07-24 David Malcolm <dmalcolm@redhat.com>
6691 PR tree-optimization/86636
6692 * json.cc (json::object::set): Fix comment. Add assertions.
6693 (json::array::append): Move here from json.h. Add comment and an
6695 (json::string::string): Likewise.
6696 * json.h (json::array::append): Move to json.cc.
6697 (json::string::string): Likewise.
6698 * optinfo-emit-json.cc
6699 (optrecord_json_writer::impl_location_to_json): Assert that we
6700 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
6701 wrapper around it. Expand the location once, rather than three
6703 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
6704 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
6706 (optrecord_json_writer::optinfo_to_json): Likewise, in four
6707 places. Fix some overlong lines.
6709 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
6711 * config/aarch64/aarch64-simd.md
6712 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
6713 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
6714 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
6715 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
6716 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
6717 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
6718 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
6719 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
6720 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
6722 2018-07-24 Jakub Jelinek <jakub@redhat.com>
6725 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
6726 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
6727 and abs_d == d, do the power of two handling if profitable.
6729 2018-07-24 Richard Biener <rguenther@suse.de>
6731 * match.pd: Add BIT_FIELD_REF canonicalizations.
6733 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6736 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
6738 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6740 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
6741 terminated STRING_CST object.
6743 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6745 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
6747 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6749 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
6750 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
6751 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
6752 the elements into a register.
6753 (rs6000_split_v4si_init_di_reg): Delete.
6754 (rs6000_split_v4si_init): Delete.
6755 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
6756 (vsx_init_v4si): Rewrite as a define_expand.
6758 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6760 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
6761 zero_extend argument from memory): New.
6763 2018-07-22 Martin Sebor <msebor@redhat.com>
6766 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
6767 diagnosing calls with unknown arguments unless -Walloca-larger-than
6768 is restricted to less than PTRDIFF_MAX bytes.
6770 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
6772 * doc/gcov.texi (Invoking Gcov): Editorial changes.
6774 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6776 * pretty-print.c (text_info::set_location): Remove redundant
6777 "line_table" parameter from call to rich_location::set_range.
6779 2018-07-20 Martin Sebor <msebor@redhat.com>
6782 * builtins.c (expand_builtin_alloca): Adjust.
6783 * calls.c (alloc_max_size): Simplify.
6784 * cgraphunit.c (cgraph_node::expand): Adjust.
6785 * common.opt (larger_than_size, warn_frame_larger_than): Remove
6787 (frame_larger_than_size): Same.
6788 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
6789 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
6791 * doc/invoke.texi (GCC Command Options): Document option arguments.
6792 Explain byte-size arguments and suffixes.
6793 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
6794 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
6795 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
6796 * doc/options.texi (UInteger): Expand.
6797 (Host_Wide_Int, ByteSize): Document new properties.
6798 * final.c (final_start_function_1): Include sizes in an error message.
6799 * function.c (frame_offset_overflow): Same.
6800 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
6801 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
6802 Diagnose unbounded alloca calls only for limits of less than
6804 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
6805 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
6807 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
6808 only for limits of less than PTRDIFF_MAX.
6809 * langhooks-def.h (lhd_handle_option): Change function argument
6811 * langhooks.c (lhd_handle_option): Same.
6812 * langhooks.h (handle_option): Same.
6813 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
6815 (var_type, var_type_struct): Same.
6816 (var_set): Handle ByteSize flag.
6817 * optc-gen.awk: Add comments to output to ease debugging. Make
6818 use of HOST_WIDE_INT where appropriate.
6819 * opts-gen-save.awk: Use %lx to format unsigned long.
6820 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
6821 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
6822 arguments. Parse bytes-size suffixes.
6823 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
6824 (enum_value_to_arg): Same.
6825 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
6826 (handle_option): Adjust.
6827 (generate_option): Change function argument to HOST_WIDE_INT.
6828 (cmdline_handle_error): Adjust.
6829 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
6830 (set_option): Change function argument to HOST_WIDE_INT.
6831 (option_enabled): Handle cl_host_wide_int.
6832 (get_option_state): Handle CLVC_SIZE.
6833 (control_warning_option): Same.
6834 * opts.c (common_handle_option): Change function argument to
6835 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
6836 OPT_Wvla_larger_than_.
6837 * opts.h (enum cl_var_type): Add an enumerator.
6838 * stor-layout.c (layout_decl): Print a more meaningful warning.
6839 * toplev.c (output_stack_usage): Adjust.
6841 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
6843 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
6844 call to inline_expand_builtin_string_cmp.
6845 (expand_builtin_strcmp): Likewise.
6846 (expand_builtin_strncmp): Likewise.
6847 (inline_string_cmp): Delete the last parameter, change char_type_node
6848 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
6850 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
6851 the inlining expansion on target where the type of the call has same or
6852 narrower precision than unsigned char.
6854 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6856 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
6857 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
6858 * common.opt (fsave-optimization-record): New option.
6859 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
6860 * doc/invoke.texi (-fsave-optimization-record): New option.
6861 * dumpfile.c: Include "optinfo-emit-json.h".
6862 (struct kv_pair): Move to coretypes.h.
6863 (optgroup_options): Make non-static.
6864 (dump_context::end_scope): Call
6865 optimization_records_maybe_pop_dump_scope.
6866 * dumpfile.h (optgroup_options): New decl.
6867 * json.cc: New file.
6869 * optinfo-emit-json.cc: New file.
6870 * optinfo-emit-json.h: New file.
6871 * optinfo.cc: Include "optinfo-emit-json.h".
6872 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
6873 (optinfo_enabled_p): Check optimization_records_enabled_p.
6874 (optinfo_wants_inlining_info_p): Likewise.
6875 * optinfo.h: Update comment.
6876 * profile-count.c (profile_quality_as_string): New function.
6877 * profile-count.h (profile_quality_as_string): New decl.
6878 (profile_count::quality): New accessor.
6879 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
6880 and optinfo_emit_json_cc_tests.
6881 * selftest.h (selftest::json_cc_tests): New decl.
6882 (selftest::optinfo_emit_json_cc_tests): New decl.
6883 * toplev.c: Include "optinfo-emit-json.h".
6884 (compile_file): Call optimization_records_finish.
6885 (do_compile): Call optimization_records_start.
6886 * tree-ssa-live.c: Include optinfo.h.
6887 (remove_unused_scope_block_p): Retain inlining information if
6888 optinfo_wants_inlining_info_p returns true.
6890 2018-07-20 Richard Biener <rguenther@suse.de>
6893 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
6894 to cover -flto-partition=none.
6896 2018-07-20 Martin Liska <mliska@suse.cz>
6898 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
6899 (get_decl_source_range): Remove unused function.
6901 2018-07-20 Richard Biener <rguenther@suse.de>
6903 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
6904 (struct vn_phi_s): Likewise.
6905 (struct vn_reference_s): Likewise.
6906 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
6907 for searching the slot of an entry known to be in the hash itself.
6908 (vn_phi_hasher::equal): Likewise.
6909 (vn_reference_hasher::equal): Likewise.
6910 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
6912 (optimistic_info, current_info): Remove, keeping only valid_info.
6913 (vn_reference_lookup_1): Remove fallback lookup.
6914 (vn_reference_lookup_2): Likewise.
6915 (vn_nary_op_lookup_1): Likewise.
6916 (vn_phi_lookup): Likewise.
6917 (vn_nary_build_or_lookup_1): Make sure to not chain the built
6919 (vn_reference_insert): Adjust, chain the inserted hash element
6920 at last_inserted_ref.
6921 (vn_reference_insert_pieces): Likewise.
6922 (visit_reference_op_call): Likewise.
6923 (vn_nary_op_insert_into): Chain the inserted hash element at
6925 (vn_nary_op_insert_pieces): Adjust.
6926 (vn_nary_op_insert): Likewise.
6927 (vn_nary_op_insert_stmt): Likewise.
6928 (vn_phi_insert): Adjust, chain the inserted hash element at
6930 (process_scc): Remove clearing and copying the optimistic
6931 table. Instead remove elements inserted during an optimistic
6932 iteration from the single table we maintain.
6933 (init_scc_vn): Adjust.
6934 (free_scc_vn): Likewise.
6935 (sccvn_dom_walker::record_cond): Likewise.
6936 (sccvn_dom_walker::after_dom_children): Likewise.
6938 2018-07-19 Martin Sebor <msebor@redhat.com>
6940 PR tree-optimization/84047
6941 PR tree-optimization/83776
6942 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
6943 (check_array_bounds): Call it.
6945 2018-07-19 Martin Sebor <msebor@redhat.com>
6947 * align.h (align_flags): Use member initialization.
6949 2018-07-19 David Malcolm <dmalcolm@redhat.com>
6951 * Makefile.in (OBJS): Add optinfo.o.
6952 * coretypes.h (class symtab_node): New forward decl.
6953 (struct cgraph_node): New forward decl.
6954 (class varpool_node): New forward decl.
6955 * dump-context.h: New file.
6956 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
6958 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
6959 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
6960 (set_alt_dump_file): Likewise.
6961 (dump_context::~dump_context): New dtor.
6962 (dump_gimple_stmt): Move implementation to...
6963 (dump_context::dump_gimple_stmt): ...this new member function.
6964 Add the stmt to any pending optinfo, creating one if need be.
6965 (dump_gimple_stmt_loc): Move implementation to...
6966 (dump_context::dump_gimple_stmt_loc): ...this new member function.
6967 Start a new optinfo and add the stmt to it.
6968 (dump_gimple_expr): Move implementation to...
6969 (dump_context::dump_gimple_expr): ...this new member function.
6970 Add the stmt to any pending optinfo, creating one if need be.
6971 (dump_gimple_expr_loc): Move implementation to...
6972 (dump_context::dump_gimple_expr_loc): ...this new member function.
6973 Start a new optinfo and add the stmt to it.
6974 (dump_generic_expr): Move implementation to...
6975 (dump_context::dump_generic_expr): ...this new member function.
6976 Add the tree to any pending optinfo, creating one if need be.
6977 (dump_generic_expr_loc): Move implementation to...
6978 (dump_context::dump_generic_expr_loc): ...this new member
6979 function. Add the tree to any pending optinfo, creating one if
6981 (dump_printf): Move implementation to...
6982 (dump_context::dump_printf_va): ...this new member function. Add
6983 the text to any pending optinfo, creating one if need be.
6984 (dump_printf_loc): Move implementation to...
6985 (dump_context::dump_printf_loc_va): ...this new member function.
6986 Start a new optinfo and add the stmt to it.
6987 (dump_dec): Move implementation to...
6988 (dump_context::dump_dec): ...this new member function. Add the
6989 value to any pending optinfo, creating one if need be.
6990 (dump_context::dump_symtab_node): New member function.
6991 (dump_context::get_scope_depth): New member function.
6992 (dump_context::begin_scope): New member function.
6993 (dump_context::end_scope): New member function.
6994 (dump_context::ensure_pending_optinfo): New member function.
6995 (dump_context::begin_next_optinfo): New member function.
6996 (dump_context::end_any_optinfo): New member function.
6997 (dump_context::s_current): New global.
6998 (dump_context::s_default): New global.
6999 (dump_scope_depth): Delete global.
7000 (dumpfile_ensure_any_optinfo_are_flushed): New function.
7001 (dump_symtab_node): New function.
7002 (get_dump_scope_depth): Reimplement in terms of dump_context.
7003 (dump_begin_scope): Likewise.
7004 (dump_end_scope): Likewise.
7005 (selftest::temp_dump_context::temp_dump_context): New ctor.
7006 (selftest::temp_dump_context::~temp_dump_context): New dtor.
7007 (selftest::verify_item): New function.
7008 (ASSERT_IS_TEXT): New macro.
7009 (ASSERT_IS_TREE): New macro.
7010 (ASSERT_IS_GIMPLE): New macro.
7011 (selftest::test_capture_of_dump_calls): New test.
7012 (selftest::dumpfile_c_tests): Call it.
7013 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7014 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7015 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7016 descriptive comment.
7017 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7018 (dump_node, dump_bb): Move these unrelated decls.
7019 (class dump_manager): Add leading comment.
7020 * optinfo.cc: New file.
7021 * optinfo.h: New file.
7023 2018-07-19 Michael Collison <michael.collison@arm.com>
7024 Richard Henderson <rth@redhat.com>
7026 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7027 (subti): Handle op1 zero.
7028 (subvti4, usub4ti4): New.
7029 (*sub<GPI>3_compare1_imm): New.
7030 (sub<GPI>3_carryinCV): New.
7031 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7032 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7034 2018-07-19 Michael Collison <michael.collison@arm.com>
7035 Richard Henderson <rth@redhat.com>
7037 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7038 (addti3): Create simpler code if low part is already known to be 0.
7039 (addvti4, uaddvti4): New.
7040 (*add<GPI>3_compareC_cconly_imm): New.
7041 (*add<GPI>3_compareC_cconly): New.
7042 (*add<GPI>3_compareC_imm): New.
7043 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7044 handle constants within this pattern..
7045 (*add<GPI>3_compareV_cconly_imm): New.
7046 (*add<GPI>3_compareV_cconly): New.
7047 (*add<GPI>3_compareV_imm): New.
7048 (add<GPI>3_compareV): New.
7049 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7050 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7051 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7052 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7053 with 'comparison' operator.
7054 (*add<GPI>3_compareV_cconly_imm): Ditto.
7055 (*add<GPI>3_compareV_cconly): Ditto.
7056 (*add<GPI>3_compareV_imm): Ditto.
7057 (add<GPI>3_compareV): Ditto.
7058 (add<mode>3_carryinC): Ditto.
7059 (*add<mode>3_carryinC_zero): Ditto.
7060 (*add<mode>3_carryinC): Ditto.
7061 (add<mode>3_carryinV): Ditto.
7062 (*add<mode>3_carryinV_zero): Ditto.
7063 (*add<mode>3_carryinV): Ditto.
7065 2018-07-19 Michael Collison <michael.collison@arm.com>
7066 Richard Henderson <rth@redhat.com>
7068 * config/aarch64/aarch64-modes.def (CC_V): New.
7069 * config/aarch64/aarch64-protos.h
7070 (aarch64_addti_scratch_regs): Declare
7071 (aarch64_subvti_scratch_regs): Declare.
7072 (aarch64_expand_subvti): Declare.
7073 (aarch64_gen_unlikely_cbranch): Declare
7074 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7075 for signed overflow using CC_Vmode.
7076 (aarch64_get_condition_code_1): Handle CC_Vmode.
7077 (aarch64_gen_unlikely_cbranch): New function.
7078 (aarch64_addti_scratch_regs): New function.
7079 (aarch64_subvti_scratch_regs): New function.
7080 (aarch64_expand_subvti): New function.
7082 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7084 * config/aarch64/aarch64-option-extensions.def: New entry for profile
7086 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7087 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7090 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7093 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7094 address check not strict.
7096 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7098 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7100 (store_pair_lanes<mode>): Likewise.
7101 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7102 enum value 'ADDR_QUERY_LDP_STP_N'.
7103 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7104 (aarch64_print_address_internal): Add declaration.
7105 (aarch64_print_ldpstp_address): Remove.
7106 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7107 (aarch64_print_operand): Change printing of 'y'.
7108 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7109 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7110 'true' rather than '1'.
7111 * config/aarch64/constraints.md (Uml): Likewise.
7112 (Uml): Rename to Umn.
7115 2018-07-19 Richard Biener <rguenther@suse.de>
7117 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7119 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7120 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7121 (vn_reference_hasher): Likewise.
7122 (struct vn_tables_s): Remove obstack and alloc-pool members.
7123 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7124 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7125 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7126 (vn_reference_insert_pieces): Likewise.
7127 (alloc_vn_nary_op_noinit): Adjust.
7128 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7129 (vn_phi_eq): Adjust.
7130 (shared_lookup_phiargs): Remove.
7131 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7132 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7133 (visit_reference_op_call): Likewise.
7134 (copy_nary, copy_phi, copy_reference): Remove.
7135 (process_scc): Rewind the obstack when iterating. Do not
7136 copy the elements to valid_info but just move them from one
7137 hashtable to the other.
7138 (allocate_vn_table): Adjust.
7139 (free_vn_table): Likewise.
7140 (init_scc_vn): Likewise.
7141 (free_scc_vn): Likewise.
7143 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
7146 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7147 indirect_return as function type attribute.
7148 (ix86_attribute_table): Change indirect_return to function
7150 * doc/extend.texi: Update indirect_return attribute.
7152 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
7154 * wide-int.h (widest2_int): New.
7155 * gimple-fold.c (arith_overflowed_p): Use it.
7156 * tree.h (widest2_int_cst): New.
7157 * tree-vrp.c (wide_int_binop_overflow): Rename from
7158 vrp_int_const_binop.
7159 Rewrite to work on trees.
7160 (extract_range_from_multiplicative_op_1): Abstract code to...
7161 (wide_int_range_min_max): ...here.
7162 (wide_int_range_cross_product): ...and here.
7163 (extract_range_from_binary_expr_1): Abstract overflow code to...
7164 (wide_int_range_mult_wrapping): ...here.
7165 * tree-vrp.h (wide_int_range_cross_product): New.
7166 (wide_int_range_mult_wrapping): New.
7168 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
7169 Julia Koval <julia.koval@intel.com>
7171 * config/i386/x86-tune-costs.h (skylake_memcpy,
7172 skylake_memset): Replace rep_prefix with unrolling for size 512.
7174 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
7177 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7178 comparision with EQ_EXPR in last stmt.
7180 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
7182 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7183 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7184 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7185 previously known as "PowerPC AltiVec Built-in Functions". Move
7186 some material to new subsubsections "PowerPC AltiVec Built-in
7187 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7189 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7190 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7191 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7192 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7194 2018-07-18 Richard Biener <rguenther@suse.de>
7196 PR tree-optimization/86557
7197 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7200 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
7202 * config/s390/s390.c (s390_function_profiler): Generate CFI.
7204 2018-07-17 Jeff Law <law@redhat.com>
7206 * config/arm/arm.c (get_label_padding): Update for recent
7207 changes to label_to_alignment.
7209 PR tree-optimization/86010
7210 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7212 * config/mips/mips.c (vr4130_align_insns): Update for recent
7213 changes to label_to_alignment.
7215 * config/frv/frv.c (frv_label_align): Update for recent changes
7216 to label_to_alignment.
7218 * config/nios2/nios2.c (nios2_label_align): Update for recent
7219 changes which dropped ALIGN_LABELS_LOG.
7221 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
7223 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7224 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7226 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
7228 * config/arc/arc.c (arc_label_align): Use align_labels instead of
7229 deprecated align_labels_log.
7231 2018-07-17 Richard Biener <rguenther@suse.de>
7234 * dwarf2out.c (init_sections_and_labels): Always generate
7235 a debug_line_str_section for early LTO debug.
7236 (dwarf2out_finish): Reset debug_line_str_hash output early.
7237 Bump counter for extra dwarf5 .debug_loc labels to not conflict
7239 (dwarf2out_early_finish): Output debug_line_str.
7241 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7243 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7244 index register on z196 or later.
7246 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7248 * config/s390/s390.c (s390_default_align): Set default function
7250 (s390_override_options_after_change): Call s390_default align.
7251 (s390_option_override_internal): Call s390_default_align.
7252 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7254 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7257 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7258 to _looptemp_ clauses, other than the first two.
7260 2018-07-17 Martin Liska <mliska@suse.cz>
7262 * opts.c: Do not enable OPT_falign_* for -Os.
7264 2018-07-17 Martin Liska <mliska@suse.cz>
7266 * align.h (MAX_CODE_ALIGN): New.
7267 (MAX_CODE_ALIGN_VALUE): New.
7268 * common/config/i386/i386-common.c (ix86_handle_option):
7269 (MAX_CODE_ALIGN): Moved to align.h.
7270 * final.c (MAX_CODE_ALIGN): Likewise.
7271 * opts.c (parse_and_check_align_values):
7272 (MAX_CODE_ALIGN): Likewise.
7273 (MAX_CODE_ALIGN_VALUE): Likewise.
7275 2018-07-17 Martin Liska <mliska@suse.cz>
7277 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7278 in order to fulfil coding style.
7279 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7280 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7281 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7282 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7283 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7284 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7285 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7286 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7287 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7289 2018-07-17 Martin Liska <mliska@suse.cz>
7291 * align.h: New file.
7292 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7294 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7295 align_flags of label_to_alignment.
7296 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7298 * config/m68k/m68k.c: Do not use removed align_labels_value and
7300 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7301 (LOOP_ALIGN): Likewise.
7302 (LABEL_ALIGN): Likewise.
7303 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7304 Remove not used macro.
7305 (rs6000_loop_align): Change return type to align_flags.
7306 (rs6000_loop_align_max_skip): Remove.
7307 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7308 Change return type to align_flags.
7309 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7310 Remove not used macro.
7311 (rs6000_loop_align): Change return type to align_flags.
7312 (rs6000_loop_align_max_skip): Remove.
7313 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7314 * config/rx/rx-protos.h (rx_align_for_label): Make it
7316 * config/rx/rx.c (rx_align_for_label): Change return type
7318 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7320 into align_flags class.
7321 (LABEL_ALIGN): Likewise.
7322 (LOOP_ALIGN): Likewise.
7323 * config/s390/s390.c (s390_label_align): Use align_flags
7325 (s390_asm_output_function_label): Likewise.
7326 * config/sh/sh.c (sh_override_options_after_change):
7327 Use align_flags class directly without macros.
7328 (find_barrier): Likewise.
7329 (barrier_align): Likewise.
7330 (sh_loop_align): Likewise.
7331 * config/spu/spu.c (spu_option_override):
7332 Use align_flags_tuple::get_value instead of removed macros.
7333 (spu_sched_init): Likewise.
7334 * config/spu/spu.h (GTY): Likewise.
7335 * config/visium/visium.c (visium_option_override):
7336 Set "8" as default secondary alignment.
7337 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
7338 in order to guarantee secondary alignment of 8.
7339 * coretypes.h: Include align.h header file.
7340 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
7341 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
7342 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
7343 * doc/tm.texi.in: Likewise.
7344 * final.c (struct label_alignment): Remove not used structure.
7345 (LABEL_ALIGN): Change type to align_flags.
7346 (LOOP_ALIGN): Likewise.
7347 (JUMP_ALIGN): Likewise.
7348 (default_loop_align_max_skip): Remove.
7349 (default_label_align_max_skip): Likewise.
7350 (default_jump_align_max_skip): Likewise.
7351 (default_label_align_after_barrier_max_skip):
7352 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
7353 (LABEL_TO_MAX_SKIP): Remove.
7354 (label_to_alignment): Return align_flags type instead of integer.
7355 (label_to_max_skip): Remove.
7356 (align_fuzz): Use align_flags type.
7357 (compute_alignments): Use align_flags type and use align_flags::max
7358 to combine multiple alignments.
7359 (grow_label_align): Grow vec instead of C array.
7360 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
7361 (shorten_branches): Use align_flags type and use align_flags::max
7362 to combine multiple alignments.
7363 (final_scan_insn_1): Remove usage of secondary alignment that comes
7364 from label alignment, but instead use proper secondary alignment
7365 which is computed in grow_label_align.
7366 * flags.h (struct align_flags_tuple): Move to align.h.
7367 (struct align_flags): Likewise.
7368 (state_align_loops): Rename to align_loops.
7369 (state_align_jumps): Rename to align_jumps.
7370 (state_align_labels): Rename to align_labels.
7371 (state_align_functions): Rename to align_functions.
7372 (align_loops_log): Remove.
7373 (align_jumps_log): Remove.
7374 (align_labels_log): Remove.
7375 (align_functions_log): Remove.
7376 (align_loops_max_skip): Remove.
7377 (align_jumps_max_skip): Remove.
7378 (align_labels_max_skip): Remove.
7379 (align_functions_max_skip): Remove.
7380 (align_loops_value): Remove.
7381 (align_jumps_value): Remove.
7382 (align_labels_value): Remove.
7383 (align_functions_value): Remove.
7384 * output.h (label_to_alignment): Change return type to align_flags.
7385 (label_to_max_skip): Remove.
7386 * target.def: Remove loop_align_max_skip, label_align_max_skip,
7387 jump_align_max_skip macros.
7388 * targhooks.h (default_loop_align_max_skip): Remove.
7389 (default_label_align_max_skip): Likewise.
7390 (default_jump_align_max_skip): Likewise.
7391 (default_label_align_after_barrier_max_skip): Remove.
7392 * toplev.c (read_log_maxskip): Use ::normalize function.
7393 (parse_N_M): Remove not used argument and also call ::normalize.
7394 (parse_alignment_opts): Do not pass unused arguments.
7395 * varasm.c (assemble_start_function): Use directly align_functions
7396 instead of removed macros.
7397 * system.h: Do not poison removed macros.
7399 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7402 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
7403 and cond temporaries don't have reference type if iterator has
7404 pointer type. For init use &for_pre_body instead of pre_p if
7405 for_pre_body is non-empty.
7407 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7409 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
7410 double-double modes to SFmode directly directly.
7411 (trunc<mode>sf2_fprs): Delete.
7413 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7415 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
7416 for conversions between IFmode and the decimal floating point modes.
7417 (init_float128_ieee): Use the correct names for conversions between
7418 KFmode and the decimal floating point modes.
7420 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7422 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
7423 for the conversions between TDmode and IFmode.
7424 (init_float128_ieee): Use more correct names for the conversions
7425 between TDmode and KFmode.
7427 2018-07-16 Jakub Jelinek <jakub@redhat.com>
7429 PR tree-optimization/86526
7430 * builtins.c (expand_builtin_memcmp): Formatting fixes.
7431 (inline_expand_builtin_string_cmp): Likewise.
7432 (inline_string_cmp): Likewise. Use c_readstr instead of
7433 builtin_memcpy_read_str. Add unit_mode temporary.
7435 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
7438 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
7439 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
7441 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7443 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
7444 Alphabetize prototypes of built-in functions, separating out
7445 built-in functions that are listed in this section but should be
7446 described elsewhere.
7448 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
7451 * expmed.c (emit_store_flag): Do not emit setcc followed by a
7452 conditional move when trapping comparison was split to a
7453 non-trapping one (and vice versa).
7455 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7457 * config/s390/s390.c (s390_function_profiler): Generate nops
7458 instead of profiler call sequences.
7459 * config/s390/s390.opt: Add the new option.
7461 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7463 * config/s390/s390.c (s390_function_profiler): Generate
7464 __mcount_loc section.
7465 * config/s390/s390.opt: Add the new option.
7467 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7469 * common.opt: Add the new warning.
7470 * config/s390/s390.c (s390_function_profiler): Emit "brasl
7471 %r0,__fentry__" when -mfentry is specified.
7472 (s390_option_override_internal): Disallow -mfentry for 31-bit
7474 * config/s390/s390.opt: Add the new option.
7476 2018-07-16 Richard Biener <rguenther@suse.de>
7479 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
7480 for function-local FUNCTION_DECL and RESULT_DECL immediately.
7482 2018-07-16 Martin Liska <mliska@suse.cz>
7485 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
7488 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
7490 * config/arc/arcHS.md: Update ARCHS scheduling rules.
7492 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
7494 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
7496 * config/arc/arc-cpus.def (hs4x): New cpu.
7498 * config/arc/arc-tables.opt: Regenerate.
7499 * config/arc/arc.c (arc_sched_issue_rate): New function.
7500 (TARGET_SCHED_ISSUE_RATE): Define.
7501 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
7502 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
7504 (attr tune): Add ARCHS4x tune values.
7505 (attr tune_dspmpy): Define.
7506 (*tst): Correct instruction type.
7507 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
7508 * config/arc/arcHS4x.md: New file.
7509 * config/arc/fpu.md: Update instruction type attributes.
7510 * config/arc/t-multilib: Regenerate.
7512 2018-07-16 Tom de Vries <tdevries@suse.de>
7515 * var-tracking.c (vt_initialize): Fix pre_dec handling.
7517 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7519 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
7522 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
7524 PR tree-optimization/86514
7525 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
7526 conversion to a boolean type from a type with greater precision.
7528 2018-07-16 Tom de Vries <tdevries@suse.de>
7530 * var-tracking.c (vt_initialize): Print adjusted insn slim if
7531 dump_flags request TDF_SLIM.
7533 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
7535 * fold-const.c (int_const_binop_1): Abstract...
7536 (wide_int_binop): ...wide int code here.
7537 (poly_int_binop): ...poly int code here.
7538 Abstract the rest of int_const_binop_1 into int_const_binop.
7539 * fold-const.h (wide_int_binop): New.
7540 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
7541 Remove useless PLUS/MINUS_EXPR case.
7542 (zero_nonzero_bits_from_vr): Move wide int code...
7543 (zero_nonzero_bits_from_bounds): ...here.
7544 (extract_range_from_binary_expr_1): Move mask optimization code...
7545 (range_easy_mask_min_max): ...here.
7546 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
7547 (range_easy_mask_min_max): New.
7549 2018-07-15 Jeff Law <law@redhat.com>
7552 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
7555 2018-07-14 Jim Wilson <jimw@sifive.com>
7557 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
7559 2018-07-14 Paul Koning <ni1d@arrl.net>
7561 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
7563 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
7565 * lto-streamer-out.c (copy_function_or_variable): Dump info about
7568 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
7569 Steve Munroe <munroesj52@gmail.com>
7571 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
7572 (_mm_andnot_si128): Likewise.
7573 (_mm_or_si128): Likewise.
7574 (_mm_xor_si128): Likewise.
7576 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
7579 * builtins.c (expand_builtin_memcmp): Inline the calls first
7580 when result_eq is false.
7581 (expand_builtin_strcmp): Inline the calls first.
7582 (expand_builtin_strncmp): Likewise.
7583 (inline_string_cmp): New routine. Expand a string compare
7584 call by using a sequence of char comparison.
7585 (inline_expand_builtin_string_cmp): New routine. Inline expansion
7586 a call to str(n)cmp/memcmp.
7587 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
7589 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
7591 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
7593 * config/arm/driver-arm.c: Include arm-native.h.
7594 (host_detect_local_cpu): Use auto-generated data tables.
7595 (vendors, arm_cpu_table): Delete. Move part information to ...
7596 * config/arm/arm-cpus.in: ... here.
7597 * config/arm/parsecpu.awk (gen_native): New function.
7598 (vendor, part): New CPU fields.
7599 (END): Add support for building the native CPU detection tables.
7600 * config/arm/t-arm (arm-native.h): Add build rule.
7601 (driver-arm.o): Add dependency on arm-native.h.
7603 2018-07-13 Richard Biener <rguenther@suse.de>
7606 * match.pd (addr1 - addr2): Allow either of the operand to
7609 2018-07-13 Tom de Vries <tdevries@suse.de>
7611 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
7614 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
7616 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
7617 arrays instead of numbered variables.
7619 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
7621 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
7622 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
7624 2018-07-13 Richard Biener <rguenther@suse.de>
7627 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
7628 instead of get_context_die.
7630 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
7631 Richard Biener <rguenther@suse.de>
7634 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
7635 that the loop latch destination where phi is defined.
7637 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
7639 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
7640 (riscv_expand_epilogue): Add assertion to check interrupt mode.
7641 (riscv_set_current_function): Extract getting interrupt type to new
7643 (riscv_get_interrupt_type): New function.
7644 (riscv_merge_decl_attributes): New function, checking interrupt type is
7646 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
7648 2018-07-12 Paul Koning <ni1d@arrl.net>
7650 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
7653 2018-07-12 Paul Koning <ni1d@arrl.net>
7655 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
7656 zero reference, add doloop_end instead.
7657 * doc/md.texi (decrement_and_branch_until_zero): Remove.
7658 (Looping patterns): Remove decrement_and_branch_until_zero. Add
7659 detail for doloop_end.
7661 2018-07-12 Martin Sebor <msebor@redhat.com>
7664 * attribs.c (decl_attributes): Reject conflicting attributes before
7665 calling attribute handlers.
7667 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7669 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
7671 (gcc::dump_manager::get_dump_file_name): likewise.
7672 (dump_begin): Likewise.
7673 * dumpfile.h (dump_begin): Update prototype.
7674 (gcc::dump_manager::get_dump_file_name,
7675 gcc::dump_manager::get_dump_file_name): Update prototype.
7677 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7679 * internal-fn.h (vectorizable_internal_fn_p): New function.
7680 * tree-vect-slp.c (compatible_calls_p): Likewise.
7681 (vect_build_slp_tree_1): Remove nops argument. Handle calls
7682 to internal functions.
7683 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
7685 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7687 * fold-const.h (inverse_conditions_p): Declare.
7688 * fold-const.c (inverse_conditions_p): New function.
7689 * match.pd: Use inverse_conditions_p. Add folds of view_converts
7690 that test the inverse condition of a conditional internal function.
7691 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
7692 * internal-fn.c (internal_fn_mask_index): Handle conditional
7694 (vectorized_internal_fn_supported_p): New function.
7695 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
7696 (any_pred_load_store): Replace with...
7697 (need_to_predicate): ...this new variable.
7698 (redundant_ssa_names): New variable.
7699 (ifcvt_can_use_mask_load_store): Move initial checks to...
7700 (ifcvt_can_predicate): ...this new function. Handle tree codes
7701 for which a conditional internal function exists.
7702 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
7703 instead of ifcvt_can_use_mask_load_store. Update after variable
7705 (predicate_load_or_store): New function, split out from
7706 predicate_mem_writes.
7707 (check_redundant_cond_expr): New function.
7708 (value_available_p): Likewise.
7709 (predicate_rhs_code): Likewise.
7710 (predicate_mem_writes): Rename to...
7711 (predicate_statements): ...this. Use predicate_load_or_store
7712 and predicate_rhs_code.
7713 (combine_blocks, tree_if_conversion): Update after above name changes.
7714 (ifcvt_local_dce): Handle redundant_ssa_names.
7715 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
7716 general conditional functions.
7717 * tree-vect-stmts.c (vectorizable_call): Likewise.
7719 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7720 Alan Hayward <alan.hayward@arm.com>
7721 David Sherwood <david.sherwood@arm.com>
7723 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
7724 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
7725 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
7726 plus and minus and convert them into IFN_COND_FMA-based sequences.
7727 (convert_mult_to_fma): Handle conditional plus and minus.
7729 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7731 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
7732 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
7733 (cond_fnms_optab): New optabs.
7734 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
7736 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
7737 * internal-fn.h (get_conditional_internal_fn): Declare.
7738 (get_unconditional_internal_fn): Likewise.
7739 * internal-fn.c (cond_ternary_direct): New macro.
7740 (expand_cond_ternary_optab_fn): Likewise.
7741 (direct_cond_ternary_optab_supported_p): Likewise.
7742 (FOR_EACH_COND_FN_PAIR): Likewise.
7743 (get_conditional_internal_fn): New function.
7744 (get_unconditional_internal_fn): Likewise.
7745 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
7746 (gimple_match_op::gimple_match_op): Add a new overload for 5
7748 (gimple_match_op::set_op): Likewise.
7749 (gimple_resimplify5): Declare.
7750 * genmatch.c (decision_tree::gen): Generate simplifications for
7752 * gimple-match-head.c (gimple_simplify): Define an overload for
7753 5 operands. Handle calls with 5 arguments in the top-level overload.
7754 (convert_conditional_op): Handle conversions from unconditional
7755 internal functions to conditional ones.
7756 (gimple_resimplify5): New function.
7757 (build_call_internal): Pass a fifth operand.
7758 (maybe_push_res_to_seq): Likewise.
7759 (try_conditional_simplification): Try converting conditional
7760 internal functions to unconditional internal functions.
7761 Handle 3-operand unconditional forms.
7762 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
7763 Define ternary equivalents of the current rules for binary conditional
7765 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
7767 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
7768 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
7769 (optab): Handle them.
7770 (SVE_COND_FP_TERNARY): New int iterator.
7771 (sve_fmla_op, sve_fmad_op): New int attributes.
7772 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
7773 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
7774 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
7776 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7778 * target.def (preferred_else_value): New target hook.
7779 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
7780 * doc/tm.texi: Regenerate.
7781 * targhooks.h (default_preferred_else_value): Declare.
7782 * targhooks.c (default_preferred_else_value): New function.
7783 * internal-fn.h (conditional_internal_fn_code): Declare.
7784 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
7785 (get_conditional_internal_fn): Use it.
7786 (conditional_internal_fn_code): New function.
7787 * gimple-match.h (gimple_match_cond): New struct.
7788 (gimple_match_op): Add a cond member function.
7789 (gimple_match_op::gimple_match_op): Update all forms to take a
7791 * genmatch.c (expr::gen_transform): Use the same condition as res_op
7792 for the suboperation, but don't specify a particular else_value.
7793 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
7794 (visit_nary_op, visit_reference_op_load): Pass
7795 gimple_match_cond::UNCOND to the gimple_match_op constructor.
7796 * gimple-match-head.c: Include tree-eh.h
7797 (convert_conditional_op): New function.
7798 (maybe_resimplify_conditional_op): Likewise.
7799 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
7800 (gimple_resimplify2): Likewise.
7801 (gimple_resimplify3): Likewise.
7802 (gimple_resimplify4): Likewise.
7803 (maybe_push_res_to_seq): Return null for conditional operations.
7804 (try_conditional_simplification): New function.
7805 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
7807 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
7809 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
7811 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
7813 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7815 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7817 (hash_tree): Do not hash DECL_FCONTEXT
7818 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
7819 Do not stream DECL_FCONTEXT.
7820 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
7821 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
7823 2018-07-12 Richard Biener <rguenther@suse.de>
7826 * dwarf2out.c (gen_block_die): Only output blocks when they have
7827 at least one !DECL_IGNORED_P variable.
7829 2018-07-12 Richard Biener <rguenther@suse.de>
7832 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
7833 Remove -mieee-fp handling.
7835 2018-07-12 Richard Biener <rguenther@suse.de>
7837 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
7838 left-over from last patch.
7840 2018-07-12 Jakub Jelinek <jakub@redhat.com>
7842 PR tree-optimization/86492
7843 * gimple-ssa-store-merging.c
7844 (imm_store_chain_info::coalesce_immediate_stores): Call
7845 check_no_overlap even for the merge_overlapping case. Formatting fix.
7847 2018-07-12 Richard Biener <rguenther@suse.de>
7850 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
7851 move possibly trapping operations into the conditional.
7853 2018-07-12 Richard Biener <rguenther@suse.de>
7855 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
7856 (vn_lookup_simplify_result): Remove recursion limit applied
7858 (vn_nary_build_or_lookup_1): Adjust.
7859 (try_to_simplify): Likewise.
7860 * gimple-match-head.c (gimple_resimplify1): Instead apply one
7862 (gimple_resimplify2): Likewise.
7863 (gimple_resimplify3): Likewise.
7864 (gimple_resimplify4): Likewise.
7866 2018-07-11 Jakub Jelinek <jakub@redhat.com>
7868 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
7869 Use __mmask64 type instead of __mmask8 for __M argument.
7870 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
7871 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
7873 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
7874 __mmask16 for __M argument.
7875 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
7876 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
7877 to __mmask16 instead of __mmask8.
7878 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
7879 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
7880 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
7881 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
7882 instead of __mmask16 for __U argument.
7883 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
7884 __mmask16 instead of __mmask8 for __U argument.
7885 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
7887 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
7889 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
7891 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
7893 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
7895 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
7897 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
7899 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
7901 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
7903 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
7905 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
7906 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
7907 return type as well as __M argument type and all casts from __mmask8
7909 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
7910 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
7911 return type as well as __M argument type and all casts from __mmask8
7913 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
7914 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
7915 return type as well as __M argument type and all casts from __mmask8
7917 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
7918 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
7919 return type as well as __M argument type and all casts from __mmask8
7921 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
7922 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
7925 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
7927 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
7928 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
7931 2018-07-11 Paul Koning <ni1d@arrl.net>
7933 * doc/md.texi (define_subst): Document how multiple occurrences of
7934 the same argument in the replacement pattern are handled.
7936 2018-07-11 Paul Koning <ni1d@arrl.net>
7938 * doc/extend.texi (Common Variable Attributes): Move "mode" into
7940 (Common Type Attributes): Add "mode" attribute.
7942 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
7944 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
7945 stream DECL_ORIGINAL_TYPE.
7946 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
7947 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
7948 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
7949 Do not walk original type.
7950 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
7952 (write_ts_decl_non_common_tree_pointers): Do not stream
7954 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
7955 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
7957 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7959 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
7960 thread twice from the same starting edge.
7962 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7964 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
7965 * gimple.c (gimple_call_nonnull_result_p): ...here...
7966 (gimple_call_nonnull_arg): ...and here.
7967 * gimple.h (gimple_call_nonnull_result_p): New.
7968 (gimple_call_nonnull_arg): New.
7970 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
7972 * config/arm/arm-cpus.in: Move information from fpu field of each
7973 cpu definition to the isa field.
7974 * config/arm/parsecpu.awk (fpu): Delete match rule.
7975 (gen_comm_data): Don't add bits from the CPU's FPU entry.
7977 2018-07-11 Richard Biener <rguenther@suse.de>
7980 * dwarf2out.c (init_sections_and_labels): Use
7981 output_asm_line_debug_info consistently.
7982 (dwarf2out_early_finish): Likewise.
7983 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
7986 2018-07-11 Richard Biener <rguenther@suse.de>
7988 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
7989 Rework father_bb setting in a way to avoid propagating constants
7990 multiple times on a loop body.
7992 2018-07-10 Mark Wielaard <mark@klomp.org>
7995 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
7997 2018-07-10 Richard Biener <rguenther@suse.de>
7999 * hash-map.h (hash_map::iterator::operator*): Return
8000 references to key and value.
8002 2018-07-10 Jakub Jelinek <jakub@redhat.com>
8005 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8006 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8007 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8008 (gimplify_omp_for): For composite loops, move outer
8009 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8010 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
8011 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
8012 TREE_LIST for both the original class iterator and the "last" helper
8013 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
8014 loop, remember has_decl_expr from outer composite loops for the
8015 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8017 2018-07-09 Martin Sebor <msebor@redhat.com>
8021 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8022 accessing implicitly initialized array elements.
8023 * expr.c (string_constant): Handle string initializers of
8024 character arrays within aggregates.
8025 * gimple-fold.c (fold_array_ctor_reference): Add argument.
8026 Store element offset. As a special case, handle zero size.
8027 (fold_nonarray_ctor_reference): Same.
8028 (fold_ctor_reference): Add argument. Store subobject offset.
8029 * gimple-fold.h (fold_ctor_reference): Add argument.
8031 2018-07-09 Paul Koning <ni1d@arrl.net>
8033 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8034 (pdp11_insn_cost): New function.
8035 (pdp11_md_asm_adjust): New function.
8036 (TARGET_INVALID_WITHIN_DOLOOP): Define.
8037 (pdp11_rtx_costs): Update to match machine better.
8038 (output_addr_const_pdp11): Correct format mismatch warnings.
8039 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8040 * config/pdp11/pdp11.md: General change to add base_cost and/or
8041 length attributes for use by new pdp11_insn_cost function.
8042 (MIN_BRANCH): Correct definition.
8044 (doloop_end): Use standard pattern name for looping pattern.
8045 (doloop_end_nocc): New.
8046 (movsf): Add another constraint alternative.
8047 (zero_extendqihi2): Add constraint alternatives for not in place
8049 (zero_extendhisi2): Remove.
8050 (shift patterns): Add CC handling variants.
8054 (define_peephole2): New peephole to recognize mov that sets CC for
8057 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8059 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8060 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8063 2018-07-09 Jakub Jelinek <jakub@redhat.com>
8066 * real.c (real_nextafter): Return true if result is denormal.
8068 2018-07-09 Martin Liska <mliska@suse.cz>
8070 * common.opt: Add back wrongly removed attribute.
8072 2018-07-09 Richard Biener <rguenther@suse.de>
8075 * dwarf2out.c (gen_block_die): For an early generated DIE
8076 always output high/low PC attributes.
8078 2018-07-09 Tom de Vries <tdevries@suse.de>
8080 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8081 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8082 onto VAR_DECL with abstract origin.
8084 2018-07-07 Jim Wilson <jimw@sifive.com>
8086 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8088 2018-07-07 Tom de Vries <tdevries@suse.de>
8090 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8092 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
8094 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8096 (combine_bound): Use wide-int overflow calculation instead of
8098 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8100 * fold-const.c (int_const_binop_2): Same.
8101 (extract_muldiv_1): Same.
8102 (fold_div_compare): Same.
8103 (fold_abs_const): Same.
8105 * poly-int.h (add): Same.
8109 * predict.c (predict_iv_comparison): Same.
8110 * profile-count.c (slow_safe_scale_64bit): Same.
8111 * simplify-rtx.c (simplify_const_binary_operation): Same.
8112 * tree-chrec.c (tree_fold_binomial): Same.
8113 * tree-data-ref.c (split_constant_offset_1): Same.
8114 * tree-if-conv.c (idx_within_array_bound): Same.
8115 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8116 * tree-ssa-phiopt.c (minmax_replacement): Same.
8117 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8118 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8119 * vr-values.c (vr_values::adjust_range_with_scev): Same.
8120 * wide-int.cc (wi::add_large): Same.
8121 (wi::mul_internal): Same.
8122 (wi::sub_large): Same.
8123 (wi::divmod_internal): Same.
8124 * wide-int.h: Change overflow type to overflow_type for neg, add,
8125 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8126 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8127 mul_internal, divmod_internal.
8128 (overflow_type): New enum.
8129 (accumulate_overflow): New.
8131 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
8133 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8134 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8136 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8138 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8139 argument is checked for zero before entering loop, avoid checking again.
8141 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8143 * gimplify.h (generic_expr_could_trap_p): Set as global function.
8144 * gimplify.h (generic_expr_could_trap_p): Likwise.
8145 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8147 2018-07-06 Jakub Jelinek <jakub@redhat.com>
8149 PR tree-optimization/86401
8150 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8151 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8152 (fold_bit_and_mask): ... here. New helper function for match.pd.
8153 * fold-const.h (fold_bit_and_mask): Declare.
8154 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8156 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
8159 * target.def (translate_mode_attribute): New hook.
8160 * targhooks.h (default_translate_mode_attribute): Declare.
8161 * targhooks.c (default_translate_mode_attribute): New function.
8162 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8163 * doc/tm.texi: Regenerate.
8164 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8165 (rs6000_translate_mode_attribute): New function.
8167 2018-07-06 Paul Koning <ni1d@arrl.net>
8169 * doc/md.texi (define_split): Document DONE and FAIL.
8170 (define_peephole2): Ditto.
8172 2018-07-05 Jeff Law <law@redhat.com>
8174 PR tree-optimization/86010
8175 * tree-ssa-dse.c (compute_trims): More aggressively trim at
8176 both the head and tail of mem* and str* calls.
8178 2018-07-05 Jim Wilson <jimw@sifive.com>
8180 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8182 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
8184 * config/aarch64/aarch64-simd.md: correct flags text for
8185 MIN_EXPR replacement.
8187 2018-07-05 James Clarke <jrtc27@jrtc27.com>
8189 * configure: Regenerated.
8191 2018-07-05 Carl Love <cel@us.ibm.com>
8193 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8194 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8195 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8196 VSX_BUILTIN_DOUBLEL_V4SF.
8198 2018-07-05 Martin Sebor <msebor@redhat.com>
8201 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8202 than its domain to compute its the upper bound of a char array.
8204 2018-07-05 Nathan Sidwell <nathan@acm.org>
8206 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8207 * doc/cpp.texi: Update comment.
8208 * doc/tm.texi: Rebuilt.
8209 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8210 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8211 * doc/extend.texi (Backwards Compatibility): Clarify it is system
8212 headers affected by extern "C".
8213 * system.h: Poison NO_IMPLICIT_EXTERN_C.
8214 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8215 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8216 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8217 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8218 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8219 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8220 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8221 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8222 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8223 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8224 config/sparc/sp64-elf.h, config/spu/spu.h,
8225 config/stormy16/stormy16.h, config/v850/v850.h,
8226 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8227 define NO_IMPLICIT_EXTERN_C.
8228 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8230 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8233 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8234 instead of GET_MODE_SIZE when comparing Units.
8236 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8239 * rtlanal.c (set_noop_p): Constrain on mode change,
8240 include hard-reg-set.h
8242 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8244 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8246 2018-07-05 Jakub Jelinek <jakub@redhat.com>
8249 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8252 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8254 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8256 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8259 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8261 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8263 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
8266 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8268 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
8270 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8271 neg_*_op* variables.
8273 2018-07-04 Martin Liska <mliska@suse.cz>
8275 * tree-switch-conversion.c: Define
8276 max_ratio_for_speed and max_ratio_for_size constants.
8278 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
8279 Martin Liska <mliska@suse.cz>
8284 * common.opt: Rename align options with 'str_' prefix.
8285 * common/config/i386/i386-common.c (set_malign_value): New
8287 (ix86_handle_option): Use it to set -falign-* options/
8288 * config/aarch64/aarch64-protos.h (struct tune_params): Change
8289 type from int to string.
8290 * config/aarch64/aarch64.c: Update default values from int
8292 * config/alpha/alpha.c (alpha_override_options_after_change):
8294 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8295 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8296 max skip conditionally.
8297 * config/i386/freebsd.h (SUBALIGN_LOG): New.
8298 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8299 max skip conditionally.
8300 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8301 max skip conditionally.
8302 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8303 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8304 max skip conditionally.
8305 * config/i386/i386.c (struct ptt): Change type from int to
8307 (ix86_default_align): Set default values.
8308 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8309 max skip conditionally.
8310 * config/i386/iamcu.h (SUBALIGN_LOG): New.
8311 (ASM_OUTPUT_MAX_SKIP_ALIGN):
8312 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8313 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8314 max skip conditionally.
8315 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8316 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8317 * config/i386/x86-64.h (SUBALIGN_LOG): New.
8318 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8319 max skip conditionally.
8320 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8321 * config/ia64/ia64.c (ia64_option_override): Set default values
8322 for alignment options.
8323 * config/m68k/m68k.c: Handle new str_align_* options.
8324 * config/mips/mips.c (mips_set_compression_mode): Change
8326 (mips_option_override): Set default values for options.
8327 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8329 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8331 * config/rx/rx.c (rx_option_override): Likewise.
8332 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8333 (LABEL_ALIGN): Use align_labels_log.
8334 (LOOP_ALIGN): Use align_loops_align.
8335 * config/s390/s390.c (s390_asm_output_function_label): Use new
8337 * config/sh/sh.c (sh_override_options_after_change):
8338 Change type of constants.
8339 * config/spu/spu.c (spu_sched_init): Likewise.
8340 * config/sparc/sparc.c (sparc_option_override): Set default
8342 * config/visium/visium.c (visium_option_override): Likewise.
8343 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
8344 emit p2align format with last argument if it's not needed.
8345 * doc/invoke.texi: Document extended format of -falign-*.
8346 * final.c: Use align_labels alignment.
8347 * flags.h (struct target_flag_state): Change type to use
8349 (struct align_flags_tuple): New.
8350 (struct align_flags): Likewise.
8351 (align_loops_log): Redefine macro to use new types.
8352 (align_loops_max_skip): Redefine macro to use new types.
8353 (align_jumps_log): Redefine macro to use new types.
8354 (align_jumps_max_skip): Redefine macro to use new types.
8355 (align_labels_log): Redefine macro to use new types.
8356 (align_labels_max_skip): Redefine macro to use new types.
8357 (align_functions_log): Redefine macro to use new types.
8358 (align_loops): Redefine macro to use new types.
8359 (align_jumps): Redefine macro to use new types.
8360 (align_labels): Redefine macro to use new types.
8361 (align_functions): Redefine macro to use new types.
8362 (align_functions_max_skip): Redefine macro to use new types.
8363 (align_loops_value): New macro.
8364 (align_jumps_value): New macro.
8365 (align_labels_value): New macro.
8366 (align_functions_value): New macro.
8367 * function.c (invoke_set_current_function_hook): Propagate
8368 alignment values from flags to global variables default in
8370 * ipa-icf.c (sem_function::equals_wpa): Use
8371 cl_optimization_option_eq instead of memcmp.
8372 * lto-streamer.h (cl_optimization_stream_out): Support streaming
8374 (cl_optimization_stream_in): Likewise.
8375 * optc-save-gen.awk: Support strings in cl_optimization.
8376 * opth-gen.awk: Likewise.
8377 * opts.c (finish_options): Remove error checking of invalid
8379 (MAX_CODE_ALIGN): Remove.
8380 (MAX_CODE_ALIGN_VALUE): Likewise.
8381 (parse_and_check_align_values): New function.
8382 (check_alignment_argument): Likewise.
8383 (common_handle_option): Use check_alignment_argument.
8384 * opts.h (parse_and_check_align_values): Declare.
8385 * toplev.c (init_alignments): Remove.
8386 (read_log_maxskip): New.
8387 (parse_N_M): Likewise.
8388 (parse_alignment_opts): Likewise.
8389 (backend_init_target): Remove usage of init_alignments.
8390 * toplev.h (parse_alignment_opts): Declare.
8391 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
8393 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
8394 * tree.c (cl_option_hasher::equal): New.
8395 * varasm.c: Use new global macros.
8397 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
8399 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
8400 Use a simpler align directive also if MAXSKIP = ALIGN-1.
8401 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8402 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8403 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8404 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8405 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
8406 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
8407 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
8408 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8409 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8410 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8411 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8413 2018-07-04 Martin Liska <mliska@suse.cz>
8414 Jonathan Wakely <jwakely@redhat.com>
8416 * coverage.c: Use correct type.
8417 * doc/invoke.texi: Language correction.
8419 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
8422 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
8423 ENDBRANCH for non-tail call which may return via indirect branch.
8424 * doc/extend.texi: Document indirect_return attribute.
8426 2018-07-03 Martin Sebor <msebor@redhat.com>
8428 PR tree-optimization/86274
8429 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
8431 (format_floating): Correct handling of infinities and NaNs.
8433 2018-07-03 Martin Sebor <msebor@redhat.com>
8435 * print-tree.c (print_real_cst): New function.
8436 (print_node_brief): Call it.
8437 (print_node): Ditto.
8439 2018-07-03 Jeff Law <law@redhat.com>
8441 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
8442 into a single pattern.
8444 * config/h8300/h8300.md (ors code_iterator): New.
8445 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
8446 a single pattern and single splitter.
8447 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
8448 (iorqi3_1, xorqi3_1): Likewise.
8449 (iorqi3, xorqi3 expanders): Similarly.
8451 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
8452 (movmd_internal) into a single pattern using the P mode iterator.
8453 (movmd splitters): Similarly.
8454 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
8455 (movsd splitters): Similarly.
8457 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
8458 ADDB, ADDW and ADDL into a single ADD attribute which selects the
8459 right table based on the size of the operand.
8460 * config/h8300/h8300.md (length_table): Corresponding changes. All
8461 references to "addb", "addw" and "addl" changed to "add".
8462 (btst patterns): Merge two variants into a single pattern.
8463 (tstqi, tsthi): Likewise.
8464 (addhi3_incdec, addsi3_incdec): Likewise.
8465 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
8466 (mulhi3, mulsi3): Likewise.
8467 (udivhi3, udivsi3): Likewise.
8468 (divhi3, divsi3): Likewise.
8469 (andorqi3, andorhi3, andorsi3): Likewise.
8471 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
8474 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
8475 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
8477 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8479 PR tree-optimization/85694
8480 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
8481 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
8483 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
8484 (<u>avg<mode>3_ceil): New patterns.
8486 2018-07-03 David Malcolm <dmalcolm@redhat.com>
8488 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
8489 scan-tree-dump directive.
8490 * gcc.dg/vect/slp-perm-2.c: Likewise.
8491 * gcc.dg/vect/slp-perm-3.c: Likewise.
8492 * gcc.dg/vect/slp-perm-5.c: Likewise.
8493 * gcc.dg/vect/slp-perm-6.c: Likewise.
8494 * gcc.dg/vect/slp-perm-7.c: Likewise.
8495 * gcc.dg/vect/slp-perm-8.c: Likewise.
8497 2018-07-03 Marek Polacek <polacek@redhat.com>
8500 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
8502 2018-07-03 Richard Biener <rguenther@suse.de>
8505 * tree-ssa-structalias.c (find_func_clobbers): Properly
8506 handle indirect calls.
8508 2018-07-03 Jeff Law <law@redhat.com>
8510 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
8511 (shifts): New code iterator.
8512 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
8513 expander. Fix HImode handling on H8/SX.
8514 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
8515 (subqi3, subhi3, subsi3 expanders): Likewise.
8516 (andqi3, andhi3, andsi3 expanders): Likewise.
8517 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
8518 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
8519 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
8520 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
8521 (zero_extendqihi2, zero_extendqisi2): Likewise.
8522 (extendqihi2, extendqisi2): Likewise.
8523 (rotlqi3, rotlhi3, rotlsi3): Likewise.
8524 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
8525 (rotlqi3_1, rotlhi3_1): Likewise.
8526 (logicalhi3_sn, logicalsi3_sn): Likewise.
8527 (logicalhi3, logicalsi3): Likewise.
8529 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8531 * tree-vect-patterns.c (vect_recog_rotate_pattern)
8532 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8533 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
8534 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
8535 type to append_pattern_def_seq instead of creating a stmt_vec_info
8537 (build_mask_conversion): Likewise. Remove vinfo argument.
8538 (vect_add_conversion_to_patterm): Likewise, renaming to...
8539 (vect_add_conversion_to_pattern): ...this.
8540 (vect_recog_mask_conversion_pattern): Update call to
8541 build_mask_conversion. Pass the vector type to
8542 append_pattern_def_seq here too.
8543 (vect_recog_gather_scatter_pattern): Update call to
8544 vect_add_conversion_to_pattern.
8546 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8548 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
8549 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8550 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
8551 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
8552 STMT_VINFO_PATTERN_DEF_SEQ to null here.
8553 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
8554 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
8555 append_pattern_def_seq instead of new_pattern_def_seq.
8556 (vect_recog_divmod_pattern): Do both of the above.
8557 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
8560 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8562 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
8563 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
8564 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
8565 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
8566 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
8567 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8568 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
8569 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8570 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8571 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
8572 parameter with a single stmt_vec_info.
8573 (vect_recog_func_ptr): Likewise.
8574 (vect_recog_gather_scatter_pattern): Likewise, folding in...
8575 (vect_try_gather_scatter_pattern): ...this.
8576 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
8577 the stmt_vec_info of the statement to be matched. Don't clear
8578 STMT_VINFO_RELATED_STMT.
8579 (vect_pattern_recog): Update call accordingly.
8581 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8583 PR tree-optimization/85694
8584 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
8585 (uavgM3_ceil): Document new optabs.
8586 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
8587 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
8589 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
8590 (savg_ceil_optab): New optabs.
8591 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
8592 (vect_vect_recog_func_ptrs): Add it.
8593 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
8594 constant directly from the associated lhs.
8596 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8598 * tree-vect-patterns.c (vect_split_statement): New function.
8599 (vect_convert_input): Use it to try to split an existing cast.
8601 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8603 * poly-int.h (print_hex): New function.
8604 * dumpfile.h (dump_dec, dump_hex): Declare.
8605 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
8606 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
8607 min_input_precision, operation_precision and operation_sign.
8608 * tree-vect-patterns.c (vect_get_range_info): New function.
8609 (vect_same_loop_or_bb_p, vect_single_imm_use)
8610 (vect_operation_fits_smaller_type): Delete.
8611 (vect_look_through_possible_promotion): Add an optional
8612 single_use_p parameter.
8613 (vect_recog_over_widening_pattern): Rewrite to use new
8614 stmt_vec_info infomration. Handle one operation at a time.
8615 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
8616 (vect_truncatable_operation_p, vect_set_operation_type)
8617 (vect_set_min_input_precision): New functions.
8618 (vect_determine_min_output_precision_1): Likewise.
8619 (vect_determine_min_output_precision): Likewise.
8620 (vect_determine_precisions_from_range): Likewise.
8621 (vect_determine_precisions_from_users): Likewise.
8622 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
8623 (vect_vect_recog_func_ptrs): Put over_widening first.
8625 (vect_pattern_recog): Call vect_determine_precisions.
8627 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8629 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
8630 statements that have been replaced by further pattern statements.
8631 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
8633 2018-07-03 Richard Biener <rguenther@suse.de>
8635 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
8636 always set *dt. Dump vectype in vectype overload.
8637 * dumpfile.h (dump_gimple_expr): New function.
8638 (dump_gimple_expr_loc): Likewise.
8639 * dumpfile.c (dump_gimple_expr): New function.
8640 (dump_gimple_expr_loc): Likewise.
8642 2018-07-02 Jeff Law <law@redhat.com>
8644 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
8645 the H8/300, H8/300H and H8/S variants into a single pattern.
8646 (movhi_h8300, movqi_h8300hs): Similarly.
8647 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
8648 (QHI mode iterator): New.
8650 * config/h8300/h8300.md: Remove trailing whitespace.
8652 2018-07-02 Jim Wilson <jimw@sifive.com>
8654 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
8655 instead of emit_insn for interrupt returns.
8656 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
8657 (riscv_sret, riscv_uret): Likewise.
8659 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8661 * pretty-print.c (selftest::test_pp_format): Move save and restore
8662 of quotes to class auto_fix_quotes, and add an instance.
8663 * selftest.c: Include "intl.h".
8664 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
8665 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
8666 * selftest.h (selftest::auto_fix_quotes): New class.
8668 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
8670 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
8671 (aarch64_sve_prepare_conditional_op): Remove.
8672 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
8673 Allow aarch64_simd_reg_or_zero as select operand; remove
8674 the aarch64_sve_prepare_conditional_op call.
8675 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
8676 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8677 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
8678 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
8679 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
8680 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
8681 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
8682 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
8683 and a splitters to match all of the *_any patterns.
8684 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
8686 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
8687 (SVE_COND_FP_BINARY_REV): Remove.
8688 (sve_int_op_rev, sve_fp_op_rev): New.
8689 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
8690 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
8691 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
8692 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
8693 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
8694 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
8695 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
8696 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
8697 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
8699 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
8700 Remove match_dup 1 from the inner unspec.
8701 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8703 * config/aarch64/aarch64.md (movprfx): New attr.
8704 (length): Default movprfx to 8.
8705 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
8706 (*madd<SVE_I>, *msub<SVE_I): Likewise.
8707 (*<su>mul<SVE_I>3_highpart): Likewise.
8708 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
8709 (*v<ASHIFT><SVE_I>3): Likewise.
8710 (*<su><MAXMIN><SVE_I>3): Likewise.
8711 (*<su><MAXMIN><SVE_F>3): Likewise.
8712 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
8713 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
8714 (*div<SVE_F>4): Likewise.
8716 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
8718 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
8721 2018-07-02 Richard Biener <rguenther@suse.de>
8723 PR tree-optimization/86363
8724 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
8725 memset argument refers to a non-variable address.
8727 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8729 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
8730 {PLUS,MINUS}_EXPR code to...
8731 (adjust_symbolic_bound): ...here,
8732 (combine_bound): ...here,
8733 (set_value_range_with_overflow): ...and here.
8735 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8737 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
8739 (extract_range_from_abs_expr): ...here.
8741 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
8743 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
8744 -fno-omit-frame-pointer when not optimizing.
8746 2018-07-02 Martin Liska <mliska@suse.cz>
8749 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
8750 (propagate_nothrow): Likewise.
8752 2018-07-02 Martin Liska <mliska@suse.cz>
8755 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
8757 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8759 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
8760 function in r262149, changing "loc" param from source_location to
8761 const dump_location_t &.
8762 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
8763 declaration, as above.
8765 2018-07-01 Paul Koning <ni1d@arrl.net>
8767 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8768 -munit-asm, -mgnu-asm, -mdec-asm.
8769 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
8770 (pdp11_output_labelref): New.
8771 (pdp11_output_def): New.
8772 (pdp11_output_addr_vec_elt): New.
8773 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
8774 %# and %@ format codes.
8775 (pdp11_option_override): New.
8776 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
8777 (pdp11_output_ident): New.
8778 (pdp11_asm_named_section): New.
8779 (pdp11_asm_init_sections): New.
8780 (pdp11_file_start): New.
8781 (pdp11_file_end): New.
8782 (output_ascii): Use .ascii/.asciz for -mdec-asm.
8783 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
8784 %o, like %c but octal.
8785 (pdp11_option_override): New.
8786 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
8788 (DATA_SECTION_ASM_OP): Ditto.
8789 (READONLY_DATA_SECTION_ASM_OP): New.
8790 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
8791 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
8792 (ASM_OUTPUT_LABELREF): Ditto.
8793 (ASM_OUTPUT_DEF): Ditto.
8794 (ASM_OUTPUT_EXTERNAL): New.
8795 (ASM_OUTPUT_SOURCE_FILENAME): New.
8796 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
8797 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
8798 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
8799 %# and %@ format codes.
8800 * config/pdp11/pdp11.opt (mgnu-asm): New.
8801 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
8802 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
8803 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
8805 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
8807 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
8808 dereferencing path[] beyond its length.
8810 (debug_all_paths): New.
8811 (rewire_first_differing_edge): New.
8812 (adjust_paths_after_duplication): New.
8813 (duplicate_thread_path): Call adjust_paths_after_duplication.
8815 (thread_through_all_blocks): Add new argument to
8816 duplicate_thread_path.
8818 2018-06-30 Jim Wilson <jimw@sifive.com>
8820 * config/riscv/predicates.md (p2m1_shift_operand): New.
8821 (high_mask_shift_operand): New.
8822 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
8823 pattern using p2m1_shift_operand.
8824 (lshsi3_zero_extend_3+2): New combiner pattern using
8825 high_mask_shift_operand.
8827 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8829 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
8831 (vect_recog_rotate_pattern): ...here.
8832 (vect_convert_input): Try to insert casts of invariants in the
8834 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
8835 preheader to be empty.
8837 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8839 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
8840 vector type. If given, install it in the new statement's
8842 (vect_element_precision): New function.
8843 (vect_unpromoted_value): New struct.
8844 (vect_unpromoted_value::vect_unpromoted_value): New function.
8845 (vect_unpromoted_value::set_op): Likewise.
8846 (vect_look_through_possible_promotion): Likewise.
8847 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
8848 (vect_widened_op_tree, vect_convert_input): Likewise.
8849 (vect_convert_inputs, vect_convert_output): Likewise.
8850 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
8851 to handle the optional cast of the multiplication result and
8852 vect_widened_op_tree to detect the widened multiplication itself.
8853 Do not require the input and output of promotion casts to have
8854 the same sign, but base the signedness of the operation on the
8855 input rather than the result. If the pattern includes two
8856 promotions, check that those promotions have the same sign.
8857 Do not restrict the MULT_EXPR handling to a double-width result;
8858 handle quadruple-width results and wider. Use vect_convert_inputs
8859 to convert the inputs to the common type.
8860 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
8861 to handle the optional cast of the ABS result. Also allow a sign
8862 change or a sign extension between the ABS and MINUS.
8863 Use vect_widened_op_tree to detect the widened subtraction and use
8864 vect_convert_inputs to convert the inputs to the common type.
8865 (vect_handle_widen_op_by_const): Delete.
8866 (vect_recog_widen_op_pattern): New function.
8867 (vect_recog_widen_mult_pattern): Use it.
8868 (vect_recog_widen_shift_pattern): Likewise.
8869 (vect_recog_widen_sum_pattern): Use
8870 vect_look_through_possible_promotion to handle the promoted
8873 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8875 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
8876 the containing gimple_seq *.
8877 * gimple-iterator.h (gsi_for_stmt): Declare it.
8878 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
8879 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
8880 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8881 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8882 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
8884 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
8886 (vect_mark_pattern_stmts): ...here. Handle cases in which the
8887 statement being replaced is part of an existing pattern
8888 definition sequence, inserting the new pattern statements before
8890 (vect_pattern_recog_1): Don't return a bool. If the statement
8891 is already part of a pattern, instead apply pattern matching
8892 to the pattern definition statements. Don't clear the
8893 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
8894 (vect_pattern_recog): Don't break after the first match;
8895 continue processing the pattern definition statements instead.
8896 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
8898 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8900 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8901 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8902 (vect_recog_widen_sum_pattern): Use it.
8904 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8906 * tree-vect-loop.c (vectorizable_reduction): Assert that the
8907 phi is not a pattern statement and has not been replaced by
8908 a pattern statement.
8909 * tree-vect-patterns.c (type_conversion_p): Don't check
8910 STMT_VINFO_IN_PATTERN_P.
8911 (vect_recog_vector_vector_shift_pattern): Likewise.
8912 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
8913 the pattern statement rather than the original statement; check
8914 directly for a WIDEN_MULT_EXPR here.
8915 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
8916 vect_is_simple_use to return the pattern statement rather
8917 than the original statement; use is_pattern_stmt_p to check
8918 for such a pattern statement.
8919 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
8920 to return the pattern statement rather than the original statement;
8921 don't do the same transformation here.
8922 (vect_is_simple_use): If the defining statement has been replaced
8923 by a pattern statement, return the pattern statement instead.
8924 Remove the corresponding (local) transformation from the vectype
8927 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8929 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
8930 end and default to null.
8931 * tree-vect-loop.c (vect_create_epilog_for_reduction)
8932 (vectorizable_reduction): Update calls accordingly, dropping the
8933 gimple ** argument if the passed-back statement isn't needed.
8934 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
8935 (vect_recog_rotate_pattern): Likewise.
8936 (vect_recog_mask_conversion_pattern): Likewise.
8937 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8938 (vect_mask_constant_operand_p): Likewise.
8939 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
8940 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
8941 (get_group_load_store_type, get_load_store_type): Likewise.
8942 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
8943 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
8944 (vectorizable_conversion, vectorizable_assignment): Likewise.
8945 (vectorizable_shift, vectorizable_operation): Likewise.
8946 (vectorizable_store, vect_is_simple_cond): Likewise.
8947 (vectorizable_condition, vectorizable_comparison): Likewise.
8948 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
8949 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
8950 and move it to the end. Cope with null def_stmt_outs.
8952 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8954 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
8956 2018-06-29 Jeff Law <law@redhat.com>
8958 * config/v850/v850.c (v850_legitimate_address_p): Handle large
8959 displacements for TARGET_V850E2V3 and newer.
8960 (TARGET_LRA_P): Remove. Defaults to LRA now.
8961 * config/v850/v850.md (sign23byte_load): Remove.
8962 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
8963 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
8965 2018-06-29 Martin Liska <mliska@suse.cz>
8968 * coverage.c (coverage_init): Mangle full path name.
8969 * doc/invoke.texi: Document the change.
8970 * gcov-io.c (mangle_path): New.
8971 * gcov-io.h (mangle_path): Likewise.
8972 * gcov.c (mangle_name): Use mangle_path for path mangling.
8974 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8976 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
8977 if starting source register is not even.
8979 2018-06-29 Martin Liska <mliska@suse.cz>
8981 PR tree-optimization/86263
8982 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
8983 Make edge redirection.
8985 2018-06-29 David Malcolm <dmalcolm@redhat.com>
8987 * dumpfile.c (dump_loc): Add indentation based on scope depth.
8988 (dump_scope_depth): New variable.
8989 (get_dump_scope_depth): New function.
8990 (dump_begin_scope): New function.
8991 (dump_end_scope): New function.
8992 * dumpfile.h (get_dump_scope_depth): New declaration.
8993 (dump_begin_scope): New declaration.
8994 (dump_end_scope): New declaration.
8995 (class auto_dump_scope): New class.
8996 (AUTO_DUMP_SCOPE): New macro.
8997 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9000 2018-06-29 Richard Biener <rguenther@suse.de>
9002 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9003 compute_all_dependences succeeds.
9004 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9005 exceed --param loop-max-datarefs-for-datadeps.
9007 2018-06-29 Jakub Jelinek <jakub@redhat.com>
9009 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9011 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
9014 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9015 alternative 0 in preferred_for_speed attribute.
9017 2018-06-28 Paul Koning <ni1d@arrl.net>
9019 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9020 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9021 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9022 * config/pdp11/pdp11.md: Correct "length" attribute calculation
9023 for shift insn patterns.
9025 2018-06-28 David Malcolm <dmalcolm@redhat.com>
9027 * cgraph.c (cgraph_node::get_body): Replace assignments to
9028 "dump_file" with calls to set_dump_file.
9029 * dumpfile.c (alt_dump_file): Make static, and group with...
9030 (alt_flags): ...this definition.
9031 (dumps_are_enabled): New variable.
9032 (refresh_dumps_are_enabled): New function.
9033 (set_dump_file): New function.
9034 (set_alt_dump_file): New function.
9035 (gcc::dump_manager::dump_start): Replace assignments to
9036 "dump_file" and "alt_dump_file" with calls to set_dump_file and
9038 (gcc::dump_manager::dump_finish): Likewise.
9039 * dumpfile.h (alt_dump_file): Delete decl.
9040 (dumps_are_enabled): New variable decl.
9041 (set_dump_file): New function decl.
9042 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9044 * tree-nested.c (lower_nested_functions): Replace assignments to
9045 "dump_file" with calls to set_dump_file.
9047 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9049 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9050 goto_locus of each outgoing edge of each basic block.
9052 2018-06-28 Richard Biener <rguenther@suse.de>
9054 * dwarf2out.c (decl_scope_table): Remove.
9055 (push_decl_scope): Likewise.
9056 (pop_decl_scope): Likewise.
9057 (gen_type_die_for_member): Do not call push/pop_decl_scope.
9058 (gen_struct_or_union_type_die): Likewise.
9059 (gen_tagged_type_die): Likewise.
9060 (dwarf2out_init): Do not initialize decl_scope_table.
9061 (dwarf2out_c_finalize): Do not free it.
9063 2018-06-28 Richard Biener <rguenther@suse.de>
9065 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9066 deciding whether to not re-use a DIE.
9068 2018-06-28 Richard Biener <rguenther@suse.de>
9070 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9071 DW_AT_abstract_origin attribute.
9073 2018-06-28 Martin Liska <mliska@suse.cz>
9075 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9076 Use newly introduced constants.
9077 * tree-switch-conversion.h (struct jump_table_cluster):
9078 Define max_ratio_for_size and max_ratio_for_speed.
9080 2018-06-28 Martin Liska <mliska@suse.cz>
9082 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9083 Add new checking assert to catch invalid state.
9084 (jump_table_cluster::can_be_handled): Handle single case
9086 (jump_table_cluster::is_beneficial): Bail out for such case.
9087 (bit_test_cluster::find_bit_tests):
9088 Add new checking assert to catch invalid state.
9089 (bit_test_cluster::can_be_handled): Handle single case
9091 (bit_test_cluster::is_beneficial): Bail out for such case.
9092 (switch_decision_tree::analyze_switch_statement):
9095 2018-06-28 Martin Liska <mliska@suse.cz>
9097 * common.opt: Introduce -completion option.
9098 * gcc.c (driver_handle_option): Handle it.
9099 (driver::main): Print completions if completion
9101 * opt-suggestions.c (option_proposer::get_completions):
9103 (option_proposer::suggest_completion): Likewise.
9104 (option_proposer::find_param_completions): Likewise.
9105 (verify_autocompletions): Likewise.
9106 (test_completion_valid_options): Likewise.
9107 (test_completion_valid_params): Likewise.
9108 (in_completion_p): Likewise.
9109 (empty_completion_p): Likewise.
9110 (test_completion_partial_match): Likewise.
9111 (test_completion_garbage): Likewise.
9112 (opt_proposer_c_tests): Likewise.
9113 * opt-suggestions.h: Declare new functions.
9114 * opts.c (common_handle_option): Handle OPT__completion_.
9115 * selftest-run-tests.c (selftest::run_tests): Add
9116 opt_proposer_c_tests.
9117 * selftest.c (assert_str_startswith): New.
9118 * selftest.h (assert_str_startswith): Likewise.
9119 (opt_proposer_c_tests): New.
9120 (ASSERT_STR_STARTSWITH): Likewise.
9122 2018-06-28 Martin Liska <mliska@suse.cz>
9124 * Makefile.in: Add opt-suggestions.o.
9125 * gcc-main.c: Include opt-suggestions.h.
9126 * gcc.c (driver::driver): Likewise.
9127 (driver::~driver): Remove m_option_suggestions.
9128 (driver::build_option_suggestions): Moved to option_proposer.
9129 (driver::suggest_option): Likewise.
9130 (driver::handle_unrecognized_options): Use option_proposer.
9131 * gcc.h (class driver): Add new memver m_option_proposer.
9132 * opt-suggestions.c: New file.
9133 * opt-suggestions.h: New file.
9135 2018-06-28 Martin Liska <mliska@suse.cz>
9137 * vec.h (class auto_string_vec): New (moved from auto_argvec).
9138 (auto_string_vec::~auto_string_vec): Likewise.
9140 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9142 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9143 prevent_decl_creation_for_types fields up and add reset_location field.
9144 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9145 statement if id->reset_location is true.
9146 (copy_edges_for_bb): Do not set goto_locus on the new edges if
9147 id->reset_location is true.
9148 (copy_phis_for_bb): Force input_location on the arguments if
9149 id->reset_location is true.
9150 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9151 is set on the function to be inlined.
9153 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
9155 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9157 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
9159 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9160 registers for Pmode.
9161 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9162 hard registers for the clobbered pseudo.
9164 2018-06-27 Paul Koning <ni1d@arrl.net>
9166 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9167 mutually exclusive options.
9168 * config/pdp11/constraints.md (h): New constraint.
9169 (O): Update definition to match shift code generation.
9170 (D): New constraint.
9171 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9173 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9175 (output_jump): Change arguments.
9176 (pdp11_fixed_cc_regs): New function.
9177 (pdp11_cc_mode): Ditto.
9178 (pdp11_expand_shift): Ditto.
9179 (pdp11_assemble_shift): Ditto.
9180 (pdp11_small_shift): Ditto.
9181 (pdp11_branch_cost): Remove.
9182 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9184 (pdp11_register_move_cost): Update for CC registers.
9185 (pdp11_rtx_costs): Add case for LSHIFTRT.
9186 (pdp11_output_jump): Add CCNZ mode conditional branches.
9187 (notice_update_cc_on_set): Remove.
9188 (pdp11_cc_mode): New function.
9189 (simple_memory_operand): Correct pre/post decrement case.
9190 (no_side_effect_operand): New function.
9191 (pdp11_regno_reg_class): Add CC_REGS class.
9192 (pdp11_fixed_cc_regs): New function.
9193 (pdp11_small_shift): New function.
9194 (pdp11_expand_shift): New function to expand shift insns.
9195 (pdp11_assemble_shift): New function to output shifts.
9196 (pdp11_branch_cost): Remove.
9197 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9198 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9199 (WCHAR_TYPE): Ditto.
9200 (PTRDIFF_TYPE): Ditto.
9201 (ADJUST_INSN_LENGTH): New macro.
9202 (FIXED_REGISTERS): Add CC registers.
9203 (CALL_USED_REGISTERS): Ditto.
9205 (REG_CLASS_NAMES): Ditto.
9206 (REG_CLASS_CONTENTS): Ditto.
9207 (SELECT_CC_MODE): Use new function.
9208 (TARGET_FLAGS_REGNUM): New macro.
9209 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9210 (cc0_reg_rtx): Remove.
9211 (CC_STATUS_MDEP): Remove.
9212 (CC_STATUS_MDEFP_INIT): Remove.
9213 (CC_IN_FPU): Remove.
9214 (NOTICE_UPDATE_CC): Remove.
9215 (REGISTER_NAMES): Add CC registers.
9216 (BRANCH_COST): Change to constant 1.
9217 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9219 * config/pdp11/pdp11.opt (mbcopy): Remove.
9220 (mbcopy-builtin): Remove.
9221 (mbranch-cheap): Remove.
9222 (mbranch-expensive): Remove.
9223 * config/pdp11/predicates.md (expand_shift_operand): Update to
9224 match shift code generation.
9225 (ccnz_operator): New predicate.
9226 * doc/invoke.texi (PDP-11 Options): Remove deleted options
9227 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9228 Remove non-existent option -mabshi, -mno-abshi. Document mutually
9230 * doc/md.texi (PDP-11): Document new D and h constraints. Update
9231 description of O constraint.
9233 2018-06-27 Jeff Law <law@redhat.com>
9234 Austin Law <austinklaw@gmail.com>
9236 * config/v850/v850.md (addsi3_set_flags): New pattern.
9237 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9238 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9239 (zero_extendhisi2_v850_set_flags): Likewise.
9240 (zero_extendqisi2_v850_set_flags): Likewise.
9241 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9242 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9243 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9245 * config/v850/v850-protos.h (notice_update_cc): Remove.
9246 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9247 (v850_print_operand): Handle 'D' and "d".
9248 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9249 Add handling of arithmetic/logical operations compared against zero.
9250 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9251 Do not look at v850_compare_op, instead get mode from last argument.
9252 (v850_gen_compare): Remove
9253 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9254 after reload for prologue insns.
9255 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9257 (construct_save_jarl): Likewise.
9258 (TARGET_FLAGS_REGNUM): Define.
9259 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9260 (NOTICE_UPDATE_CC): Remove.
9261 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9262 than cc0. Conditionalize on reload_completed.
9263 (cmpsi_insn, setfcc_insn): Likewise.
9264 (tst1 splitter): Turn into define_and_split which sets the flags
9266 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9267 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9268 (cstoresf4, cstoredf4): Clobber the flags.
9269 (cmpsi, cmpsf, cmpdf): Remove expanders.
9270 (setf_insn): Remove pattern.
9271 (addsi3): Turn into define_and_split which clobbers the flags after
9272 reload and a suitable pattern (addsi3_clobber_flags) for use after
9274 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9275 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9276 (ashrsi3, ashrsi3_v850e2): Likewise.
9277 (bins): Clobber the flags.
9278 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9279 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9280 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9281 (call_value_internal_short, call_value_internal_long): Likewise.
9282 (callt_save_interrupt, callt_return_interrupt): Likewise.
9283 (save_interrupt, return_interrupt): Likewise.
9284 (callt_save_all_interrupt, save_all_interrupt): Likewise.
9285 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9286 (restore_all_interrupt, _restore_all_interrupt): Likewise.
9287 (All FP comparisons): Only allow after reload has completed.
9289 (divh, divhu): Tweak output template.
9290 (branch_z_normal, branch_z_invert): Remove
9291 (branch_nz_normal, branch_nz_invert): Likewise.
9292 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9294 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9295 * config/v850/v850.c (notice_update_cc): Remove.
9296 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
9297 (CC_NO_CARRY): Likewise.
9298 (NOTICE_UPDATE_CC): Define to nothing.
9299 * config/v850/v850.md: Remove block comment on cc0 handling
9300 Remove "cc" attribute from all patterns. Remove cc_status handling
9301 from all patterns. Minor formatting fixes.
9303 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9305 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9306 (cortex-a76.cortex-a55): Likewise.
9307 * config/aarch64/aarch64-tune.md: Regenerate.
9308 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9309 cortex-a76.cortex-a55.
9311 2018-06-27 Jeff Law <law@redhat.com>
9313 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9314 (MULTILIB_DIRNAMES): Similarly.
9316 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
9318 * gimple.h (gimple_return_retbnd): Delete.
9319 (gimple_return_set_retbnd): Likewise.
9320 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9321 gimple_return_set_retbnd.
9322 * gimple-pretty-print.c (dump_gimple_return): Remove call to
9323 gimple_return_retbnd and adjust.
9324 * tree-inline.h (struct copy_body_data): Remove retbnd field.
9325 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9326 Explicitly return NULL in a couple more cases. Move assertion
9327 on debug statements and remove unreachable code.
9328 (reset_debug_binding): Do not test id->retbnd.
9329 (expand_call_inline): Do not set it.
9331 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9333 * configure.ac: Add --disable-gcov option.
9334 * configure: Regenerate.
9335 * Makefile.in: Honour @enable_gcov@.
9336 * doc/install.texi: Document --disable-gcov.
9338 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9340 * config/arm/arm-cpus.in (cortex-a76): New entry.
9341 (cortex-a76.cortex-a55): Likewise.
9342 * config/arm/arm-tables.opt: Regenerate.
9343 * config/arm/arm-tune.md: Likewise.
9344 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
9345 * doc/invoke.texi (ARM Options): Document cortex-a76 and
9346 cortex-a76.cortex-a55.
9348 2018-06-27 Tamar Christina <tamar.christina@arm.com>
9351 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
9353 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
9355 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
9357 (EPILOGUE_USES): Likewise.
9359 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
9361 * tree-inline.c (remap_location): New function extracted from...
9362 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
9363 (copy_phis_for_bb): ...here. Call remap_location.
9364 (copy_cfg_body): Adjust call to copy_edges_for_bb.
9366 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
9368 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
9369 unaligned vsx for 16B memset.
9371 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
9374 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
9375 ieee128_float_type_node to long_double_type_node unless
9376 TARGET_LONG_DOUBLE_128 is set.
9378 2018-06-26 David Malcolm <dmalcolm@redhat.com>
9380 * cfgloop.c (get_loop_location): Convert return type from
9381 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
9382 by implicit construction from rtx_insn *, and using
9383 dump_user_location_t::from_function_decl for the fallback case.
9384 * cfgloop.h (get_loop_location): Convert return type from
9385 location_t to dump_user_location_t.
9386 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
9387 dump_printf_loc to pass in a dump_location_t rather than a
9388 location_t, via the gimple stmt.
9389 * coverage.c (get_coverage_counts): Update calls to
9390 dump_printf_loc to pass in dump_location_t rather than a
9392 * doc/optinfo.texi (Dump types): Convert example of
9393 dump_printf_loc from taking "locus" to taking "insn". Update
9394 description of the "_loc" calls to cover dump_location_t.
9395 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
9397 (dump_user_location_t::dump_user_location_t): New constructors,
9398 from gimple *stmt and rtx_insn *.
9399 (dump_user_location_t::from_function_decl): New function.
9400 (dump_loc): Make static.
9401 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
9402 const dump_location_t &.
9403 (dump_generic_expr_loc): Delete.
9404 (dump_printf_loc): Convert param "loc" from location_t to
9405 const dump_location_t &.
9406 (selftest::test_impl_location): New function.
9407 (selftest::dumpfile_c_tests): New function.
9408 * dumpfile.h: Include "profile-count.h".
9409 (class dump_user_location_t): New class.
9410 (struct dump_impl_location_t): New struct.
9411 (class dump_location_t): New class.
9412 (dump_printf_loc): Convert 2nd param from source_location to
9413 const dump_location_t &.
9414 (dump_generic_expr_loc): Delete.
9415 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
9416 const dump_location_t &.
9417 * gimple-fold.c (fold_gimple_assign): Update call to
9418 dump_printf_loc to pass in a dump_location_t rather than a
9419 location_t, via the gimple stmt.
9420 (gimple_fold_call): Likewise.
9421 * gimple-loop-interchange.cc
9422 (loop_cand::analyze_iloop_reduction_var): Update for change to
9423 check_reduction_path.
9424 (tree_loop_interchange::interchange): Update for change to
9426 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
9427 change in return-type of find_loop_location.
9428 (graphite_regenerate_ast_isl): Likewise.
9429 * graphite-optimize-isl.c (optimize_isl): Likewise.
9430 * graphite.c (graphite_transform_loops): Likewise.
9431 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
9432 pass in a dump_location_t rather than a location_t, via the
9434 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9435 * ipa.c (walk_polymorphic_call_targets): Likewise.
9436 * loop-unroll.c (report_unroll): Convert "locus" param from
9437 location_t to dump_location_t.
9438 (decide_unrolling): Update for change to get_loop_location's
9440 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
9441 location_t to dump_user_location_t.
9442 (grid_find_single_omp_among_assignments_1): Updates calls to
9443 dump_printf_loc to pass in a dump_location_t rather than a
9444 location_t, via the gimple stmt.
9445 (grid_parallel_clauses_gridifiable): Convert "tloc" from
9446 location_t to dump_location_t. Updates calls to dump_printf_loc
9447 to pass in a dump_location_t rather than a location_t, via the
9449 (grid_inner_loop_gridifiable_p): Likewise.
9450 (grid_dist_follows_simple_pattern): Likewise.
9451 (grid_gfor_follows_tiling_pattern): Likewise.
9452 (grid_target_follows_gridifiable_pattern): Likewise.
9453 (grid_attempt_target_gridification): Convert initialization
9454 of local "grid" from memset to zero-initialization; FIXME: does
9455 this require C++11? Update call to dump_printf_loc to pass in a
9456 optinfo_location rather than a location_t, via the gimple stmt.
9457 * profile.c (read_profile_edge_counts): Updates call to
9458 dump_printf_loc to pass in a dump_location_t rather than a
9460 (compute_branch_probabilities): Likewise.
9461 * selftest-run-tests.c (selftest::run_tests): Call
9463 * selftest.h (dumpfile_c_tests): New decl.
9464 * tree-loop-distribution.c (pass_loop_distribution::execute):
9465 Update for change in return type of find_loop_location.
9466 * tree-parloops.c (parallelize_loops): Likewise.
9467 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
9468 "locus" from location_t to dump_user_location_t.
9469 (canonicalize_loop_induction_variables): Likewise.
9470 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
9471 for change in return type of find_loop_location.
9472 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
9473 to dump_printf_loc to pass in a dump_location_t rather than a
9474 location_t, via the stmt.
9475 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
9477 * tree-vect-loop-manip.c (find_loop_location): Convert return
9478 type from source_location to dump_user_location_t.
9479 (vect_do_peeling): Update for above change.
9480 (vect_loop_versioning): Update for change in type of
9482 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
9483 from location_t to dump_user_location_t.
9484 (vect_estimate_min_profitable_iters): Update for change in type
9486 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
9487 location_t to dump_location_t.
9488 (vect_slp_bb): Update for change in type of vect_location.
9489 * tree-vectorizer.c (vect_location): Convert from source_location
9490 to dump_user_location_t.
9491 (try_vectorize_loop_1): Update for change in vect_location's type.
9492 (vectorize_loops): Likewise.
9493 (increase_alignment): Likewise.
9494 * tree-vectorizer.h (vect_location): Convert from source_location
9495 to dump_user_location_t.
9496 (find_loop_location): Convert return type from source_location to
9497 dump_user_location_t.
9498 (check_reduction_path): Convert 1st param from location_t to
9499 dump_user_location_t.
9500 * value-prof.c (check_counter): Update call to dump_printf_loc to
9501 pass in a dump_user_location_t rather than a location_t; update
9502 call to error_at for change in type of "locus".
9503 (check_ic_target): Update call to dump_printf_loc to
9504 pass in a dump_user_location_t rather than a location_t, via the
9507 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
9509 * config/s390/s390.h (enum processor_flags): Do not use
9510 default tune parameter when -march was specified.
9512 2018-06-26 Jakub Jelinek <jakub@redhat.com>
9515 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
9516 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
9519 2018-06-26 Richard Biener <rguenther@suse.de>
9521 PR tree-optimization/86287
9523 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
9524 (vect_analyze_loop): Initialize n_stmts.
9526 2018-06-26 Richard Biener <rguenther@suse.de>
9529 * fold-const.c (fold_convertible_p): Pointer extension
9532 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
9535 * dwarf2out.c (loc_list_has_views): Adjust comments.
9536 (dw_loc_list): Split single cross-partition range with
9539 2018-06-25 Jeff Law <law@redhat.com>
9541 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
9542 on -mbig-switch by default.
9544 * config/v850/predicates.md (const_float_1_operand): Fix match_code
9546 (const_float_0_operand): Remove unused predicate.
9547 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
9548 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
9549 (recipsf2): New expander. Original pattern now called
9551 (recipdf2, recipdf2_insn): Similarly.
9552 (rsqrtsf2, rsqrtsf2_insn): Similarly
9553 (rsqrtdf2, rsqrtdf2_insn): Similarly
9555 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
9557 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
9558 Simplify logic for FreeBSD (twice).
9560 2018-06-25 Martin Sebor <msebor@redhat.com>
9562 PR tree-optimization/86204
9563 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
9564 a strnlen result if it's less than the length of the string.
9566 2018-06-25 Martin Sebor <msebor@redhat.com>
9568 PR tree-optimization/85700
9569 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
9570 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
9571 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
9573 2018-06-25 Martin Sebor <msebor@redhat.com>
9575 * doc/extend.texi (Zero-length arrays): Update and clarify.
9577 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
9579 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
9580 added IEEE/IBM long double multilib support on PowerPC little
9581 endian Linux systems.
9582 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
9583 (MULTILIB_DEFAULTS): Likewise.
9584 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9586 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
9587 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
9588 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
9590 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
9593 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
9594 (REORDER_45): Likewise.
9596 2018-06-25 Jeff Law <law@redhat.com>
9598 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
9599 dividend to 32 bits. Adjust length.
9600 (udivmodhi4): Cleanup output template. Fix length.
9602 2018-06-25 Carl Love <cel@us.ibm.com>
9604 * config/rs6000/vsx.md: Change word selector to prefered location.
9606 2018-06-25 Richard Biener <rguenther@suse.de>
9608 PR tree-optimization/86304
9609 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
9610 epilogue-if-converted loops as well.
9612 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
9614 * lto-section-out.c (lto_begin_section): Do not print section
9615 name for noaddr and unnumbered dumps.
9617 2018-06-25 Richard Biener <rguenther@suse.de>
9619 * tree-vectorizer.h (struct vec_info_shared): New structure
9620 with parts split out from struct vec_info and loop_nest from
9621 struct _loop_vec_info.
9622 (struct vec_info): Adjust accordingly.
9623 (struct _loop_vec_info): Likewise.
9624 (LOOP_VINFO_LOOP_NEST): Adjust.
9625 (LOOP_VINFO_DATAREFS): Likewise.
9626 (LOOP_VINFO_DDRS): Likewise.
9627 (struct _bb_vec_info): Likewise.
9628 (BB_VINFO_DATAREFS): Likewise.
9629 (BB_VINFO_DDRS): Likewise.
9630 (struct _stmt_vec_info): Add dr_aux member.
9631 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
9632 (DR_MISALIGNMENT_UNINITIALIZED): New.
9633 (set_dr_misalignment): Adjust.
9634 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
9635 (vect_analyze_loop): Adjust prototype.
9636 (vect_analyze_loop_form): Likewise.
9637 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9638 Compute dependences lazily.
9639 (vect_record_base_alignments): Use shared datarefs/ddrs.
9640 (vect_verify_datarefs_alignment): Likewise.
9641 (vect_analyze_data_refs_alignment): Likewise.
9642 (vect_analyze_data_ref_accesses): Likewise.
9643 (vect_analyze_data_refs): Likewise.
9644 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
9645 constructor parameter for shared part.
9646 (vect_analyze_loop_form): Pass in shared part and adjust.
9647 (vect_analyze_loop_2): Pass in storage for the number of
9648 stmts. Move loop nest finding to the caller. Compute
9650 (vect_analyze_loop): Pass in shared part.
9651 (vect_transform_loop): Verify shared datarefs are unchanged.
9652 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
9653 constructor parameter for shared part.
9654 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
9655 (vect_slp_bb): Verify shared datarefs are unchanged before
9657 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
9659 (new_stmt_vec_info): Initialize DR_AUX misalignment to
9660 DR_MISALIGNMENT_UNINITIALIZED.
9661 * tree-vectorizer.c (vec_info::vec_info): Add constructor
9662 parameter for shared part.
9663 (vec_info::~vec_info): Adjust.
9664 (vec_info_shared::vec_info_shared): New.
9665 (vec_info_shared::~vec_info_shared): Likewise.
9666 (vec_info_shared::save_datarefs): Likewise.
9667 (vec_info_shared::check_datarefs): Likewise.
9668 (try_vectorize_loop_1): Construct shared part live for analyses
9669 of a single loop for multiple vector sizes.
9670 * tree-parloops.c (gather_scalar_reductions): Adjust.
9672 2018-06-25 Richard Biener <rguenther@suse.de>
9674 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
9675 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
9676 (vect_analyze_data_refs): Remove similar code from here and
9677 simplify accordingly.
9679 2018-06-25 Richard Biener <rguenther@suse.de>
9681 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
9682 for reverse storage order accesses rather than asserting
9683 they cannot happen here.
9685 2018-06-25 Tom de Vries <tdevries@suse.de>
9688 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
9689 Use data16 instead of .byte for insn prefix.
9691 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
9694 * parser.c (make_char_string_pack): Pass this literal chars
9695 through cpp_interpret_string.
9696 (cp_parser_userdef_numeric_literal): Check the result of
9697 make_char_string_pack.
9699 2018-06-24 Maya Rashish <coypu@sdf.org>
9701 * ginclude/stddef.h: Simplify conditions around avoiding
9702 re-definition of __size_t.
9704 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9706 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
9707 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
9709 2018-06-22 Maya Rashish <coypu@sdf.org>
9711 * doc/invoke.texi (mno-fancy-math-387): Update for changes
9712 made to OpenBSD and NetBSD through the years.
9714 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9716 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9717 behavior of vec_pack (vector double, vector double) to match
9718 behavior of vec_float2 (vector double, vector double).
9720 2018-06-22 Olivier Hainque <hainque@adacore.com>
9722 * gimplify.c (gimplify_function_tree): Prevent creation
9723 of a trampoline for the address of the current function
9724 passed to entry/exit instrumentation hooks.
9726 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
9729 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
9732 2018-06-22 Martin Liska <mliska@suse.cz>
9734 PR tree-optimization/86263
9735 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9736 Bail out if is_enabled is false.
9737 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
9739 (jump_table_cluster::is_enabled): New function.
9741 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9743 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9744 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
9745 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
9746 (lto_input_ts_binfo_tree_pointers): Likewise.
9747 * tree-streamer-out.c (streamer_write_tree_bitfields,
9748 write_ts_binfo_tree_pointers): Likewise.
9749 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
9751 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9753 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
9755 2018-06-22 Martin Liska <mliska@suse.cz>
9757 * symbol-summary.h (get): Make it pure and inline move
9758 functionality from ::get function.
9759 (get): Remove and inline into ::get and ::get_create.
9760 (get_create): Move code from ::get function.
9762 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9765 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
9766 -x assembler-with-cpp.
9768 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9770 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
9771 _FILE_OFFSET_BITS=64 for C++.
9773 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
9775 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
9776 conversion insn that shows up when pr85657-3.c is compiled using
9777 IEEE 128-bit long double.
9778 (neg<mode>2_internal): Use the correct mode to check whether the
9779 mode is IBM extended.
9780 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
9781 multiply and divide external functions from being created more
9784 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9786 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
9788 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
9789 the edge can be forwarded.
9790 (cfg_layout_merge_blocks): Likewise.
9792 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9794 * except.c (finish_eh_generation): Commit edge insertions only after
9795 the EH edges have been redirected from post-landing to landing pads.
9797 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9799 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
9800 create_tmp_var_for to create the FRAME decl.
9801 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
9803 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9805 * tree-inline.c (copy_edges_for_bb): Minor tweak.
9806 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
9807 debug statement when resetting its value.
9808 (expand_call_inline): Copy the locus of the call onto the assignment
9809 of the return value, if any. Use local variable in more cases.
9811 2018-06-21 Martin Liska <mliska@suse.cz>
9813 * ipa-pure-const.c (propagate_nothrow): Use
9814 funct_state_summaries->get.
9815 (dump_malloc_lattice): Likewise.
9816 (propagate_malloc): Likewise.
9818 2018-06-21 Richard Biener <rguenther@suse.de>
9820 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
9821 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
9822 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
9824 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
9825 BLOCK_ABSTRACT_ORIGIN unconditionally.
9827 2018-06-21 David Malcolm <dmalcolm@redhat.com>
9829 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
9831 * ipa-reference.c (ipa_reference_c_finalize): Delete
9832 ipa_ref_opt_sum_summaries and set it to NULL.
9834 2018-06-21 Tom de Vries <tdevries@suse.de>
9836 PR tree-optimization/85859
9837 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
9838 test with comment from bb_no_side_effects_p.
9840 2018-06-21 Richard Biener <rguenther@suse.de>
9842 PR tree-optimization/86232
9843 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
9844 max for constant niter.
9846 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9848 * config/aarch64/aarch64-simd.md
9849 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
9851 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9853 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
9854 Make opernads of the unspec commutative.
9856 2018-06-21 Richard Biener <rguenther@suse.de>
9858 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
9859 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
9860 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
9861 (vect_analyze_data_ref_dependence): Re-order checks to deal with
9863 (vect_record_base_alignments): Do not record base alignment
9864 for gathers or scatters.
9865 (vect_compute_data_ref_alignment): Drop return value that is always
9866 true. Bail out early for gathers or scatters.
9867 (vect_enhance_data_refs_alignment): Bail out early for gathers
9869 (vect_find_same_alignment_drs): Likewise.
9870 (vect_analyze_data_refs_alignment): Remove dead code.
9871 (vect_slp_analyze_and_verify_node_alignment): Likewise.
9872 (vect_analyze_data_refs): For possible gathers or scatters do
9873 not create an alternate DR, just check their possible validity
9874 and mark them. Adjust DECL_NONALIASED handling to not rely
9876 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
9877 update inits of gathers or scatters.
9878 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
9879 Also copy gather/scatter flag to pattern vinfo.
9881 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
9883 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9884 behavior of vec_packsu (vector unsigned long long, vector unsigned
9885 long long) to match behavior of vec_packs with same signature.
9887 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
9888 Thomas Schwinge <thomas@codesourcery.com>
9889 Cesar Philippidis <cesar@codesourcery.com>
9891 * gimplify.c (gimplify_scan_omp_clauses): Add support for
9892 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9893 (gimplify_adjust_omp_clauses): Likewise.
9894 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
9895 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
9896 (gimplify_omp_target_update): Update handling of acc update and
9898 * omp-low.c (install_var_field): Remove unused parameter
9899 base_pointers_restrict.
9900 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
9901 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
9903 (omp_target_base_pointers_restrict_p): Delete.
9904 (scan_omp_target): Update call to scan_sharing_clauses.
9905 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
9907 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
9908 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9909 (convert_local_omp_clauses): Likewise.
9910 * tree-pretty-print.c (dump_omp_clause): Likewise.
9911 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
9913 (omp_clause_code_name): Likewise.
9915 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9918 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
9921 PR tree-optimization/86231
9922 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
9923 anti-range don't overwrite *vr0min before using it to compute *vr0max.
9925 2018-06-20 Tom de Vries <tdevries@suse.de>
9927 PR tree-optimization/86097
9928 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
9929 iv type if signedness of iv type is not the same as that of *nit.
9931 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9933 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
9934 EDGE_EH edges, verify they are all EDGE_EH.
9936 2018-06-20 Maya Rashish <coypu@sdf.org>
9938 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
9940 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9942 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
9943 * config/aarch64/aarch64.c (xgene1_tunings): Add
9944 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
9945 (aarch64_mode_valid_for_sched_fusion_p):
9946 Allow 16-byte modes.
9947 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
9948 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
9950 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
9952 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
9953 * config/aarch64/iterators.md (VQ2): New mode iterator.
9955 2018-06-20 Martin Liska <mliska@suse.cz>
9957 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9958 Change default ratio from 10 to 8.
9960 2018-06-20 Martin Liska <mliska@suse.cz>
9962 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9964 (bit_test_cluster::find_bit_tests): Likewise.
9965 (switch_decision_tree::analyze_switch_statement): Find clusters.
9966 * tree-switch-conversion.h (struct jump_table_cluster): Document
9969 2018-06-20 Martin Liska <mliska@suse.cz>
9971 * tree-switch-conversion.c (switch_conversion::collect):
9972 Record m_uniq property.
9973 (switch_conversion::expand): Bail out for special conditions.
9974 (group_cluster::~group_cluster): New.
9975 (group_cluster::group_cluster): Likewise.
9976 (group_cluster::dump): Likewise.
9977 (jump_table_cluster::emit): New.
9978 (switch_decision_tree::fix_phi_operands_for_edges): New.
9979 (struct case_node): Remove struct.
9980 (jump_table_cluster::can_be_handled): New.
9981 (case_values_threshold): Moved to header.
9982 (reset_out_edges_aux): Likewise.
9983 (jump_table_cluster::is_beneficial): New.
9984 (bit_test_cluster::can_be_handled): Likewise.
9985 (add_case_node): Remove.
9986 (bit_test_cluster::is_beneficial): New.
9987 (case_bit_test::cmp): New.
9988 (bit_test_cluster::emit): New.
9989 (expand_switch_as_decision_tree_p): Remove.
9990 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
9991 (fix_phi_operands_for_edge): Likewise.
9992 (switch_decision_tree::analyze_switch_statement): New.
9993 (compute_cases_per_edge): Move ...
9994 (switch_decision_tree::compute_cases_per_edge): ... here.
9995 (try_switch_expansion): Likewise.
9996 (switch_decision_tree::try_switch_expansion): Likewise.
9997 (record_phi_operand_mapping): Likewise.
9998 (switch_decision_tree::record_phi_operand_mapping): Likewise.
9999 (emit_case_decision_tree): Likewise.
10000 (switch_decision_tree::emit): Likewise.
10001 (balance_case_nodes): Likewise.
10002 (switch_decision_tree::balance_case_nodes): Likewise.
10003 (dump_case_nodes): Likewise.
10004 (switch_decision_tree::dump_case_nodes): Likewise.
10005 (emit_jump): Likewise.
10006 (switch_decision_tree::emit_jump): Likewise.
10007 (emit_cmp_and_jump_insns): Likewise.
10008 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10009 (emit_case_nodes): Likewise.
10010 (switch_decision_tree::emit_case_nodes): Likewise.
10011 (conditional_probability): Remove.
10012 * tree-switch-conversion.h (enum cluster_type): New.
10014 (struct cluster): Likewise.
10015 (cluster::cluster): Likewise.
10016 (struct simple_cluster): Likewise.
10017 (simple_cluster::simple_cluster): Likewise.
10018 (struct group_cluster): Likewise.
10019 (struct jump_table_cluster): Likewise.
10020 (struct bit_test_cluster): Likewise.
10021 (struct min_cluster_item): Likewise.
10022 (struct case_tree_node): Likewise.
10023 (case_tree_node::case_tree_node): Likewise.
10024 (jump_table_cluster::case_values_threshold): Likewise.
10025 (struct case_bit_test): Likewise.
10026 (struct switch_decision_tree): Likewise.
10027 (struct switch_conversion): Likewise.
10028 (switch_decision_tree::reset_out_edges_aux): Likewise.
10030 2018-06-20 Martin Liska <mliska@suse.cz>
10032 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10033 (hoist_edge_and_branch_if_true): Likewise.
10034 (expand_switch_using_bit_tests_p): Likewise.
10035 (struct case_bit_test): Likewise.
10036 (case_bit_test_cmp): Likewise.
10037 (emit_case_bit_tests): Likewise.
10038 (switch_conversion::switch_conversion): New class.
10039 (struct switch_conv_info): Remove old struct.
10040 (collect_switch_conv_info): More to ...
10041 (switch_conversion::collect): ... this.
10042 (check_range): Likewise.
10043 (switch_conversion::check_range): Likewise.
10044 (check_all_empty_except_final): Likewise.
10045 (switch_conversion::check_all_empty_except_final): Likewise.
10046 (check_final_bb): Likewise.
10047 (switch_conversion::check_final_bb): Likewise.
10048 (create_temp_arrays): Likewise.
10049 (switch_conversion::create_temp_arrays): Likewise.
10050 (free_temp_arrays): Likewise.
10051 (gather_default_values): Likewise.
10052 (switch_conversion::gather_default_values): Likewise.
10053 (build_constructors): Likewise.
10054 (switch_conversion::build_constructors): Likewise.
10055 (constructor_contains_same_values_p): Likewise.
10056 (switch_conversion::contains_same_values_p): Likewise.
10057 (array_value_type): Likewise.
10058 (switch_conversion::array_value_type): Likewise.
10059 (build_one_array): Likewise.
10060 (switch_conversion::build_one_array): Likewise.
10061 (build_arrays): Likewise.
10062 (switch_conversion::build_arrays): Likewise.
10063 (gen_def_assigns): Likewise.
10064 (switch_conversion::gen_def_assigns): Likewise.
10065 (prune_bbs): Likewise.
10066 (switch_conversion::prune_bbs): Likewise.
10067 (fix_phi_nodes): Likewise.
10068 (switch_conversion::fix_phi_nodes): Likewise.
10069 (gen_inbound_check): Likewise.
10070 (switch_conversion::gen_inbound_check): Likewise.
10071 (process_switch): Use the newly created class.
10072 (switch_conversion::expand): New.
10073 (switch_conversion::~switch_conversion): New.
10074 * tree-switch-conversion.h: New file.
10076 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10078 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10079 tree-vect-patterns.c.
10080 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10081 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
10082 (vect_recog_sad_pattern): Likewise.
10083 (vect_recog_widen_sum_pattern): Likewise.
10084 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
10085 (vect_recog_widen_shift_pattern): Remove the type_in argument.
10086 (vect_recog_rotate_pattern): Likewise.
10087 (vect_recog_mult_pattern): Likewise.
10088 (vect_recog_vector_vector_shift_pattern): Likewise.
10089 (vect_recog_divmod_pattern): Likewise.
10090 (vect_recog_mixed_size_cond_pattern): Likewise.
10091 (vect_recog_bool_pattern): Likewise.
10092 (vect_recog_mask_conversion_pattern): Likewise.
10093 (vect_try_gather_scatter_pattern): Likewise.
10094 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
10095 (vect_recog_over_widening_pattern): Likewise.
10096 (vect_recog_gather_scatter_pattern): Likewise.
10097 (vect_recog_func_ptr): Move from tree-vectorizer.h
10098 (vect_vect_recog_func_ptrs): Move further down the file.
10099 (vect_recog_func): Likewise. Remove the third argument.
10100 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10101 (vect_pattern_recog_1): Expect the pattern function to do any
10102 necessary target tests. Also expect it to provide a vector type.
10103 Remove the type_in handling.
10105 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10107 * tree-vect-patterns.c (vect_pattern_detected): New function.
10108 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10109 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10110 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10111 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10112 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10113 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10114 (vect_recog_mask_conversion_pattern)
10115 (vect_try_gather_scatter_pattern): Likewise.
10117 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10119 * tree-vect-patterns.c (vect_get_internal_def): New function.
10120 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10121 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10122 (search_type_for_mask_1): Use it.
10124 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10126 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10127 redundant WIDEN_SUM_EXPR handling.
10128 (vect_recog_sad_pattern): Likewise.
10130 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10132 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10133 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10134 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10135 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10136 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10138 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10140 * tree-vect-stmts.c (vectorizable_call): Make sure that we
10141 use the stmt_vec_info of the original bb statement for the
10142 new zero assignment, even if the call is part of a pattern.
10144 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10146 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10147 that the sequence is attached to the original statement rather
10148 than the pattern statement.
10149 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10150 PATTERN_DEF_SEQ from the original statement rather than
10151 the main pattern statement.
10152 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10153 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10154 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10156 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10158 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10159 definition statements before the early exit for statements that aren't
10161 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10163 (vect_transform_loop): ...here. Process pattern definition
10164 statements without first checking whether the main pattern
10165 statement is live or relevant.
10167 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10169 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10170 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10172 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
10174 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10175 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10176 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10177 (expand_block_compare): Change select_block_compare_mode call.
10178 (expand_strncmp_align_check): Use new functions, fix comment.
10179 (emit_final_str_compare_gpr): New function.
10180 (expand_strn_compare): Refactor and clean up code.
10181 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10183 2018-06-19 Tony Reix <tony.reix@atos.com>
10184 Damien Bergamini <damien.bergamini@atos.com>
10185 David Edelsohn <dje.gcc@gmail.com>
10187 * collect2.c (static_obj): New variable.
10188 (static_libs): New variable.
10189 (is_in_list): Uncomment declaration.
10190 (main): Track AIX libraries linked statically.
10191 (is_in_list): Uncomment definition.
10192 (scan_prog_file): Don't add AIX shared libraries initializer
10193 to constructor list if linking statically.
10195 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
10197 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10199 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10201 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
10203 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10206 2018-06-19 Martin Liska <mliska@suse.cz>
10208 * config/i386/i386.c (ix86_can_inline_p): Do not use
10209 ipa_fn_summaries::get_create.
10210 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10212 (devirtualization_time_bonus): Likewise.
10213 (ipcp_propagate_stage): Likewise.
10214 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10215 (edge_set_predicate): Likewise.
10216 (evaluate_conditions_for_known_args): Likewise.
10217 (evaluate_properties_for_edge): Likewise.
10218 (ipa_call_summary::reset): Tranform to ...
10219 (ipa_call_summary::~ipa_call_summary): ... this.
10220 (ipa_fn_summary::reset): Transform to ...
10221 (ipa_fn_summary::~ipa_fn_summary): ... this.
10222 (ipa_fn_summary_t::remove): Rename to ...
10223 (ipa_fn_summary_t::remove_callees): ... this.
10224 (ipa_fn_summary_t::duplicate): Use placement new
10225 instead of memory copy.
10226 (ipa_call_summary_t::duplicate): Likewise.
10227 (ipa_call_summary_t::remove): Remove.
10228 (dump_ipa_call_summary): Change get_create to get.
10229 (ipa_dump_fn_summary): Dump only when summary exists.
10230 (analyze_function_body): Use symbol_summary::get instead
10232 (compute_fn_summary): Likewise.
10233 (estimate_edge_devirt_benefit): Likewise.
10234 (estimate_edge_size_and_time): Likewise.
10235 (inline_update_callee_summaries): Likewise.
10236 (remap_edge_change_prob): Likewise.
10237 (remap_edge_summaries): Likewise.
10238 (ipa_merge_fn_summary_after_inlining): Likewise.
10239 (write_ipa_call_summary): Likewise.
10240 (ipa_fn_summary_write): Likewise.
10241 (ipa_free_fn_summary): Likewise.
10242 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10243 (struct ipa_call_summary): Likewise.
10244 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10246 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10247 (estimate_size_after_inlining): Likewise.
10248 (estimate_growth): Likewise.
10249 (growth_likely_positive): Likewise.
10250 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10251 (inline_call): Likewise.
10252 * ipa-inline.c (caller_growth_limits): Likewise.
10253 (can_inline_edge_p): Likewise.
10254 (can_inline_edge_by_limits_p): Likewise.
10255 (compute_uninlined_call_time): Likewise.
10256 (compute_inlined_call_time): Likewise.
10257 (want_inline_small_function_p): Likewise.
10258 (edge_badness): Likewise.
10259 (update_caller_keys): Likewise.
10260 (update_callee_keys): Likewise.
10261 (inline_small_functions): Likewise.
10262 (inline_to_all_callers_1): Likewise.
10263 (dump_overall_stats): Likewise.
10264 (early_inline_small_functions): Likewise.
10265 (early_inliner): Likewise.
10266 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10267 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10268 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10269 * ipa-split.c (execute_split_functions): Likewise.
10270 * symbol-summary.h: Likewise.
10271 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10273 2018-06-19 Richard Biener <rguenther@suse.de>
10275 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10276 (vectorize_loops): ... here. Fix dbgcnt handling.
10277 (try_vectorize_loop): Wrap try_vectorize_loop_1.
10279 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
10282 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10283 ieee128 argument takes up only one (vector) register, not two (floating
10286 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10288 * gimplify.c (gimplify_init_constructor): Really never clear for an
10289 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10291 2018-06-19 Richard Biener <rguenther@suse.de>
10293 PR tree-optimization/86179
10294 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10295 after failed recognition.
10297 2018-06-18 Martin Sebor <msebor@redhat.com>
10299 PR middle-end/85602
10300 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10301 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10302 Handle integer subtraction.
10303 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10304 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10306 2018-06-18 David Malcolm <dmalcolm@redhat.com>
10308 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10309 param from rtx to rtx_insn *.
10310 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10312 (frv_ifcvt_modify_insn): Likwise.
10313 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10314 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
10315 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10316 as_a <rtx_insn *> cast to local "unprotected_region" once
10317 it's been established that it's not NULL or pc_rtx.
10318 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10319 param "sethi" from rtx to rtx_insn *.
10320 (nds32_group_float_insns): Likewise for param "insn".
10321 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10323 (vax_output_int_subtract): Likewise.
10324 * config/vax/vax.c (vax_output_int_add): Likewise for param
10326 (vax_output_int_subtract): Likewise.
10327 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10328 (emit_pattern_after): Likewise for param "after".
10329 (emit_insn_after): Likewise.
10330 (emit_jump_insn_after): Likewise.
10331 (emit_call_insn_after): Likewise.
10332 (emit_debug_insn_after): Likewise.
10333 (emit_pattern_before): Likewise for param "before".
10334 (emit_insn_before): Likewise.
10335 (emit_jump_insn_before): Likewise.
10336 * final.c (get_insn_template): Likewise for param "insn", removing
10338 * output.h (get_insn_template): Likewise for 2nd param.
10339 * rtl.h (emit_insn_before): Likewise.
10340 (emit_jump_insn_before): Likewise.
10341 (emit_debug_insn_before_noloc): Likewise.
10342 (emit_insn_after): Likewise.
10343 (emit_jump_insn_after): Likewise.
10344 (emit_call_insn_after): Likewise.
10345 (emit_debug_insn_after): Likewise.
10346 (set_insn_deleted): Likewise for param.
10348 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
10351 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
10352 floating point modes, so that IFmode is numerically greater than
10353 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
10354 to declare the ordering. This prevents IFmode from being
10355 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
10356 machine. Include rs6000-modes.h to share the fractional values
10357 between genmodes* and the rest of the compiler.
10358 (IFmode): Likewise.
10359 (KFmode): Likewise.
10360 (TFmode): Likewise.
10361 * config/rs6000/rs6000-modes.h: New file.
10362 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
10363 meaning of rs6000_long_double_size so that 126..128 selects an
10364 appropriate 128-bit floating point type.
10365 (rs6000_option_override_internal): Likewise.
10366 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
10367 (TARGET_LONG_DOUBLE_128): Change the meaning of
10368 rs6000_long_double_size so that 126..128 selects an appropriate
10369 128-bit floating point type.
10370 (LONG_DOUBLE_TYPE_SIZE): Update comment.
10371 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
10372 source and destination to match the standard usage.
10373 (truncifkf2): Likewise.
10374 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
10375 ISA 2.07 to use an explicit clobber, instead of passing in a
10377 (copysign<mode>3_soft): Likewise.
10379 2018-06-18 David Malcolm <dmalcolm@redhat.com>
10381 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10382 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
10383 (vect_slp_analyze_instance_dependence): Likewise.
10384 (vect_enhance_data_refs_alignment): Likewise.
10385 (vect_analyze_data_refs_alignment): Likewise.
10386 (vect_slp_analyze_and_verify_instance_alignment
10387 (vect_analyze_data_ref_accesses): Likewise.
10388 (vect_prune_runtime_alias_test_list): Likewise.
10389 (vect_analyze_data_refs): Likewise.
10390 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
10391 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10392 (vect_analyze_scalar_cycles_1): Likewise.
10393 (vect_get_loop_niters): Likewise.
10394 (vect_analyze_loop_form_1): Likewise.
10395 (vect_update_vf_for_slp): Likewise.
10396 (vect_analyze_loop_operations): Likewise.
10397 (vect_analyze_loop): Likewise.
10398 (vectorizable_induction): Likewise.
10399 (vect_transform_loop): Likewise.
10400 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
10401 * tree-vect-slp.c (vect_analyze_slp): Likewise.
10402 (vect_make_slp_decision): Likewise.
10403 (vect_detect_hybrid_slp): Likewise.
10404 (vect_slp_analyze_operations): Likewise.
10405 (vect_slp_bb): Likewise.
10406 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
10407 (vectorizable_bswap): Likewise.
10408 (vectorizable_call): Likewise.
10409 (vectorizable_simd_clone_call): Likewise.
10410 (vectorizable_conversion): Likewise.
10411 (vectorizable_assignment): Likewise.
10412 (vectorizable_shift): Likewise.
10413 (vectorizable_operation): Likewise.
10414 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
10416 2018-06-18 Martin Sebor <msebor@redhat.com>
10418 PR tree-optimization/81384
10419 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
10420 * builtins.c (expand_builtin_strnlen): New function.
10421 (expand_builtin): Call it.
10422 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
10423 * builtins.def (BUILT_IN_STRNLEN): New.
10424 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
10425 Warn for bounds in excess of maximum object size.
10426 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
10427 single-value ranges. Handle strnlen.
10428 (handle_builtin_strlen): Handle strnlen.
10429 (strlen_check_and_optimize_stmt): Same.
10430 * doc/extend.texi (Other Builtins): Document strnlen.
10432 2018-06-18 Maya Rashish <coypu@sdf.org>
10434 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
10435 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
10436 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
10438 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
10440 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
10442 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10444 * tree.c (escaped_string::escape): Replace cast to char * by
10445 const_cast<char *> (unescaped).
10447 2018-06-18 Nick Clifton <nickc@redhat.com>
10450 * tree.c (escaped_string): New class. Converts an unescaped
10451 string into its escaped equivalent.
10452 (warn_deprecated_use): Use the new class to convert the
10453 deprecation message, if present.
10454 (test_escaped_strings): New self test.
10455 (test_c_tests): Add test_escaped_strings.
10456 * doc/extend.texi (deprecated): Add a note that the
10457 deprecation message is affected by the -fmessage-length
10458 option, and that control characters will be escaped.
10459 (#pragma GCC error): Document this pragma.
10460 (#pragma GCC warning): Likewise.
10461 * doc/invoke.texi (-fmessage-length): Document this option's
10462 effect on the #warning and #error preprocessor directives and
10463 the deprecated attribute.
10465 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
10467 * tree.c (decl_value_expr_lookup): Revert latest change.
10468 (decl_value_expr_insert): Likewise.
10470 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
10472 * gimplify.c (nonlocal_vlas): Delete.
10473 (nonlocal_vla_vars): Likewise.
10474 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
10476 (gimplify_body): Do not create and destroy nonlocal_vlas.
10477 * tree-nested.c: Include diagnostic.h.
10478 (use_pointer_in_frame): Tweak.
10479 (lookup_field_for_decl): Add assertion and declare the transformation.
10480 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
10481 internal error when the reference is in a wrong context. Do not
10482 create a debug decl by default.
10483 (note_nonlocal_block_vlas): Delete.
10484 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
10485 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
10486 create a debug decl by default.
10487 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
10488 call is in a wrong context.
10489 (fixup_vla_decls): New function.
10490 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
10491 debug variables were created.
10492 * tree.c (decl_value_expr_lookup): Add checking assertion.
10493 (decl_value_expr_insert): Likewise.
10495 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
10497 PR middle-end/82479
10498 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
10499 * tree-scalar-evolution.c (interpret_expr): Likewise.
10500 (expression_expensive_p): Likewise.
10501 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
10502 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
10503 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
10504 (ssa_defined_by_minus_one_stmt_p): New.
10506 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
10508 PR middle-end/64946
10509 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
10510 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
10511 * dojump.c (do_jump): Likewise.
10512 * expr.c (expand_expr_real_2): Check operand type's sign.
10513 * fold-const.c (const_unop): Handle ABSU_EXPR.
10514 (fold_abs_const): Likewise.
10515 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
10516 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
10517 (strip_sign_op_1): Likesise.
10518 * match.pd: Add new pattern to generate ABSU_EXPR.
10519 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
10520 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10521 * tree-eh.c (operation_could_trap_helper_p): Likewise.
10522 * tree-inline.c (estimate_operator_cost): Likewise.
10523 * tree-pretty-print.c (dump_generic_node): Likewise.
10524 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
10525 * tree.def (ABSU_EXPR): New.
10527 2018-06-16 Jakub Jelinek <jakub@redhat.com>
10529 PR middle-end/86095
10530 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
10531 documented as preserved for backward compatibility only.
10532 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
10534 PR rtl-optimization/86108
10535 * bb-reorder.c (create_forwarder_block): Renamed to ...
10536 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
10537 jump from new landing pad to the second part.
10538 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
10541 2018-06-15 Jakub Jelinek <jakub@redhat.com>
10543 PR middle-end/85878
10544 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
10545 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
10546 Only call store_expr for halves if the mode is the same.
10548 PR middle-end/86123
10549 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
10550 Fix up comment formatting.
10552 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
10554 * typed-splay-tree.h (typed_splay_tree::remove): New function.
10555 (typed_splay_tree::closure,
10556 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
10557 (typed_splay_tree::typed_splay_tree,
10558 typed_splay_tree::operator =): Declared private.
10559 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
10560 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
10561 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
10562 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
10563 typed_splay_tree::splay_tree_splay,
10564 typed_splay_tree::splay_tree_foreach_helper,
10565 typed_splay_tree::splay_tree_insert,
10566 typed_splay_tree::splay_tree_remove,
10567 typed_splay_tree::splay_tree_lookup,
10568 typed_splay_tree::splay_tree_predecessor,
10569 typed_splay_tree::splay_tree_successor,
10570 typed_splay_tree::splay_tree_min,
10571 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
10572 (typed_splay_tree::root, typed_splay_tree::comp,
10573 typed_splay_tree::delete_key,
10574 typed_splay_tree::delete_value): New data members.
10575 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
10576 typed_splay_tree::remove.
10578 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
10580 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
10581 -mginv and -mno-ginv to the assembler.
10582 * config/mips/mips.opt (-mcrc): New option.
10583 (-mginv): Likewise.
10584 * doc/invoke.text (-mcrc): Document.
10585 (-mginv): Likewise.
10587 2018-06-15 Nick Clifton <nickc@redhat.com>
10590 * tree.c (escaped_string): New class. Converts an unescaped
10591 string into its escaped equivalent.
10592 (warn_deprecated_use): Use the new class to convert the
10593 deprecation message, if present.
10594 (test_escaped_strings): New self test.
10595 (test_c_tests): Add test_escaped_strings.
10596 * doc/extend.texi (deprecated): Add a note that the
10597 deprecation message is affected by the -fmessage-length
10598 option, and that control characters will be escaped.
10599 (#pragma GCC error): Document this pragma.
10600 (#pragma GCC warning): Likewise.
10601 * doc/invoke.texi (-fmessage-length): Document this option's
10602 effect on the #warning and #error preprocessor directives and
10603 the deprecated attribute.
10605 2018-06-15 Richard Biener <rguenther@suse.de>
10607 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
10608 here, also noting vector size used.
10609 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
10610 size used in MSG_OPTIMIZED_LOCATIONS dump.
10611 (pass_slp_vectorize::execute): Adjust.
10613 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
10616 * config/arc/arc.c (arc_return_address_register): Fix
10619 2018-06-15 Richard Biener <rguenther@suse.de>
10621 PR middle-end/86159
10622 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
10623 leave useless conversion stripping to force_gimple_operand_gsi.
10624 (gimplify_build2): Likewise.
10625 (gimplify_build1): Likewise.
10627 2018-06-15 Richard Biener <rguenther@suse.de>
10629 PR middle-end/86076
10630 * tree-cfg.c (move_stmt_op): unshare invariant addresses
10631 before adjusting their block.
10633 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
10635 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
10636 multilibs for *-*-rtems*.
10637 * config/riscv/t-rtems: New file.
10639 2018-06-14 Jakub Jelinek <jakub@redhat.com>
10641 PR middle-end/86122
10642 * match.pd ((A +- CST1) +- CST2): Punt if last resort
10643 unsigned_type_for returns NULL.
10646 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
10647 subregs of multi-word pseudos unless the float mode has word size.
10649 2018-06-14 Richard Biener <rguenther@suse.de>
10651 PR middle-end/86139
10652 * tree-vect-generic.c (build_word_mode_vector_type): Remove
10653 duplicate and harmful type_hash_canon.
10654 * tree.c (type_hash_canon): Assert we didn't find ourselves.
10656 2018-06-14 Richard Biener <rguenther@suse.de>
10659 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
10662 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
10664 * config/rtems.h (STDINT_LONG32): Define.
10666 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
10667 Prachi Godbole <prachi.godbole@imgtec.com>
10669 * config/mips/mips-cpus.def: Define P6600.
10670 * config/mips/mips-tables.opt: Regenerate.
10671 * config/mips/mips.c (mips_ucbranch_type): New enum.
10672 (mips_rtx_cost_data): Add support for P6600.
10673 (mips_issue_rate): Likewise.
10674 (mips_multipass_dfa_lookahead): Likewise.
10675 (mips_avoid_hazard): Likewise.
10676 (mips_reorg_process_insns): Likewise.
10677 (mips_classify_branch_p6600): New function.
10678 * config/mips/mips.h (TUNE_P6600): New define.
10679 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
10680 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
10681 * config/mips/mips.md: Include p6600.md.
10682 (processor): Add p6600.
10683 * config/mips/p6600.md: New file.
10684 * doc/invoke.texi: Add p6600 to supported architectures.
10686 2018-06-13 Martin Sebor <msebor@redhat.com>
10688 PR tree-optimization/86114
10689 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
10691 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
10693 2018-06-13 Richard Biener <rguenther@suse.de>
10695 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
10696 Properly set vector type of the intermediate stmt.
10697 * tree-vect-stmts.c (vectorizable_operation): The destination
10698 var always has vectype_out type.
10700 2018-06-13 Jeff Law <law@redhat.com>
10702 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
10703 integer 0 for argument to print_rtl_with_bb.
10704 (rl78_reorg): Likewise.
10706 2018-06-13 David Malcolm <dmalcolm@redhat.com>
10708 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
10709 from rtx to rtx_insn *.
10710 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
10712 (add_sched_insns_for_speculation): Likewise for local "target",
10713 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10714 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
10715 from rtx_insn ** to rtx_code_label **.
10716 (reorg_emit_nops): Likewise.
10717 (c6x_reorg): Likewise for local "call_labels".
10718 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
10720 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
10721 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
10722 the loops over LABEL_REFS.
10723 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
10725 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10726 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
10727 (split_branches): Strengthen local "olabel" from rtx to
10728 rtx_insn *, adding a safe_as_a cast.
10729 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
10731 (add_insn_after): Likewise for first two params.
10732 (add_insn_before): Likewise.
10733 (remove_insn): Likewise for param.
10734 (emit_pattern_before_noloc): Likewise for second and third params.
10735 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
10736 (emit_call_insn_before_noloc): Likewise.
10737 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
10739 (emit_barrier_before): Likewise.
10740 (emit_label_before): Strengthen "label" param from "rtx" to
10741 "rtx_code_label *". Strengthen "before" param from "rtx" to
10743 (emit_insn_after_1): Strengthen "after" param from "rtx" to
10745 (emit_pattern_after_noloc): Likewise.
10746 (emit_insn_after_noloc): Likewise.
10747 (emit_jump_insn_after_noloc): Likewise.
10748 (emit_call_insn_after_noloc): Likewise.
10749 (emit_debug_insn_after_noloc): Likewise.
10750 (emit_barrier_after): Likewise.
10751 (emit_label_after): Likewise for both params.
10752 (emit_pattern_after_setloc): Likewise for "after" param. Convert
10753 "loc" param from "int" to "location_t".
10754 (emit_insn_after_setloc): Likewise.
10755 (emit_jump_insn_after_setloc): Likewise.
10756 (emit_call_insn_after_setloc): Likewise.
10757 (emit_debug_insn_after_setloc): Likewise.
10758 (emit_pattern_before_setloc): Likewise for "before" param. Convert
10759 "loc" param from "int" to "location_t".
10760 (emit_pattern_before): Convert NULL_RTX to NULL.
10761 (emit_insn_before_setloc): Convert "loc" param from "int" to
10763 (emit_jump_insn_before_setloc): Likewise.
10764 (emit_call_insn_before_setloc): Likewise.
10765 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
10766 rtx_insn *. Convert "loc" param from "int" to "location_t".
10767 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
10768 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
10769 Convert 3rd param from "int" to "location_t".
10770 (emit_barrier_before, emit_barrier_after, next_real_insn):
10771 Strengthen param from rtx to rtx_insn *.
10772 (emit_label_before): Strengthen 1st param from "rtx" to
10773 "rtx_code_label *". Strengthen 2nd param from "rtx" to
10775 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
10776 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
10777 Strengthen 2nd param from "rtx" to "rtx_insn *".
10778 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
10779 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
10780 Likewise. Convert 3rd param from "int" to "location_t".
10781 (emit_label_after): Strengthen 1st param from "rtx" to
10782 "rtx_code_label *".
10783 (next_real_insn, remove_insn): Strengthen param from "rtx" to
10785 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
10786 from "rtx" to "rtx_insn *".
10788 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
10790 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
10791 bodies streamed in with -Q.
10792 * dumpfile.c (dump_files): Add lto-stream-out dump file.
10793 * dumpfile.h (tree_dump_index): Add lto_stream_out.
10794 * gimple-streamer-out.c: Include gimple-pretty-print.h
10795 (output_bb): Dump stmts streamed.
10796 * lto-section-out.c: Include print-tree.h
10797 (lto_begin_section): Dump sections created.
10798 (lto_output_decl_index): Dump decl encoded.
10799 * lto-streamer-out.c: Include print-tree.h
10800 (create_output_block): Dump output block created.
10801 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
10802 (output_function): Dump function output.
10803 (output_constructor): Dump constructor streamed.
10804 (write_global_stream): Output indexes encoded.
10805 (produce_asm_for_decls): Dump streams encoded.
10806 * lto-streamer.c (streamer_dump_file): New global var.
10807 * lto-streamer.h (streamer_dump_file): Declare.
10808 * passes.c (ipa_write_summaries): Initialize streamer dump.
10809 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
10812 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
10815 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
10816 offsets for register save directives. Emit a second batch of save
10817 directives, if need be, when the function accesses prior frames.
10819 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10821 * config/arc/fpu.md (fmasf4): Force operand to register.
10822 (fnmasf4): Likewise.
10824 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10826 * config/arc/arc-protos.h (arc_pad_return): Remove.
10827 * config/arc/arc.c (machine_function): Remove force_short_suffix
10829 (arc_print_operand): Adjust printing of '&'.
10830 (arc_verify_short): Remove conditional printing of short suffix.
10831 (arc_final_prescan_insn): Remove reference to size_reason.
10832 (pad_return): New function.
10833 (arc_reorg): Call pad_return.
10834 (arc_pad_return): Remove.
10835 (arc_init_machine_status): Remove reference to force_short_suffix.
10836 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
10837 (attr length): When attribute iscompact is true force to 2
10838 regardless; in the case of maybe check if we want to force the
10839 instruction to have 4 bytes length.
10840 (nopv): Change it to generate 4 byte long nop as well.
10841 (blockage): New pattern.
10842 (simple_return): Remove call to arc_pad_return.
10843 (p_return_i): Likewise.
10845 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10847 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
10849 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10851 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
10854 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10856 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
10857 for ARC700 and ARCv2.
10859 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
10862 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
10863 operands[2] instead of operands[1].
10866 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10868 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
10869 case, check whether the outer register overlaps an unallocatable
10870 register, not just whether it fits the required class.
10872 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10874 * poly-int.h (can_div_trunc_p): Add new overload in which all values
10876 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
10877 (memrefs_conflict_p): Likewise.
10878 (init_alias_analysis): Likewise.
10879 * cfgexpand.c (expand_debug_expr): Likewise.
10880 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
10881 * cse.c (fold_rtx): Likewise.
10882 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
10883 * expr.c (emit_block_move_hints): Likewise.
10884 (clear_storage_hints, push_block, emit_push_insn): Likewise.
10885 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
10886 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
10887 (emit_group_store): Likewise.
10888 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
10889 to read the PRE/POST_MODIFY increment.
10890 * calls.c (store_one_arg): Use strip_offset.
10891 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
10893 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
10895 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
10896 (simplify_binary_operation_1): Extend CONST_INT handling to
10898 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
10899 than a HOST_WIDE_INT.
10900 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
10902 (adjust_mems, add_stores): Update accodingly.
10903 (vt_canonicalize_addr): Track polynomial offsets.
10904 (emit_note_insn_var_location): Likewise.
10905 (vt_add_function_parameter): Likewise.
10906 (vt_initialize): Likewise.
10908 2018-06-12 Jeff Law <law@redhat.com>
10910 * config.gcc (alpha*-*-freebsd*): Remove.
10911 * config/alpha/freebsd.h: Remove.
10913 2018-06-12 David Malcolm <dmalcolm@redhat.com>
10916 * spellcheck-tree.c (levenshtein_distance): Rename to...
10917 (get_edit_distance): ...this, and update for underlying renaming.
10918 * spellcheck-tree.h (levenshtein_distance): Rename to...
10919 (get_edit_distance): ...this.
10920 * spellcheck.c (levenshtein_distance): Rename to...
10921 (get_edit_distance): ...this. Convert from Levenshtein distance
10922 to Damerau-Levenshtein distance by supporting transpositions of
10923 adjacent characters. Rename "v1" to "v_next" and "v0" to
10925 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
10926 (selftest::test_edit_distance_unit_test_oneway): ...this, and
10927 update for underlying renaming.
10928 (selftest::levenshtein_distance_unit_test): Rename to...
10929 (selftest::test_get_edit_distance_unit): ...this, and update for
10930 underlying renaming.
10931 (selftest::test_find_closest_string): Add example from PR 69968
10932 where transposition helps
10933 (selftest::test_metric_conditions): Update for renaming.
10934 (selftest::test_metric_conditions): Likewise.
10935 (selftest::spellcheck_c_tests): Likewise.
10936 * spellcheck.h (levenshtein_distance): Rename both overloads to...
10937 (get_edit_distance): ...this.
10938 (best_match::consider): Update for renaming.
10940 2018-06-12 Martin Sebor <msebor@redhat.com>
10942 PR tree-optimization/85259
10943 * builtins.c (compute_objsize): Handle constant offsets.
10944 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
10945 true iff a warning has been issued.
10946 * gimple.h (gimple_nonartificial_location): New function.
10947 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
10948 gimple_nonartificial_location and handle -Wno-system-headers.
10949 (handle_builtin_stxncpy): Same.
10951 2018-06-12 Martin Sebor <msebor@redhat.com>
10954 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
10956 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
10958 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
10959 BUILTIN_VEC_XST entries for pointer to double and long long.
10961 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
10964 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10966 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10969 2018-06-12 Martin Liska <mliska@suse.cz>
10971 * doc/options.texi: Document IntegerRange.
10973 2018-06-12 Martin Liska <mliska@suse.cz>
10975 * config/i386/i386.opt: Make MPX-related options as Deprecated.
10976 * opt-functions.awk: Handle Deprecated flag.
10977 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
10979 (read_cmdline_option): Report warning for a deprecated option.
10980 * opts.h (struct cl_option): Add new field cl_deprecated.
10981 (CL_ERR_DEPRECATED): New.
10983 2018-06-12 Martin Liska <mliska@suse.cz>
10985 * doc/options.texi: Document Deprecated option flag.
10987 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10989 * config/arc/arc-arch.h (arc_extras): New enum.
10990 (arc_cpu_t):Add field extra.
10991 (arc_cpu_types): Consider the extras.
10992 * config/arc/arc-cpus.def: Add extras info.
10993 * config/arc/arc-opts.h (processor_type): Consider extra field.
10994 * config/arc/arc.c (arc_override_options): Handle extra field.
10996 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10998 * config/arc/arc-arch.h: Update ARC_OPTX macro.
10999 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11001 * config/arc/arc.c (arc_init): Update pic warning.
11002 (irq_range): Update irq range parsing warnings.
11003 (arc_override_options): Update various warning messages.
11004 (arc_handle_aux_attribute): Likewise.
11006 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
11008 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11010 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11012 * doc/sourcebuild.texi: Document usage of line number 0 in verify
11013 compiler messages directives.
11015 2018-06-12 Matthew Fortune <mfortune@gmail.com>
11017 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11018 * config/mips/mips-tables.opt: Regenerate.
11019 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11021 * doc/invoke.texi: Document -march=i6500.
11023 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
11025 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11026 (i6400_gpmul): Add cpu_unit.
11027 (i6400_gpdiv): Likewise.
11028 (i6400_msa_add_d): Update reservations.
11029 (i6400_msa_int_add) Likewise.
11030 (i6400_msa_short_logic3) Likewise.
11031 (i6400_msa_short_logic2) Likewise.
11032 (i6400_msa_short_logic) Likewise.
11033 (i6400_msa_move) Likewise.
11034 (i6400_msa_cmp) Likewise.
11035 (i6400_msa_short_float2) Likewise.
11036 (i6400_msa_div_d) Likewise.
11037 (i6400_msa_long_logic1) Likewise.
11038 (i6400_msa_long_logic2) Likewise.
11039 (i6400_msa_mult) Likewise.
11040 (i6400_msa_long_float2) Likewise.
11041 (i6400_msa_long_float4) Likewise.
11042 (i6400_msa_long_float5) Likewise.
11043 (i6400_msa_long_float8) Likewise.
11044 (i6400_fpu_fadd): Include frint type.
11045 (i6400_fpu_store): New define_insn_reservation.
11046 (i6400_fpu_load): Likewise.
11047 (i6400_fpu_move): Likewise.
11048 (i6400_fpu_fcmp): Likewise.
11049 (i6400_fpu_fmadd): Likewise.
11050 (i6400_int_mult): Include imul3nc type and update reservation.
11051 (i6400_int_div): Include idiv3 type and update reservation.
11052 (i6400_int_load): Update to check type not move_type.
11053 (i6400_int_store): Likewise.
11054 (i6400_int_prefetch): Set zero latency.
11056 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
11058 * gcc.c: Document new %@{...} sequence.
11059 (LINK_COMMAND_SPEC): Use it for the -L switches.
11060 (cpp_unique_options): Use it for the -I switches.
11061 (at_file_argbuf): New global variable.
11062 (in_at_file): Likewise.
11063 (alloc_args): Create at_file_argbuf.
11064 (clear_args): Truncate at_file_argbuf.
11065 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11066 (open_at_file): New function.
11067 (close_at_file): Likewise.
11068 (create_at_file): Delete.
11069 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11072 (validate_switches_from_spec): Deal with %@{...} sequence.
11073 (validate_switches): Likewise.
11074 (driver::finalize): Call clear_args.
11076 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
11078 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11080 2018-06-11 Martin Sebor <msebor@redhat.com>
11082 * doc/invoke.texi (-Wall): List -Wc++17-compat.
11083 (Wno-class-memaccess): Add @opindex.
11084 (Wno-templates, Wno-multiple-inheritance): Same.
11085 (Wno-virtual-inheritance, Wno-namespaces): Same.
11086 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11087 (Wno-format-overflow, Wno-format-truncation): Same.
11088 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11089 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11090 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11091 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11092 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11093 (Wno-misspelled-isr): Same.
11095 2018-06-11 Martin Sebor <msebor@redhat.com>
11097 * PR tree-optimization/86083
11098 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11100 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
11102 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11104 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
11107 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11108 on the correct operand.
11109 (*movdi_internal64): Ditto.
11111 2018-06-11 Martin Liska <mliska@suse.cz>
11113 PR tree-optimization/86089
11114 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11116 2018-06-11 Julia Koval <julia.koval@intel.com>
11118 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11119 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11120 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11122 2018-06-11 Olivier Hainque <hainque@adacore.com>
11124 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11125 for Ada with strict dwarf2.
11127 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
11130 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11133 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
11135 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11137 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
11139 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11142 2018-06-08 Martin Liska <mliska@suse.cz>
11144 * tree-cfg.h (debug_function): Fix argument type to match
11147 2018-06-08 Martin Liska <mliska@suse.cz>
11149 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11150 Remove usage of MPX-related (and removed) fields.
11151 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11153 2018-06-08 David Malcolm <dmalcolm@redhat.com>
11155 * cfg.c (debug): Use TDF_NONE rather than 0.
11156 * cfghooks.c (debug): Likewise.
11157 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11158 (struct dump_option_value_info): Convert to...
11159 (struct kv_pair): ...this template type.
11160 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11162 (optinfo_verbosity_options): Likewise.
11163 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11165 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11166 than int for "optgroup_flags" param.
11167 (dump_generic_expr_loc): Use dump_flags_t rather than int for
11169 (dump_dec): Likewise.
11170 (dump_finish): Use TDF_NONE rather than 0.
11171 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11172 rather than int for "optgroup_flags" param. Use TDF_NONE rather
11173 than 0. Update for change to option_ptr.
11174 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11175 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
11176 0. Update for changes to optinfo_verbosity_options and
11178 (opt_info_switch_p): Convert optgroup_flags from int to
11180 (dump_basic_block): Use dump_flags_t rather than int
11181 for "dump_kind" param.
11182 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11183 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11184 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11185 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11186 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11187 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11188 TDF_NONE): Convert from macros to...
11189 (enum dump_flag): ...this new enum.
11190 (dump_flags_t): Update to use enum.
11191 (operator|, operator&, operator~, operator|=, operator&=):
11192 Implement for dump_flags_t.
11193 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11194 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11195 Convert from macros to...
11196 (enum optgroup_flag): ...this new enum.
11197 (optgroup_flags_t): New typedef.
11198 (operator|, operator|=): Implement for optgroup_flags_t.
11199 (struct dump_file_info): Convert field "alt_flags" to
11200 dump_flags_t. Convert field "optgroup_flags" to
11202 (dump_basic_block): Use dump_flags_t rather than int for param.
11203 (dump_generic_expr_loc): Likewise.
11204 (dump_dec): Likewise.
11205 (dump_register): Convert param "optgroup_flags" to
11207 (opt_info_enable_passes): Likewise.
11208 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11210 * gimple-pretty-print.c (debug): Likewise.
11211 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11212 (merged_store_group::apply_stores): Likewise.
11213 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11214 * gimple.c (verify_gimple_pp): Likewise.
11215 * graphite-poly.c (print_pbb_body): Likewise.
11216 * passes.c (pass_manager::register_one_dump_file): Convert
11217 local "optgroup_flags" to optgroup_flags_t.
11218 * print-tree.c (print_node): Use TDF_NONE rather than 0.
11220 (debug_body): Likewise.
11221 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11222 to optgroup_flags_t.
11223 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11225 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11226 (convert_mult_to_fma): Likewise.
11227 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11228 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11229 * tree-vect-data-refs.c (dump_lower_bound): Convert param
11230 "dump_kind" to dump_flags_t.
11232 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11234 * config/rs6000/rs6000.c (min, max): Delete.
11236 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11238 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11239 -mabi=spe and -mabi=no-spe.
11241 2018-06-08 Martin Liska <mliska@suse.cz>
11243 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11244 where we expect an existing summary.
11246 2018-06-08 Martin Liska <mliska@suse.cz>
11248 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11249 * ipa-inline.h (estimate_edge_growth): Likewise.
11251 2018-06-08 Martin Liska <mliska@suse.cz>
11253 * cgraph.c (function_version_hasher::hash): Use
11254 cgraph_node::get_uid ().
11255 (function_version_hasher::equal):
11256 * cgraph.h (cgraph_node::get_uid): New method.
11257 * ipa-inline.c (update_caller_keys): Use
11258 cgraph_node::get_uid ().
11259 (update_callee_keys): Likewise.
11260 * ipa-utils.c (searchc): Likewise.
11261 (ipa_reduced_postorder): Likewise.
11262 * lto-cgraph.c (input_node): Likewise.
11263 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11264 * symbol-summary.h (symtab_insertion): Likewise.
11265 (symtab_removal): Likewise.
11266 (symtab_duplication): Likewise.
11267 * tree-pretty-print.c (dump_function_header): Likewise.
11268 * tree-sra.c (convert_callers_for_node): Likewise.
11270 2018-06-08 Martin Liska <mliska@suse.cz>
11272 * cgraph.c (symbol_table::create_edge): Always assign a new
11274 (symbol_table::free_edge): Do not recycle numbers.
11275 * cgraph.h (cgraph_edge::get): New method.
11276 * symbol-summary.h (symtab_removal): Use it.
11277 (symtab_duplication): Likewise.
11278 (call_summary::hashable_uid): Remove.
11280 2018-06-08 Martin Liska <mliska@suse.cz>
11282 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11283 (initialize_growth_caches): Remove.
11284 (free_growth_caches): Likewise.
11285 (do_estimate_edge_time): Use edge_growth_cache.
11286 (do_estimate_edge_size): Likewise.
11287 (do_estimate_edge_hints): Likewise.
11288 * ipa-inline.c (reset_edge_caches): Likewise.
11289 (recursive_inlining): Likewise.
11290 (inline_small_functions): Likewise.
11291 * ipa-inline.h (initialize_growth_caches): Remove.
11292 (estimate_edge_size): Likewise.
11293 (estimate_edge_time): Likewise.
11294 (estimate_edge_hints): Likewise.
11295 (reset_edge_growth_cache): Likewise.
11296 * symbol-summary.h (call_summary::remove): New method.
11298 2018-06-08 Martin Liska <mliska@suse.cz>
11300 * ipa-cp.c (class edge_clone_summary): New summary.
11301 (grow_edge_clone_vectors): Remove.
11302 (ipcp_edge_duplication_hook): Remove.
11303 (class edge_clone_summary_t): New call_summary class.
11304 (ipcp_edge_removal_hook): Remove.
11305 (edge_clone_summary_t::duplicate): New function.
11306 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11307 (create_specialized_node): Likewise.
11308 (ipcp_driver): Initialize edge_clone_summaries and do not
11311 2018-06-08 Martin Liska <mliska@suse.cz>
11313 * symbol-summary.h (get): New function.
11314 (call_summary::m_initialize_when_cloning): New class member.
11316 2018-06-08 Martin Liska <mliska@suse.cz>
11318 * cgraph.c (cgraph_node::remove): Do not recycle uid.
11319 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11320 (symbol_table::allocate_cgraph_symbol): Do not set uid.
11321 * passes.c (uid_hash_t): Record removed_nodes by their uids.
11322 (remove_cgraph_node_from_order): Use the removed_nodes set.
11323 (do_per_function_toporder): Likwise.
11324 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11325 instead of summary_uid.
11326 (symtab_removal): Likewise.
11327 (symtab_duplication): Likewise.
11329 2018-06-08 Martin Liska <mliska@suse.cz>
11331 * ipa-cp.c (ipcp_store_bits_results): Use
11332 ipcp_transformation_sum.
11333 (ipcp_store_vr_results): Likewise.
11334 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
11336 (ipcp_transformation_initialize): ... this.
11337 (ipa_set_node_agg_value_chain):
11338 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
11339 (write_ipcp_transformation_info): Likewise.
11340 (read_ipcp_transformation_info): Likewise.
11341 (ipcp_update_bits): Likewise.
11342 (ipcp_update_vr): Likewise.
11343 (ipcp_transform_function): Likewise.
11344 * ipa-prop.h: Rename ipcp_transformation_summary to
11345 ipcp_transformation.
11346 (class ipcp_transformation_t): New function summary.
11347 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
11348 (ipa_get_agg_replacements_for_node): Likewise.
11350 2018-06-08 Martin Liska <mliska@suse.cz>
11352 * ipa-pure-const.c (struct funct_state_d): Do it class instead
11354 (class funct_state_summary_t): New function_summary class.
11355 (has_function_state): Remove.
11356 (get_function_state): Likewise.
11357 (set_function_state): Likewise.
11358 (add_new_function): Likewise.
11359 (funct_state_summary_t::insert): New function.
11360 (duplicate_node_data): Remove.
11361 (remove_node_data): Remove.
11362 (funct_state_summary_t::duplicate): New function.
11363 (register_hooks): Create new funct_state_summaries.
11364 (pure_const_generate_summary): Use it.
11365 (pure_const_write_summary): Likewise.
11366 (pure_const_read_summary): Likewise.
11367 (propagate_pure_const): Likewise.
11368 (propagate_nothrow): Likewise.
11369 (dump_malloc_lattice): Likewise.
11370 (propagate_malloc): Likewise.
11371 (execute): Do not register hooks, just remove summary
11373 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
11376 2018-06-08 Martin Liska <mliska@suse.cz>
11378 * ipa-reference.c (remove_node_data): Remove.
11379 (duplicate_node_data): Likewise.
11380 (class ipa_ref_var_info_summary_t): New class.
11381 (class ipa_ref_opt_summary_t): Likewise.
11382 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
11383 (get_reference_optimization_summary): Use
11384 ipa_ref_opt_sum_summaries.
11385 (set_reference_vars_info): Remove.
11386 (set_reference_optimization_summary): Likewise.
11387 (ipa_init): Create summaries.
11388 (init_function_info): Use function summary.
11389 (ipa_ref_opt_summary_t::duplicate): New function.
11390 (ipa_ref_opt_summary_t::remove): New function.
11391 (get_read_write_all_from_node): Fix GNU coding style.
11392 (propagate): Use function summary.
11393 (write_node_summary_p): Fix GNU coding style.
11394 (stream_out_bitmap): Likewise.
11395 (ipa_reference_read_optimization_summary): Use function summary.
11396 (ipa_reference_c_finalize): Do not release hooks.
11398 2018-06-08 Martin Liska <mliska@suse.cz>
11400 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
11401 (analyze_function_body): Extract multiple calls of get_create.
11402 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11403 * ipa-inline.c (recursive_inlining): Use ::get method.
11404 * ipa-inline.h (estimate_edge_growth): Likewise.
11406 2018-06-08 Martin Liska <mliska@suse.cz>
11408 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
11410 (hsa_function_summary::hsa_function_summary): Use the new enum
11412 (hsa_gpu_implementation_p): Use hsa_summaries::get.
11413 * hsa-gen.c (hsa_get_host_function): Likewise.
11414 (get_brig_function_name): Likewise.
11415 * ipa-hsa.c (process_hsa_functions): Likewise.
11416 (ipa_hsa_write_summary): Likewise.
11417 * symbol-summary.h (symtab_duplication): Use ::get function/
11418 (get): New function.
11420 2018-06-08 Martin Liska <mliska@suse.cz>
11422 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
11424 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
11425 (hsa_register_kernel): Likewise.
11426 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
11427 * hsa-gen.c (hsa_get_host_function): Likewise.
11428 (get_brig_function_name): Likewise.
11429 (generate_hsa): Likewise.
11430 (pass_gen_hsail::execute): Likewise.
11431 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
11432 (devirtualization_time_bonus): Likewise.
11433 (ipcp_propagate_stage): Likewise.
11434 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11435 (edge_set_predicate): Likewise.
11436 (evaluate_conditions_for_known_args): Likewise.
11437 (evaluate_properties_for_edge): Likewise.
11438 (ipa_fn_summary::reset): Likewise.
11439 (ipa_fn_summary_t::duplicate): Likewise.
11440 (dump_ipa_call_summary): Likewise.
11441 (ipa_dump_fn_summary): Likewise.
11442 (analyze_function_body): Likewise.
11443 (compute_fn_summary): Likewise.
11444 (estimate_edge_devirt_benefit): Likewise.
11445 (estimate_edge_size_and_time): Likewise.
11446 (estimate_calls_size_and_time): Likewise.
11447 (estimate_node_size_and_time): Likewise.
11448 (inline_update_callee_summaries): Likewise.
11449 (remap_edge_change_prob): Likewise.
11450 (remap_edge_summaries): Likewise.
11451 (ipa_merge_fn_summary_after_inlining): Likewise.
11452 (ipa_update_overall_fn_summary): Likewise.
11453 (read_ipa_call_summary): Likewise.
11454 (inline_read_section): Likewise.
11455 (write_ipa_call_summary): Likewise.
11456 (ipa_fn_summary_write): Likewise.
11457 (ipa_free_fn_summary): Likewise.
11458 * ipa-hsa.c (process_hsa_functions): Likewise.
11459 (ipa_hsa_write_summary): Likewise.
11460 (ipa_hsa_read_section): Likewise.
11461 * ipa-icf.c (sem_function::merge): Likewise.
11462 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11463 (do_estimate_edge_time): Likewise.
11464 (estimate_size_after_inlining): Likewise.
11465 (estimate_growth): Likewise.
11466 (growth_likely_positive): Likewise.
11467 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11468 (inline_call): Likewise.
11469 * ipa-inline.c (caller_growth_limits): Likewise.
11470 (can_inline_edge_p): Likewise.
11471 (can_inline_edge_by_limits_p): Likewise.
11472 (compute_uninlined_call_time): Likewise.
11473 (compute_inlined_call_time): Likewise.
11474 (want_inline_small_function_p): Likewise.
11475 (edge_badness): Likewise.
11476 (update_caller_keys): Likewise.
11477 (update_callee_keys): Likewise.
11478 (recursive_inlining): Likewise.
11479 (inline_small_functions): Likewise.
11480 (inline_to_all_callers_1): Likewise.
11481 (dump_overall_stats): Likewise.
11482 (early_inline_small_functions): Likewise.
11483 (early_inliner): Likewise.
11484 * ipa-inline.h (estimate_edge_growth): Likewise.
11485 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11486 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11487 * ipa-prop.h (IPA_NODE_REF): Likewise.
11488 (IPA_EDGE_REF): Likewise.
11489 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11490 (propagate_malloc): Likewise.
11491 * ipa-split.c (execute_split_functions): Likewise.
11492 * symbol-summary.h: Rename get to get_create.
11494 (get_create): Likewise.
11495 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11497 2018-06-08 Martin Liska <mliska@suse.cz>
11499 * symbol-summary.h (release): Move definition out of class
11501 (symtab_removal): Likewise.
11502 (symtab_duplication): Likewise.
11504 2018-06-08 Martin Liska <mliska@suse.cz>
11506 * symbol-summary.h (function_summary): Move constructor
11507 implementation out of class declaration.
11508 (release): Likewise.
11509 (symtab_insertion): Likewise.
11510 (symtab_removal): Likewise.
11511 (symtab_duplication): Likewise.
11514 2018-06-08 Martin Liska <mliska@suse.cz>
11516 * Makefile.in: Remove support for MPX (macros, related functions,
11517 fields in cgraph_node, ...).
11518 * builtin-types.def (BT_BND): Likewise.
11519 (BT_FN_BND_CONST_PTR): Likewise.
11520 (BT_FN_CONST_PTR_BND): Likewise.
11521 (BT_FN_VOID_PTR_BND): Likewise.
11522 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
11523 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
11524 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
11525 (expand_builtin_mempcpy_with_bounds): Likewise.
11526 (expand_builtin_memset_with_bounds): Likewise.
11527 (expand_builtin_memset_args): Likewise.
11528 (std_expand_builtin_va_start): Likewise.
11529 (expand_builtin): Likewise.
11530 (expand_builtin_with_bounds): Likewise.
11531 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
11532 (DEF_LIB_BUILTIN_CHKP): Likewise.
11533 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
11534 (DEF_CHKP_BUILTIN): Likewise.
11535 (BUILT_IN_MEMCPY): Likewise.
11536 (BUILT_IN_MEMMOVE): Likewise.
11537 (BUILT_IN_MEMPCPY): Likewise.
11538 (BUILT_IN_MEMSET): Likewise.
11539 (BUILT_IN_STPCPY): Likewise.
11540 (BUILT_IN_STRCAT): Likewise.
11541 (BUILT_IN_STRCHR): Likewise.
11542 (BUILT_IN_STRCPY): Likewise.
11543 (BUILT_IN_STRLEN): Likewise.
11544 (BUILT_IN_MEMCPY_CHK): Likewise.
11545 (BUILT_IN_MEMMOVE_CHK): Likewise.
11546 (BUILT_IN_MEMPCPY_CHK): Likewise.
11547 (BUILT_IN_MEMSET_CHK): Likewise.
11548 (BUILT_IN_STPCPY_CHK): Likewise.
11549 (BUILT_IN_STRCAT_CHK): Likewise.
11550 (BUILT_IN_STRCPY_CHK): Likewise.
11551 * calls.c (store_bounds): Likewise.
11552 (emit_call_1): Likewise.
11553 (special_function_p): Likewise.
11554 (maybe_warn_nonstring_arg): Likewise.
11555 (initialize_argument_information): Likewise.
11556 (finalize_must_preallocate): Likewise.
11557 (compute_argument_addresses): Likewise.
11558 (expand_call): Likewise.
11559 * cfgexpand.c (expand_call_stmt): Likewise.
11560 (expand_return): Likewise.
11561 (expand_gimple_stmt_1): Likewise.
11562 (pass_expand::execute): Likewise.
11563 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
11564 (cgraph_node::remove): Likewise.
11565 (cgraph_node::dump): Likewise.
11566 (cgraph_node::verify_node): Likewise.
11567 * cgraph.h (chkp_function_instrumented_p): Likewise.
11568 (symtab_node::get_alias_target): Likewise.
11569 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
11570 (cgraph_local_p): Likewise.
11571 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
11572 (cgraph_edge::rebuild_references): Likewise.
11573 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
11574 (walk_polymorphic_call_targets): Likewise.
11575 (cgraph_node::expand_thunk): Likewise.
11576 (symbol_table::output_weakrefs): Likewise.
11577 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
11578 (ix86_handle_option): Likewise.
11579 * config/i386/constraints.md: Likewise.
11580 * config/i386/i386-builtin-types.def (BND): Likewise.
11584 * config/i386/i386-builtin.def (BDESC_END): Likewise.
11585 (BDESC_FIRST): Likewise.
11587 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
11588 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
11589 * config/i386/i386.c (enum reg_class): Likewise.
11590 (ix86_target_string): Likewise.
11591 (ix86_option_override_internal): Likewise.
11592 (ix86_conditional_register_usage): Likewise.
11593 (ix86_valid_target_attribute_inner_p): Likewise.
11594 (ix86_set_indirect_branch_type): Likewise.
11595 (ix86_set_current_function): Likewise.
11596 (ix86_function_arg_regno_p): Likewise.
11597 (init_cumulative_args): Likewise.
11598 (ix86_function_arg_advance): Likewise.
11599 (ix86_function_arg): Likewise.
11600 (ix86_pass_by_reference): Likewise.
11601 (ix86_function_value_regno_p): Likewise.
11602 (ix86_function_value_1): Likewise.
11603 (ix86_function_value_bounds): Likewise.
11604 (ix86_return_in_memory): Likewise.
11605 (ix86_setup_incoming_vararg_bounds): Likewise.
11606 (ix86_va_start): Likewise.
11607 (indirect_thunk_need_prefix): Likewise.
11608 (print_reg): Likewise.
11609 (ix86_print_operand): Likewise.
11610 (ix86_expand_call): Likewise.
11611 (ix86_output_function_return): Likewise.
11612 (reg_encoded_number): Likewise.
11613 (BDESC_VERIFYS): Likewise.
11614 (ix86_init_mpx_builtins): Likewise.
11615 (ix86_init_builtins): Likewise.
11616 (ix86_emit_cmove): Likewise.
11617 (ix86_emit_move_max): Likewise.
11618 (ix86_expand_builtin): Likewise.
11619 (ix86_builtin_mpx_function): Likewise.
11620 (ix86_get_arg_address_for_bt): Likewise.
11621 (ix86_load_bounds): Likewise.
11622 (ix86_store_bounds): Likewise.
11623 (ix86_load_returned_bounds): Likewise.
11624 (ix86_store_returned_bounds): Likewise.
11625 (ix86_class_likely_spilled_p): Likewise.
11626 (ix86_hard_regno_mode_ok): Likewise.
11627 (x86_order_regs_for_local_alloc): Likewise.
11628 (ix86_mitigate_rop): Likewise.
11629 (ix86_bnd_prefixed_insn_p): Likewise.
11630 (ix86_mpx_bound_mode): Likewise.
11631 (ix86_make_bounds_constant): Likewise.
11632 (ix86_initialize_bounds): Likewise.
11633 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
11634 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
11635 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
11636 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
11637 (TARGET_CHKP_BOUND_MODE): Likewise.
11638 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
11639 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
11640 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
11641 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
11642 * config/i386/i386.h (TARGET_MPX): Likewise.
11643 (TARGET_MPX_P): Likewise.
11644 (VALID_BND_REG_MODE): Likewise.
11645 (FIRST_BND_REG): Likewise.
11646 (LAST_BND_REG): Likewise.
11647 (enum reg_class): Likewise.
11648 (BND_REG_P): Likewise.
11649 (BND_REGNO_P): Likewise.
11650 (BNDmode): Likewise.
11651 (ADJUST_INSN_LENGTH): Likewise.
11652 * config/i386/i386.md: Likewise.
11653 * config/i386/i386.opt: Likewise.
11654 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
11655 (defined): Likewise.
11656 (LINK_MPX): Likewise.
11657 (MPX_SPEC): Likewise.
11658 (LIBMPX_SPEC): Likewise.
11659 (LIBMPXWRAPPERS_SPEC): Likewise.
11660 (CHKP_SPEC): Likewise.
11661 * config/i386/predicates.md: Likewise.
11662 * dbxout.c (dbxout_type): Likewise.
11663 * doc/extend.texi: Likewise.
11664 * doc/invoke.texi: Likewise.
11665 * doc/md.texi: Likewise.
11666 * doc/tm.texi: Likewise.
11667 * doc/tm.texi.in: Likewise.
11668 * dwarf2out.c (is_base_type): Likewise.
11669 (gen_formal_types_die): Likewise.
11670 (gen_subprogram_die): Likewise.
11671 (gen_type_die_with_usage): Likewise.
11672 (gen_decl_die): Likewise.
11673 (dwarf2out_late_global_decl): Likewise.
11674 * expr.c (expand_assignment): Likewise.
11675 (emit_storent_insn): Likewise.
11676 (store_expr_with_bounds): Likewise.
11677 (store_expr): Likewise.
11678 (expand_expr_real_1): Likewise.
11679 * expr.h (store_expr_with_bounds): Likewise.
11680 * function.c (use_register_for_decl): Likewise.
11681 (struct bounds_parm_data): Likewise.
11682 (assign_parms_augmented_arg_list): Likewise.
11683 (assign_parm_find_entry_rtl): Likewise.
11684 (assign_parm_is_stack_parm): Likewise.
11685 (assign_parm_load_bounds): Likewise.
11686 (assign_bounds): Likewise.
11687 (assign_parms): Likewise.
11688 (expand_function_start): Likewise.
11689 * gcc.c (CHKP_SPEC): Likewise.
11690 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
11691 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
11692 (wrestrict_dom_walker::check_call): Likewise.
11693 * gimple.c (gimple_build_call_from_tree): Likewise.
11694 * gimple.h (enum gf_mask): Likewise.
11695 (gimple_call_with_bounds_p): Likewise.
11696 (gimple_call_set_with_bounds): Likewise.
11697 * gimplify.c (gimplify_init_constructor): Likewise.
11698 * ipa-cp.c (initialize_node_lattices): Likewise.
11699 (propagate_constants_across_call): Likewise.
11700 (find_more_scalar_values_for_callers_subset): Likewise.
11701 * ipa-hsa.c (process_hsa_functions): Likewise.
11702 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
11703 * ipa-icf.c (sem_function::merge): Likewise.
11704 * ipa-inline.c (early_inliner): Likewise.
11705 * ipa-pure-const.c (warn_function_noreturn): Likewise.
11706 (warn_function_cold): Likewise.
11707 (propagate_pure_const): Likewise.
11708 * ipa-ref.h (enum GTY): Likewise.
11709 * ipa-split.c (find_retbnd): Likewise.
11710 (consider_split): Likewise.
11711 (split_function): Likewise.
11712 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11713 * ipa.c (walk_polymorphic_call_targets): Likewise.
11714 (symbol_table::remove_unreachable_nodes): Likewise.
11715 (process_references): Likewise.
11716 (cgraph_build_static_cdtor_1): Likewise.
11717 * lto-cgraph.c (lto_output_node): Likewise.
11718 (output_refs): Likewise.
11719 (compute_ltrans_boundary): Likewise.
11720 (input_overwrite_node): Likewise.
11721 (input_node): Likewise.
11722 (input_cgraph_1): Likewise.
11723 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
11724 * passes.c (pass_manager::execute_early_local_passes): Likewise.
11725 (class pass_chkp_instrumentation_passes): Likewise.
11726 (make_pass_chkp_instrumentation_passes): Likewise.
11727 * passes.def: Likewise.
11728 * rtl.h (struct GTY): Likewise.
11729 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
11730 * stor-layout.c (layout_type): Likewise.
11731 * symtab.c: Likewise.
11732 * target.def: Likewise.
11733 * targhooks.c (default_chkp_bound_type): Likewise.
11734 (default_chkp_bound_mode): Likewise.
11735 (default_builtin_chkp_function): Likewise.
11736 (default_chkp_function_value_bounds): Likewise.
11737 (default_chkp_make_bounds_constant): Likewise.
11738 (default_chkp_initialize_bounds): Likewise.
11739 * targhooks.h (default_chkp_bound_type): Likewise.
11740 (default_chkp_bound_mode): Likewise.
11741 (default_builtin_chkp_function): Likewise.
11742 (default_chkp_function_value_bounds): Likewise.
11743 (default_chkp_make_bounds_constant): Likewise.
11744 (default_chkp_initialize_bounds): Likewise.
11745 * toplev.c (compile_file): Likewise.
11746 (process_options): Likewise.
11747 * tree-core.h (DEF_BUILTIN): Likewise.
11748 (DEF_BUILTIN_CHKP): Likewise.
11749 * tree-inline.c (declare_return_variable): Likewise.
11750 (remap_gimple_stmt): Likewise.
11751 (copy_bb): Likewise.
11752 (initialize_inlined_parameters): Likewise.
11753 (expand_call_inline): Likewise.
11754 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
11755 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
11756 (make_pass_ipa_chkp_produce_thunks): Likewise.
11757 (make_pass_chkp): Likewise.
11758 (make_pass_chkp_opt): Likewise.
11759 (make_pass_chkp_instrumentation_passes): Likewise.
11760 * tree-pretty-print.c (dump_generic_node): Likewise.
11761 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
11762 * tree-ssa-dce.c (propagate_necessity): Likewise.
11763 (eliminate_unnecessary_stmts): Likewise.
11764 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
11765 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
11766 * tree-ssa-sccvn.h: Likewise.
11767 * tree-ssa-strlen.c (get_string_length): Likewise.
11768 (valid_builtin_call): Likewise.
11769 (adjust_last_stmt): Likewise.
11770 (handle_builtin_strchr): Likewise.
11771 (handle_builtin_strcpy): Likewise.
11772 (handle_builtin_stxncpy): Likewise.
11773 (handle_builtin_memcpy): Likewise.
11774 (handle_builtin_strcat): Likewise.
11775 (strlen_check_and_optimize_stmt): Likewise.
11776 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
11777 * tree-streamer-in.c: Likewise.
11778 * tree-streamer.c (record_common_node): Likewise.
11779 * tree.c (tree_code_size): Likewise.
11780 (wide_int_to_tree_1): Likewise.
11781 (type_contains_placeholder_1): Likewise.
11782 (build_common_tree_nodes): Likewise.
11783 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
11784 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
11785 (POINTER_BOUNDS_P): Likewise.
11786 (BOUNDED_TYPE_P): Likewise.
11787 (BOUNDED_P): Likewise.
11788 (CALL_WITH_BOUNDS_P): Likewise.
11789 (pointer_bounds_type_node): Likewise.
11790 * value-prof.c (gimple_ic): Likewise.
11791 * var-tracking.c (vt_add_function_parameters): Likewise.
11792 * varasm.c (make_decl_rtl): Likewise.
11793 (assemble_start_function): Likewise.
11794 (output_constant): Likewise.
11795 (maybe_assemble_visibility): Likewise.
11796 * varpool.c (ctor_for_folding): Likewise.
11797 * chkp-builtins.def: Remove.
11798 * ipa-chkp.c: Remove.
11799 * ipa-chkp.h: Remove.
11800 * rtl-chkp.c: Remove.
11801 * rtl-chkp.h: Remove.
11802 * tree-chkp-opt.c: Remove.
11803 * tree-chkp.c: Remove.
11804 * tree-chkp.h: Remove.
11806 2018-06-07 Carl Love <cel@us.ibm.com>
11808 * config/rs6000/vsx.md (vextract_fp_from_shorth,
11809 vextract_fp_from_shortl): Add BE support.
11811 2018-06-07 Paul Koning <ni1d@arrl.net>
11813 * compare-elim.c (try_merge_compare): Don't merge compare if
11814 address contains a side effect.
11815 (try_eliminate_compare): Likewise.
11817 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
11819 * config.gcc: Support "tremont".
11820 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
11821 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11823 * config/i386/i386.c (m_TREMONT): Define.
11824 (processor_target_table): Add "tremont".
11825 (PTA_TREMONT): Define.
11826 (ix86_lea_outperforms): Add TARGET_TREMONT.
11827 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
11828 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
11829 and M_INTEL_GOLDMONT_PLUS.
11830 (fold_builtin_cpu): Add "tremont".
11831 (ix86_add_stmt_cost): Add TARGET_TREMONT.
11832 (ix86_option_override_internal): Add "tremont".
11833 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
11834 (processor_type): Add PROCESSOR_TREMONT.
11835 * config/i386/x86-tune.def: Add m_TREMONT.
11836 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
11838 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11840 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
11841 symbol defined for msp430i* devices to be lower case.
11843 2018-06-07 Richard Biener <rguenther@suse.de>
11845 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
11846 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
11847 Properly wrap signed arithmetic if overflow wraps.
11849 2018-06-07 Jakub Jelinek <jakub@redhat.com>
11851 PR tree-optimization/69615
11852 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
11853 of a cast from a same precision integral SSA_NAME in a bb dominated
11854 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
11855 cast to utype if rhs2 has already a compatible type.
11857 2018-06-07 Richard Biener <rguenther@suse.de>
11859 PR tree-optimization/85935
11860 * graphite-scop-detection.c (find_params_in_bb): Analyze
11861 condition operands with respect to the correct loop. Assert
11862 the analysis doesn't fail.
11864 2018-06-04 Carl Love <cel@us.ibm.com>
11866 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
11867 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
11868 as it is slightly cheaper.
11869 (first_match_or_eos_index_<mode>):
11870 Calculate index using natural element order.
11871 (first_match_index_<mode>):
11872 Calculate index using natural element order.
11873 (first_match_or_eos_index_<mode>):
11874 Calculate index using natural order.
11875 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
11876 for BE and LE modes.
11877 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
11878 P9V_BUILTIN_VCLZLSBB_V16QI.
11879 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
11882 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11884 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
11885 indentation and line wrap for many prototypes. Add missing
11886 @smallexample directives around block of prototypes for vec_xl and
11889 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
11891 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
11892 track if we pass or return IEEE 128-bit floating point.
11893 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
11894 C++ mangling that is compatible with GCC 8.1.
11895 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
11896 (init_cumulative_args): Note if we pass or return IEEE 128-bit
11897 floating point types.
11898 (rs6000_function_arg_advance_1): Likewise.
11899 (rs6000_mangle_type): Optionally generate mangled names that match
11900 what GCC 8.1 generated for IEEE 128-bit floating point types.
11901 (rs6000_globalize_decl_name): If we have an external function that
11902 passes or returns IEEE 128-bit types, generate a weak reference
11903 from the mangled name used in GCC 8.1 to the current mangled
11905 (rs6000_init_builtins): Make __ibm128 use the long double type if
11906 long double is IBM extended double. Make __float128 use the long
11907 double type if long double is IEEE 128-bit.
11910 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
11911 macro for __ibm128 built-in functions.
11912 (PACK_IF): Add __ibm128 pack/unpack functions.
11913 (UNPACK_IF): Likewise.
11914 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
11915 enable long double built-in functions if long double is IEEE
11916 128-bit floating point.
11917 (rs6000_invalid_builtin): Update long double built-in function
11919 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
11920 functions, adjust the built-in function to use the long double
11921 built-in function if __ibm128 and long double are the same type.
11922 * doc/extend.texi (PowerPC builtins): Update documention for
11923 __builtin_{,un}pack_longdouble. Add documentation for
11924 __builtin_{,un}pack_ibm128.
11926 2018-06-06 Jim Wilson <jimw@sifive.com>
11928 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
11929 (struct machine_function): New field interrupt_mode.
11930 (riscv_handle_type_attribute): New function. Add forward declaration.
11931 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
11932 (riscv_expand_epilogue): Check interrupt_mode field.
11933 (riscv_set_current_function): Check interrupt attribute args and
11934 set interrupt_mode field.
11935 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
11936 (riscv_sret, riscv_uret): New.
11937 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
11938 new arguments to interrupt attribute.
11940 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
11943 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
11944 Don't handle -mcpu=power8 if -mpower9-vector is also used.
11946 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11948 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
11949 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
11950 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11951 several redundant entries.
11953 2018-06-06 David Malcolm <dmalcolm@redhat.com>
11955 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
11956 type from "rtx" to "rtx_insn *".
11957 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
11958 for local "call_insn", removing cast.
11959 (ix86_expand_call): Likewise, introducing a "call_insn" local.
11961 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
11963 PR tree-optimization/86066
11964 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
11965 for BIT_INSERT_EXPR stores.
11967 2018-06-06 Richard Biener <rguenther@suse.de>
11969 PR tree-optimization/86062
11970 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11971 component refs ontop
11972 of to be offsetted base.
11974 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11976 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
11977 to be static and remove check on interrupt attribute name.
11979 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
11981 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11982 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
11984 2018-06-05 Steve Ellcey <sellcey@cavium.com>
11987 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
11989 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
11990 Remove second argument, change how error is called.
11991 (aarch64_layout_arg): Remove second argument from
11992 aarch64_err_no_fpadvsimd call.
11993 (aarch64_init_cumulative_args): Ditto.
11994 (aarch64_gimplify_va_arg_expr): Ditto.
11995 * config/aarch64/aarch64.md (mov<mode>): Ditto.
11997 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
11999 * config/i386/i386.md (simple_return_indirect_internal): New expander.
12000 (*simple_return_indirect_internal<mode>): Rename from
12001 simple_return_indirect_internal. Use W mode iterator.
12002 (rstorssp): New expander.
12003 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
12004 (clrssbsy): New expander.
12005 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
12007 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
12009 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12011 (cmse_check_pointed_object): Likewise.
12013 2018-06-05 Martin Liska <mliska@suse.cz>
12015 PR gcov-profile/47618
12016 * doc/invoke.texi: Document how -fprofile-dir format
12019 2018-06-05 Richard Biener <rguenther@suse.de>
12021 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12022 removal pretend DOM info isn't available so we do not update
12023 it and only remove edges, not dominated blocks. Actually free
12024 DOM info in case we removed something. Remove unreachable blocks.
12025 (mfb_keep_latches): Work with either DOM info or marked backedges.
12026 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12027 first. Mark backedges if DOM info isn't available.
12028 (Re-)compute DOM info after cleanup_control_flow_pre.
12030 2018-06-05 Richard Biener <rguenther@suse.de>
12032 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12033 (locus_discrim_hasher::hash): Adjust.
12034 (locus_discrim_hasher::equal): Likewise.
12035 (next_discriminator_for_locus): Work on line directly.
12036 (same_line_p): Pass in expanded locus1 as well.
12037 (assign_discriminators): Avoid redundant location expansions.
12039 2018-06-05 Richard Biener <rguenther@suse.de>
12041 PR tree-optimization/86046
12042 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12043 if required after clearing TREE_ADDRESSABLE.
12045 2018-06-05 Richard Biener <rguenther@suse.de>
12047 PR tree-optimization/86047
12048 * tree-ssa-loop.c (for_each_index): Glob handling of all
12049 decls and constants and really handle all of them.
12051 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12054 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12055 qualifier_void_pointer and qualifier_const_void_pointer.
12056 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12057 (arm_init_builtins): Handle the above.
12058 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12059 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12060 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12063 2018-06-05 Martin Liska <mliska@suse.cz>
12065 * auto-profile.c (read_autofdo_file): Do not use
12066 gcov_ctr_summary struct.
12067 (afdo_callsite_hot_enough_for_early_inline): Likewise.
12068 * coverage.c (struct counts_entry): Likewise.
12069 (read_counts_file): Read just single summary entry.
12070 (get_coverage_counts): Use gcov_summary struct.
12071 * coverage.h (get_coverage_counts): Likewise.
12072 * gcov-dump.c (dump_working_sets): Likewise.
12073 (tag_summary): Dump just single summary.
12074 * gcov-io.c (gcov_write_summary): Write just histogram
12076 (gcov_read_summary): Read just single summary.
12077 (compute_working_sets): Use gcov_summary struct.
12078 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12079 of GCOV_COUNTERS_SUMMABLE.
12080 (GCOV_COUNTERS_SUMMABLE): Remove.
12081 (GCOV_FIRST_VALUE_COUNTER): Replace with
12082 GCOV_COUNTER_V_INTERVAL.
12083 (struct gcov_ctr_summary): Remove.
12084 (struct gcov_summary): Directly use fields of former
12086 (compute_working_sets): Use gcov_summary struct.
12087 * gcov.c (read_count_file): Do not use ctrs fields.
12088 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12090 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12092 * profile.c: Likewise.
12093 * profile.h: Likewise.
12095 2018-06-05 Martin Liska <mliska@suse.cz>
12097 PR gcov-profile/84846
12098 * gcov.c (output_lines): Print working directory only
12099 in intermediate format.
12101 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
12103 * config/s390/s390-builtin-types.def: Add void function type.
12104 * config/s390/s390-builtins.def: Use the function type for the
12107 2018-06-04 Jim Wilson <jimw@sifive.com>
12109 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12111 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12112 and maybe_eh_return. Change regno to unsigned int. Use new args to
12113 handle EH_RETURN_DATA_REGNO registers properly.
12114 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12115 (riscv_expand_epilogue): Update comment. Change argument name and
12116 type. Update code to use new name and type. Pass new args to
12117 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
12119 * config/riscv/riscv.md (NORMAL_RETURN): New.
12120 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12121 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12122 (eh_return): Call gen_eh_return_internal and emit barrier.
12123 (eh_return_internal): Call riscv_expand_epilogue.
12125 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
12127 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12128 bit_insertion field and declare can_be_merged_into method.
12129 (merged_store_group::can_be_merged_into): New method.
12130 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12131 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
12132 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12134 2018-06-04 Richard Biener <rguenther@suse.de>
12136 PR tree-optimization/85955
12137 * builtins.c (fold_builtin_sincos): Convert pointers to
12138 destination to appropriate type before dereferencing.
12140 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
12142 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12144 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
12146 * expr.c (expand_expr_real_1): Force the operand into memory if
12147 its TYPE_MODE is BLKmode and if there is no integer mode for
12148 the number of bits being extracted.
12150 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12154 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12155 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12157 2018-06-04 Richard Biener <rguenther@suse.de>
12159 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12160 (cleanup_tree_cfg_noloop): ... single caller. Do
12161 start_recording_case_labels later.
12163 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
12165 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12166 to _IMMINTRIN_H_INCLUDED.
12167 * config/i386/pconfigintrin.h: Ditto.
12168 * config/i386/waitpkgintrin.h: Ditto.
12169 * config/i386/immintrin.h: Add includes for sgxintrin.h,
12170 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12171 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12172 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12173 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12174 waitpkgintrin.h and cldemoteintrin.h.
12176 2018-06-04 Richard Biener <rguenther@suse.de>
12178 PR tree-optimization/86038
12179 * tracer.c (find_best_successor): Check probability for
12180 being initialized, bail out if not.
12182 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
12185 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12186 of bits to ignore when comparing architectures.
12188 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12190 PR tree-optimization/69615
12191 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12192 maximum or minimum of the type, try to merge it also as if
12193 range1 is + [-, x - 1] or + [x + 1, -].
12196 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12198 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
12200 PR tree-optimization/86034
12201 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12202 the unsigned bitfield type in a bit insertion sequence if it does not
12203 have a larger precision than the bitfield size.
12204 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12206 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
12208 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12210 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12212 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12213 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12214 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12215 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12217 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12219 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12220 Disable -fdelete-null-pointer-checks for ELF toolchain.
12222 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12223 Kito Cheng <kito.cheng@gmail.com>
12225 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12226 (nds32le-*-*, nds32be-*-*): Integrate checking process.
12227 (nds32*-*-*): Add glibc and uclibc conditions.
12228 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12229 (TARGET_EXCEPT_UNWIND_INFO): Define.
12230 * config/nds32/elf.h: New file.
12231 * config/nds32/linux.h: New file.
12232 * config/nds32/nds32-elf.opt: New file.
12233 * config/nds32/nds32-linux.opt: New file.
12234 * config/nds32/nds32-fp-as-gp.c
12235 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12236 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12238 (nds32_asm_file_end): Ditto.
12239 (nds32_print_operand): Ditto.
12240 (nds32_insert_attributes): Ditto.
12241 (nds32_init_libfuncs): New function.
12242 (TARGET_HAVE_TLS): Define.
12243 (TARGET_INIT_LIBFUNCS): Define.
12244 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12246 (TARGET_ELF): Apply different mcmodel setting.
12247 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12248 been migrated into elf.h and linux.h files.
12249 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12250 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12251 (mcmodel): The content has been migrated into nds32-elf.opt and
12252 nds32-linux.opt files.
12253 * config/nds32/t-elf: New file.
12254 * config/nds32/t-linux: New file.
12256 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12257 Shiva Chen <shiva0217@gmail.com>
12259 * config/nds32/constants.md (unspec_volatile_element): Add
12260 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12261 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12263 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12264 (make_pass_nds32_fp_as_gp): Declare.
12265 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12267 (nds32_asm_function_end_prologue): Remove unused asm output.
12268 (nds32_asm_function_begin_epilogue): Remove unused asm output.
12269 (nds32_asm_file_start): Output necessary fp_as_gp information.
12270 (nds32_option_override): Adjust register usage.
12271 (nds32_expand_prologue): Consider fp_as_gp situation.
12272 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12273 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12276 (simple_return): Ditto.
12277 (omit_fp_begin): Output special directive for fp_as_gp.
12278 (omit_fp_end): Output special directive for fp_as_gp.
12279 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12280 mforbid-fp-as-gp): New options.
12282 2018-06-01 Mark Wielaard <mark@klomp.org>
12284 * dwarf2out.c (dwarf2out_finish): Remove generation of
12285 DW_AT_loclists_base.
12287 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
12289 * gimple-ssa-store-merging.c: Include gimple-fold.h.
12290 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12291 (struct merged_store_group): Add bit_insertion field.
12292 (dump_char_array): Use standard hexadecimal format.
12293 (merged_store_group::merged_store_group): Set bit_insertion to false.
12294 (merged_store_group::apply_stores): Use optimal buffer size. Deal
12295 with BIT_INSERT_EXPR stores. Move up code updating the mask and
12296 also print the mask in the dump file.
12297 (pass_store_merging::gate): Minor tweak.
12298 (imm_store_chain_info::coalesce_immediate): Fix wrong association
12299 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
12300 stores with INTEGER_CST stores.
12301 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12302 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12303 and use it throughout. Generate bit insertion sequences if need be.
12304 (pass_store_merging::process_store): Remove redundant condition.
12305 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12307 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
12309 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12310 the 128-bit floating point types. Fix function comment.
12312 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12314 * config/aarch64/aarch64-simd.md
12315 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12317 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12320 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
12322 PR tree-optimization/85989
12323 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12325 (backprop::intersect_uses): Check it when deciding whether this
12326 is a backedge reference.
12327 (backprop::process_block): Add each phi to m_visited_phis
12328 after visiting it, then clear it at the end.
12330 2018-06-01 Richard Biener <rguenther@suse.de>
12332 * tree-vectorizer.h (vect_dr_stmt): New function.
12333 (vect_get_load_cost): Adjust.
12334 (vect_get_store_cost): Likewise.
12335 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
12336 Use vect_dr_stmt instead of DR_SMTT.
12337 (vect_record_base_alignments): Likewise.
12338 (vect_calculate_target_alignment): Likewise.
12339 (vect_compute_data_ref_alignment): Likewise and make static.
12340 (vect_update_misalignment_for_peel): Likewise.
12341 (vect_verify_datarefs_alignment): Likewise.
12342 (vector_alignment_reachable_p): Likewise.
12343 (vect_get_data_access_cost): Likewise. Pass down
12344 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
12345 (vect_get_peeling_costs_all_drs): Likewise.
12346 (vect_peeling_hash_get_lowest_cost): Likewise.
12347 (vect_enhance_data_refs_alignment): Likewise.
12348 (vect_find_same_alignment_drs): Likewise.
12349 (vect_analyze_data_refs_alignment): Likewise.
12350 (vect_analyze_group_access_1): Likewise.
12351 (vect_analyze_group_access): Likewise.
12352 (vect_analyze_data_ref_access): Likewise.
12353 (vect_analyze_data_ref_accesses): Likewise.
12354 (vect_vfa_segment_size): Likewise.
12355 (vect_small_gap_p): Likewise.
12356 (vectorizable_with_step_bound_p): Likewise.
12357 (vect_prune_runtime_alias_test_list): Likewise.
12358 (vect_analyze_data_refs): Likewise.
12359 (vect_supportable_dr_alignment): Likewise.
12360 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12361 (vect_gen_prolog_loop_niters): Likewise.
12362 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12363 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
12365 (vect_recog_mask_conversion_pattern): Likewise.
12366 (vect_try_gather_scatter_pattern): Likewise.
12367 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
12368 to vect_get_store_cost.
12369 (vect_get_store_cost): Get stmt_info instead of DR.
12370 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
12371 (vect_get_load_cost): Get stmt_info instead of DR.
12373 2018-06-01 Richard Biener <rguenther@suse.de>
12375 PR middle-end/86017
12376 * gimple-fold.c (var_decl_component_p): Also allow offsetted
12377 vars wrapped in MEM_REFs.
12379 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
12381 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12382 Fix subreg tests so that we only return a choice between
12383 GENERAL_REGS and FP_REGS if the original classes included both.
12385 2018-06-01 Richard Biener <rguenther@suse.de>
12388 * tree-ssa-structalias.c (get_function_part_constraint):
12389 Handle NULL fi->decl.
12390 (find_func_aliases_for_call): Properly handle indirect
12391 fi from direct call.
12392 (find_func_clobbers): Likewise.
12393 (ipa_pta_execute): Likewise.
12394 (create_variable_info_for): For functions that are ifunc_resolver
12395 resolve to a varinfo that contains the result of the resolver call.
12396 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
12399 2018-05-31 Michael Collison <michael.collison@arm.com>
12401 * config/aarch64/aarch64.md:
12402 (*fix_to_zero_extenddfdi2): New pattern.
12403 * gcc.target/aarch64/fix_extend1.c: New testcase.
12405 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
12407 PR middle-end/78809
12408 PR middle-end/83026
12409 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
12410 and BUILT_IN_STRNCMP_EQ.
12411 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
12412 BUILT_IN_STRNCMP_EQ.
12413 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
12414 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12415 (gimple_fold_builtin): Likewise.
12416 * tree-ssa-strlen.c (compute_string_length): New function.
12417 (determine_min_obsize): New function.
12418 (handle_builtin_string_cmp): New function to handle calls to
12419 string compare functions.
12420 (strlen_optimize_stmt): Add handling to builtin string compare
12422 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
12423 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12424 * tree.c (build_common_builtin_nodes): Add new defines of
12425 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
12427 2018-05-31 Jakub Jelinek <jakub@redhat.com>
12430 * bb-reorder.c (pass_partition_blocks::gate): Return false for
12431 functions with naked attribute.
12433 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
12435 * config/i386/sse.md (avx_vec_concat<mode>):
12436 Substitute concat_tg_mode mode attribute with xtg_mode.
12437 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
12438 (concat_tg_mode): Remove mode attribute.
12440 2018-05-31 Martin Sebor <msebor@redhat.com>
12443 * calls.c (alloc_max_size): Correct a logic error/typo.
12444 Treat excessive arguments as infinite. Warn for invalid arguments.
12445 * doc/invoke.texi (-Walloc-size-larger-than): Update.
12447 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
12450 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
12451 and movx for Haswell.
12453 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
12454 Cesar Philippidis <cesar@codesourcery.com>
12456 PR middle-end/85879
12457 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
12458 when emitting error on private/firstprivate reductions.
12459 * omp-low.c (lower_omp_target): Avoid reference-type processing
12460 on pointers for firstprivate clause.
12462 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12464 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
12467 * config/aarch64/aarch64-simd.md
12468 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
12469 (aarch64_ld1_x3_<mode>): Likewise
12470 (aarch64_st1x2<VALLDIF:mode>): Likewise
12471 (aarch64_st1_x2_<mode>): Likewise
12472 (aarch64_st1x3<VALLDIF:mode>): Likewise
12473 (aarch64_st1_x3_<mode>): Likewise
12474 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
12475 (vld1_s8_x3): Likewise.
12476 (vld1_u16_x3): Likewise.
12477 (vld1_s16_x3): Likewise.
12478 (vld1_u32_x3): Likewise.
12479 (vld1_s32_x3): Likewise.
12480 (vld1_u64_x3): Likewise.
12481 (vld1_s64_x3): Likewise.
12482 (vld1_f16_x3): Likewise.
12483 (vld1_f32_x3): Likewise.
12484 (vld1_f64_x3): Likewise.
12485 (vld1_p8_x3): Likewise.
12486 (vld1_p16_x3): Likewise.
12487 (vld1_p64_x3): Likewise.
12488 (vld1q_u8_x3): Likewise.
12489 (vld1q_s8_x3): Likewise.
12490 (vld1q_u16_x3): Likewise.
12491 (vld1q_s16_x3): Likewise.
12492 (vld1q_u32_x3): Likewise.
12493 (vld1q_s32_x3): Likewise.
12494 (vld1q_u64_x3): Likewise.
12495 (vld1q_s64_x3): Likewise.
12496 (vld1q_f16_x3): Likewise.
12497 (vld1q_f32_x3): Likewise.
12498 (vld1q_f64_x3): Likewise.
12499 (vld1q_p8_x3): Likewise.
12500 (vld1q_p16_x3): Likewise.
12501 (vld1q_p64_x3): Likewise.
12502 (vst1_s64_x2): Likewise.
12503 (vst1_u64_x2): Likewise.
12504 (vst1_f64_x2): Likewise.
12505 (vst1_s8_x2): Likewise.
12506 (vst1_p8_x2): Likewise.
12507 (vst1_s16_x2): Likewise.
12508 (vst1_p16_x2): Likewise.
12509 (vst1_s32_x2): Likewise.
12510 (vst1_u8_x2): Likewise.
12511 (vst1_u16_x2): Likewise.
12512 (vst1_u32_x2): Likewise.
12513 (vst1_f16_x2): Likewise.
12514 (vst1_f32_x2): Likewise.
12515 (vst1_p64_x2): Likewise.
12516 (vst1q_s8_x2): Likewise.
12517 (vst1q_p8_x2): Likewise.
12518 (vst1q_s16_x2): Likewise.
12519 (vst1q_p16_x2): Likewise.
12520 (vst1q_s32_x2): Likewise.
12521 (vst1q_s64_x2): Likewise.
12522 (vst1q_u8_x2): Likewise.
12523 (vst1q_u16_x2): Likewise.
12524 (vst1q_u32_x2): Likewise.
12525 (vst1q_u64_x2): Likewise.
12526 (vst1q_f16_x2): Likewise.
12527 (vst1q_f32_x2): Likewise.
12528 (vst1q_f64_x2): Likewise.
12529 (vst1q_p64_x2): Likewise.
12530 (vst1_s64_x3): Likewise.
12531 (vst1_u64_x3): Likewise.
12532 (vst1_f64_x3): Likewise.
12533 (vst1_s8_x3): Likewise.
12534 (vst1_p8_x3): Likewise.
12535 (vst1_s16_x3): Likewise.
12536 (vst1_p16_x3): Likewise.
12537 (vst1_s32_x3): Likewise.
12538 (vst1_u8_x3): Likewise.
12539 (vst1_u16_x3): Likewise.
12540 (vst1_u32_x3): Likewise.
12541 (vst1_f16_x3): Likewise.
12542 (vst1_f32_x3): Likewise.
12543 (vst1_p64_x3): Likewise.
12544 (vst1q_s8_x3): Likewise.
12545 (vst1q_p8_x3): Likewise.
12546 (vst1q_s16_x3): Likewise.
12547 (vst1q_p16_x3): Likewise.
12548 (vst1q_s32_x3): Likewise.
12549 (vst1q_s64_x3): Likewise.
12550 (vst1q_u8_x3): Likewise.
12551 (vst1q_u16_x3): Likewise.
12552 (vst1q_u32_x3): Likewise.
12553 (vst1q_u64_x3): Likewise.
12554 (vst1q_f16_x3): Likewise.
12555 (vst1q_f32_x3): Likewise.
12556 (vst1q_f64_x3): Likewise.
12557 (vst1q_p64_x3): Likewise.
12559 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12561 * config/msp430/msp430.c (msp430_output_labelref): Prepend
12562 user_label_prefix to name.
12564 * tree-core.h: Update comment about the format of NAME string
12565 passed to handler in attribute_spec.
12567 * config/msp430/msp430.md: Remove erroneous subreg expression from
12568 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
12569 zero_extend{q,h}isi2.
12571 2018-05-30 Borislav Petkov <bp@suse.de>
12573 * doc/extend.texi: Document some architecture specific
12574 constraints and sort entries.
12576 2018-05-30 Martin Sebor <msebor@redhat.com>
12578 PR middle-end/85369
12579 * builtins.c (expand_builtin_stpcpy_1): New function.
12580 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
12581 only if the former succeeds.
12583 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12585 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
12588 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12590 * doc/invoke.texi (-flinker-output): Document
12592 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12594 * passes.c (ipa_write_summaries): Only modify statements if body
12596 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
12597 incrementally linking.
12598 (ipa_passes): Likewise.
12599 * lto-cgraph.c (lto_output_node): When incrementally linking do not
12600 pass down resolution info.
12601 * common.opt (flag_incremental_link): Update info.
12602 * gcc.c (plugin specs): Turn flinker-output=* to
12603 -plugin-opt=-linker-output-known
12604 * toplev.c (compile_file): Also cut compilation when doing incremental
12606 * flag-types. (enum lto_partition_model): Add
12607 LTO_LINKER_OUTPUT_NOLTOREL.
12608 (invoke.texi): Add -flinker-output docs.
12609 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
12610 link same way as WPA; do not stream in dead initializers.
12612 * dwarf2out.c (dwarf2out_die_ref_for_decl,
12613 darf2out_register_external_decl): Support incremental link.
12615 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12617 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
12619 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12621 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
12622 it down to simple_object_copy_lto_debug_sections.
12623 (run_gcc): Determine incremental LTO link time and configure
12624 lto1 into non-wpa mode, disable renaming of debug sections.
12626 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
12628 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12629 descriptions of various incorrectly documented functions.
12631 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
12634 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12635 address check not strict.
12637 2018-05-30 Richard Biener <rguenther@suse.de>
12639 PR tree-optimization/85964
12640 * tracer.c (better_p): Drop initialized count check, we only
12641 call the function with initialized counts now.
12642 (find_best_successor): Do find a best edge if one
12643 has uninitialized count.
12644 (find_best_predecessor): Likewise. Do BB frequency check only
12645 if count is initialized.
12647 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
12649 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
12650 (aarch64_ldrstr_offset_compare): New.
12651 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
12652 load/store orderings.
12653 (aarch64_gen_adjusted_ldpstp): Likewise.
12655 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
12657 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12658 Check for subset of GENERAL_REGS and FP_REGS.
12659 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
12662 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
12664 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
12665 and wi::to_poly_offset. Add the current offset and then check
12666 whether the sum fits, rather than using an unchecked addition of
12667 a checked term. Check for a shwi rather than a uhwi.
12668 * expr.c (get_bit_range): Use tree_to_poly_uint64.
12669 (store_constructor): Use poly_int_tree_p.
12670 (expand_expr_real_1): Likewise.
12671 * function.c (assign_temp): Likewise.
12672 * fold-const.c (const_binop): Use poly_int_tree_p and
12673 wi::to_poly_offset.
12674 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
12676 * ipa-icf-gimple.c (func_checker::compare_operand): Use
12677 to_poly_offset for MEM offsets.
12678 * ipa-icf.c (sem_variable::equals): Likewise.
12679 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
12680 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
12681 wi::to_poly_offset for BIT_FIELD_REF offsets.
12682 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
12683 wi::to_poly_offset.
12684 * var-tracking.c (emit_note_insn_var_location): Use
12685 tree_to_poly_uint64.
12687 2018-05-29 Jim Wilson <jimw@sifive.com>
12689 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
12691 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
12694 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
12695 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
12697 (sse4_1_round<mode>2): Use nonimmediate_operand
12698 for operand 1 predicate.
12700 2018-05-29 Martin Sebor <msebor@redhat.com>
12701 Richard Biener <rguenther@suse.de>
12704 * calls.c (get_size_range): Call determine_value_range instead
12705 of get_value_range..
12706 * tree-vrp.h (determine_value_range): Declared new function.
12707 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
12709 2018-05-29 Richard Biener <rguenther@suse.de>
12711 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
12712 sure to use non-pattern stmts for get_earlier_stmt arguments.
12713 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
12714 called on pattern stmts.
12715 (get_later_stmt): Likewise.
12717 2018-05-29 Martin Liska <mliska@suse.cz>
12719 PR gcov-profile/85759
12720 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
12723 2018-05-29 Jakub Jelinek <jakub@redhat.com>
12725 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
12727 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
12731 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12732 VEC_PACK_FLOAT_EXPR): New tree codes.
12733 * tree-pretty-print.c (op_code_prio): Handle
12734 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12735 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12736 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12737 * tree-inline.c (estimate_operator_cost): Likewise.
12738 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
12739 * fold-const.c (const_binop): Likewise.
12740 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
12741 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12742 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12743 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
12744 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12745 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12746 * expr.c (expand_expr_real_2): Likewise.
12747 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
12748 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
12749 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
12751 * optabs.c (expand_widen_pattern_expr): For
12752 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
12753 sign from result type rather than operand's type.
12754 (expand_binop_directly): For vec_packu_float_optab and
12755 vec_packs_float_optab allow result type to be different from operand's
12757 * optabs-tree.c (optab_for_tree_code): Handle
12758 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12759 VEC_PACK_FLOAT_EXPR. Formatting fixes.
12760 * tree-vect-generic.c (expand_vector_operations_1): Handle
12761 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12762 VEC_PACK_FLOAT_EXPR.
12763 * tree-vect-stmts.c (supportable_widening_operation): Handle
12765 (supportable_narrowing_operation): Handle FLOAT_EXPR.
12766 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
12767 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
12768 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
12769 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
12771 (vec_pack<floatprefix>_float_<mode>): New expander.
12772 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
12774 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
12775 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
12776 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
12777 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
12778 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
12780 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
12781 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
12782 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12783 VEC_PACK_FLOAT_EXPR): Document.
12785 2018-05-29 Richard Biener <rguenther@suse.de>
12787 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
12789 (stmt_vec_info_vec): Make pointer.
12790 (init_stmt_vec_info_vec): Remove.
12791 (free_stmt_vec_info_vec): Likewise.
12792 (set_stmt_vec_info_vec): New function.
12793 (free_stmt_vec_infos): Likewise.
12794 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
12795 (set_vinfo_for_stmt): Likewise.
12796 (get_earlier_stmt): Likewise.
12797 (get_later_stmt): Likewise.
12798 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
12799 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
12800 (vec_info::~vec_info): Free stmt_vec_infos.
12801 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
12802 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
12803 (pass_slp_vectorize::execute): Likewise.
12804 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
12805 (free_stmt_vec_info_vec): Likewise.
12806 (set_stmt_vec_info_vec): New function.
12807 (free_stmt_vec_infos): Likewise.
12808 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
12809 the global stmt_vec_info_vec.
12810 * tree-parloops.c (gather_scalar_reductions): Use
12811 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
12814 2018-05-29 Richard Biener <rguenther@suse.de>
12816 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
12818 2018-05-29 Martin Liska <mliska@suse.cz>
12819 David Malcolm <dmalcolm@redhat.com>
12821 * vec.c (test_reverse): New.
12822 (vec_c_tests): Add new test.
12823 * vec.h (vl_ptr>::reverse): New function.
12825 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
12827 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
12829 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
12832 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
12834 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
12836 2018-05-28 Richard Biener <rguenther@suse.de>
12838 PR tree-optimization/85933
12839 * tree-vect-data-refs.c (vect_record_base_alignments): Only
12840 look at stmts marked as vectorizable.
12842 2018-05-28 Richard Biener <rguenther@suse.de>
12844 PR tree-optimization/85934
12845 * tree-vect-generic.c (expand_vector_operations_1): Hoist
12846 vector boolean check before scalar optimization.
12848 2018-05-28 Jakub Jelinek <jakub@redhat.com>
12850 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
12853 2018-05-28 Mark Wielaard <mark@klomp.org>
12855 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
12856 if it is an expression containing a minus sign.
12858 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
12860 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
12862 2018-05-27 Paul Koning <ni1d@arrl.net>
12864 * config/pdp11/pdp11.md (truncsihi2): Remove.
12866 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
12867 Chung-Ju Wu <jasonwucj@gmail.com>
12869 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
12871 (unaligned_store_dw): Ditto.
12872 * config/nds32/nds32-memory-manipulation.c
12873 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
12874 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
12875 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
12876 (emit_setmem_word_loop): Rename to ...
12877 (emit_setmem_doubleword_loop): ... this.
12878 (nds32_gen_dup_4_byte_to_word_value): New function.
12879 (nds32_gen_dup_8_byte_to_double_word_value): New function.
12880 (nds32_expand_setmem_loop): Refine implementation.
12881 (nds32_expand_setmem_loop_v3m): Ditto.
12882 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
12885 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12887 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
12889 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12891 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
12892 (nds32_init_machine_status): Initialize machine->attr_naked_p and
12893 machine->attr_no_prologue_p.
12894 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
12895 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
12896 (nds32_expand_epilogue): Consider attr_naked_p.
12897 (nds32_expand_epilogue_v3pop): Likewise.
12898 (nds32_can_use_return_insn): Likewise.
12899 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
12900 attr_no_prologue_p fields.
12901 * config/nds32/nds32.opt (mret-in-naked-func): New option.
12903 2018-05-27 Jakub Jelinek <jakub@redhat.com>
12906 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
12908 * config/i386/sse.md
12909 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
12911 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
12913 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
12915 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
12917 (*<floatsuffix>floatv2div2sf2): Rename to ...
12918 (*float<floatunssuffix>v2div2sf2): ... this.
12919 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
12920 (float<floatunssuffix>v2div2sf2_mask): ... this.
12921 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
12922 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
12923 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
12925 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
12927 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12929 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12931 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12933 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12935 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
12936 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
12937 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
12938 gen_ufix_truncv8dfv8si2.
12939 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
12940 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
12941 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
12942 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
12943 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
12944 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
12945 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
12946 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
12948 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12952 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
12954 2018-05-25 Jim Wilson <jimw@sifive.com>
12956 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
12957 * config/riscv/riscv.c (struct machine_function): Add
12958 interrupt_handler_p and attribute_checked_p fields.
12959 (riscv_attribute_table): Add interrupt.
12960 (riscv_interrupt_type_p): New.
12961 (riscv_save_reg_p): Save extra regs for interrupt handler.
12962 (riscv_use_save_libcall): Return false for interrupt handler.
12963 (riscv_first_stack_step): Add forward declaration.
12964 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
12965 for interrupt handler with large frame. Use it for saved reg list.
12966 (riscv_expand_prologue): Move flag_stack_usage_info support to
12967 eliminate duplication.
12968 (riscv_expand_epilogue): Generate mret for interrupt handler.
12969 (riscv_epilogue_uses): New.
12970 (riscv_can_use_return_insn): Return false for interrupt handler.
12971 (riscv_function_ok_for_sibcall): Likewise.
12972 (riscv_set_current_function): Add interrupt handler support.
12973 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
12974 * config/riscv/riscv.md (UNSPECV_MRET): New.
12976 (riscv_frflags, riscv_fsflags): Use tab after opcode.
12978 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
12980 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
12982 PR tree-optimization/85712
12983 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
12984 this candidate has already been replaced in-situ by a copy.
12986 2018-05-25 Jason Merrill <jason@redhat.com>
12988 PR c++/80485 - inline function non-zero address.
12989 * symtab.c (nonzero_address): Check DECL_COMDAT.
12991 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
12994 * config/alpha/alpha.md (ashlsi3): New insn pattern.
12995 (*ashlsi_se): Rename from *ashldi_se. Define as sign
12996 extension of SImode operation. Use const123_operand predicate.
12997 (*saddsi_1): Remove.
12998 (*saddl_se_1): Ditto.
12999 (*ssubsi_1): Ditto.
13000 (*ssubl_se_1): Ditto.
13001 * config/alpha/predicates.md (const123_operand): New predicate.
13002 * config/alpha/constraints.md (P): Use IN_RANGE.
13004 2018-05-25 Richard Biener <rguenther@suse.de>
13006 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13008 (ref_maybe_used_by_stmt_p): Likewise.
13009 (stmt_may_clobber_ref_p): Likewise.
13010 (stmt_may_clobber_ref_p_1): Likewise.
13011 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13013 (ref_maybe_used_by_stmt_p): Likewise.
13014 (stmt_may_clobber_ref_p): Likewise.
13015 (stmt_may_clobber_ref_p_1): Likewise.
13016 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13017 the alias oracle to disambiguate DRs with stmts DR analysis
13019 (vect_analyze_data_refs): Do not give up on not analyzable
13020 DRs for BB vectorization. Remove code truncating the dataref
13023 2018-05-25 Jakub Jelinek <jakub@redhat.com>
13026 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13027 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
13028 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13030 2018-05-25 Richard Biener <rguenther@suse.de>
13032 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13033 function, combining stmt data ref gathering and fatal analysis
13035 (vect_analyze_data_refs): Remove now redudnant code and simplify.
13036 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13037 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13038 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13039 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13041 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13043 PR tree-optimization/85720
13044 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13045 SCC if all partitions are builtins.
13046 (version_loop_by_alias_check): New parameter. Generate cancelable
13047 runtime alias check if all partitions are builtins.
13048 (distribute_loop): Update call to above function.
13050 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13052 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13053 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13054 (parm_default_def_partition_arg): Ditto.
13055 (set_parm_default_def_partition): Ditto.
13056 (get_parm_default_def_partitions): Ditto and make it static.
13057 (get_undefined_value_partitions): Ditto and make it static.
13058 (remove_ssa_form): Refactor call to init_var_map here.
13059 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13060 computation for loop region.
13061 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13062 (register_default_def): Delete.
13063 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13064 (parm_default_def_partition_arg): Ditto.
13065 (set_parm_default_def_partition): Ditto.
13066 (get_parm_default_def_partitions): Ditto and make it static.
13067 (get_undefined_value_partitions): Ditto and make it static.
13068 (coalesce_with_default, coalesce_with_default): Update comment.
13069 (create_coalesce_list_for_region): New func factored out from
13070 create_outofssa_var_map.
13071 (populate_coalesce_list_for_outofssa): New func factored out from
13072 create_outofssa_var_map and coalesce_ssa_name.
13073 (create_outofssa_var_map): Delete.
13074 (coalesce_ssa_name): Refactor to support live range computation.
13075 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13076 (get_parm_default_def_partitions): Delete.
13077 (get_undefined_value_partitions): Ditto.
13078 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13079 computation for loop region.
13080 (new_tree_live_info, loe_visit_block): Ditto.
13081 (live_worklist, set_var_live_on_entry): Ditto.
13082 (calculate_live_on_exit, verify_live_on_entry): Ditto.
13083 * tree-ssa-live.h (struct _var_map): New fields.
13084 (init_var_map): Change decl.
13085 (region_contains_p): New.
13087 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13089 * tree-ssa-live.h (live_merge_and_clear): Delete.
13091 2018-05-25 Richard Biener <rguenther@suse.de>
13094 * tree-dump.c (dequeue_and_dump): Remove access to removed
13095 operand 2 of a SWITCH_EXPR.
13097 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13099 * doc/sourcebuild.texi (vect_double_cond_arith): Include
13100 multiplication and division.
13101 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13102 (cond_udiv@var{m}, cond_umod@var{m}): Document.
13103 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13104 (cond_udiv_optab, cond_umod_optab): New optabs.
13105 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13106 (IFN_COND_RDIV): New internal functions.
13107 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13108 TRUNC_MOD_EXPR and RDIV_EXPR.
13109 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13110 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13112 (SVE_INT_BINARY): Include mult.
13113 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13114 (optab, sve_int_op): Handle mult.
13115 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13117 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13118 for SVE_INT_BINARY_SD.
13120 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13122 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13123 (optab, sve_int_op): Handle div and udiv.
13124 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13125 for SVE_INT_BINARY_SD.
13126 (*<optab><mode>3): New insn for the same.
13128 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13130 * tree-vect-patterns.c: Include predict.h.
13131 (vect_recog_divmod_pattern): Restrict check for division support
13132 to when optimizing for size.
13134 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13136 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13137 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13138 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13139 (gimple_match_op::set_op): Likewise.
13140 (gimple_resimplify4): Declare.
13141 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13142 (expr::gen_transform): Likewise.
13143 (decision_tree::gen): Generate a simplification routine for 4 operands.
13144 * gimple-match-head.c (gimple_simplify): Add an overload for
13145 4 operands. In the top-level function, handle up to 4 call
13146 arguments and call gimple_resimplify4.
13147 (gimple_resimplify4): New function.
13148 (build_call_internal): Pass a fourth operand.
13149 (maybe_push_to_seq): Likewise.
13150 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13151 Fold VEC_COND_EXPRs of an operation and a default value into
13152 an IFN_COND_* function if possible.
13153 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13155 (SVE_COND_FP_BINARY): Include them.
13156 (optab, sve_fp_op): Handle them.
13157 (SVE_INT_BINARY_REV): New code iterator.
13158 (SVE_COND_FP_BINARY_REV): New int iterator.
13159 (commutative): New int attribute.
13160 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13162 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13164 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13165 (*cond_<optab><mode>): New patterns for reversed operands.
13167 2018-05-25 Richard Biener <rguenther@suse.de>
13169 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13170 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13171 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13172 (STMT_VINFO_GROUPED_ACCESS): Adjust.
13173 * tree-vect-data-refs.c (everywhere): Adjust users.
13174 * tree-vect-loop.c (everywhere): Likewise.
13175 * tree-vect-slp.c (everywhere): Likewise.
13176 * tree-vect-stmts.c (everywhere): Likewise.
13177 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13179 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13181 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13183 (gcc_cv_as_section_exclude): ... this.
13184 Try Solaris as #exclude syntax.
13185 * configure: Regenerate.
13186 * config.in: Regenerate.
13187 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13189 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13190 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13192 * varasm.c (default_elf_asm_named_section): Don't check if
13193 HAVE_GAS_SECTION_EXCLUDE is defined.
13195 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13197 * doc/md.texi: Update the documentation of the cond_* optabs
13198 to mention the new final operand. Fix GET_MODE_NUNITS call.
13199 Describe the scalar case too.
13200 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13201 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13203 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13204 (get_conditional_internal_fn): Update comment.
13205 * tree-vect-loop.c (vectorizable_reduction): Pass the original
13206 accumulator value as a final argument to conditional functions.
13207 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13208 a define_expand and add an "else" operand. Assert for now that
13209 the else operand is equal to operand 2. Use SVE_INT_BINARY and
13210 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13211 (*cond_<optab><mode>): New patterns.
13212 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13213 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13214 (UNSPEC_COND_EOR): Delete.
13215 (optab): Remove associated mappings.
13216 (SVE_INT_BINARY): New code iterator.
13217 (sve_int_op): Remove int attribute and add "minus" to the code
13219 (SVE_COND_INT_OP): Delete.
13220 (SVE_COND_FP_OP): Rename to...
13221 (SVE_COND_FP_BINARY): ...this.
13223 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13225 * optabs.c (can_reuse_operands_p): New function.
13226 (maybe_legitimize_operands): Try to reuse the results for
13229 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13231 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13232 Add {q} suffix to insn mnemonic.
13234 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13236 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13237 (msp430_warn_func_return): New.
13239 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
13241 * fold-const.c (tree_nonzero_bits): New function.
13242 * fold-const.h (tree_nonzero_bits): Likewise.
13243 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13244 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13246 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13250 * varasm.c (assemble_alias): Check ifunc_resolver only on
13253 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13256 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13257 when memory input operand is handled.
13259 2018-05-24 Luis Machado <luis.machado@linaro.org>
13261 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13263 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13265 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13267 * match.pd: Delay FMA folds until after vectorization.
13269 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
13272 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13273 address check not strict.
13275 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13277 * gimple-match.h (gimple_match_op): New class.
13278 (mprts_hook): Replace parameters with a gimple_match_op *.
13279 (maybe_build_generic_op): Likewise.
13280 (gimple_simplified_result_is_gimple_val): Replace parameters with
13281 a const gimple_match_op *.
13282 (gimple_simplify): Replace code_helper * and tree * parameters with
13283 a gimple_match_op * parameter.
13284 (gimple_resimplify1): Replace code_helper *, tree and tree *
13285 parameters with a gimple_match_op * parameter.
13286 (gimple_resimplify2): Likewise.
13287 (gimple_resimplify3): Likewise.
13288 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13289 parameters with a gimple_match_op * parameter.
13290 * gimple-match-head.c (gimple_simplify): Change prototypes of
13291 auto-generated functions to take a gimple_match_op * instead of
13292 separate code_helper * and tree * parameters. Make the same
13293 change in the top-level overload and update calls to the
13294 gimple_resimplify routines. Update calls to the auto-generated
13295 functions and to maybe_push_res_to_seq in the publicly-facing
13296 operation-specific gimple_simplify overloads.
13297 (gimple_match_op::MAX_NUM_OPS): Define.
13298 (gimple_resimplify1): Replace rcode and ops with a single res_op
13299 parameter. Update call to gimple_simplify.
13300 (gimple_resimplify2): Likewise.
13301 (gimple_resimplify3): Likewise.
13302 (mprts_hook): Replace parameters with a gimple_match_op *.
13303 (maybe_build_generic_op): Likewise.
13304 (build_call_internal): Replace type, nargs and ops with
13305 a gimple_match_op *.
13306 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13307 with a single gimple_match_op *. Update calls to mprts_hook,
13308 build_call_internal and gimple_simplified_result_is_gimple_val.
13309 Factor out code that is common to the tree_code and combined_fn cases.
13310 * genmatch.c (expr::gen_transform): Replace tem_code and
13311 tem_ops with a gimple_match_op called tem_op. Update calls
13312 to the gimple_resimplify functions and maybe_push_res_to_seq.
13313 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13314 res_ops. Update call to the gimple_resimplify functions.
13315 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13316 (decision_tree::gen): Make the functions take a gimple_match_op *
13317 called res_op instead of separate res_code and res_ops parameters.
13318 Update call accordingly.
13319 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13320 and ops with a single res_op parameter. Update calls to
13321 maybe_build_generic_op and maybe_push_res_to_seq.
13322 (fold_stmt_1): Update calls to gimple_simplify and
13323 replace_stmt_with_simplification.
13324 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13325 and gimple_simplified_result_is_gimple_val.
13326 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13328 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13329 with a gimple_match_op *.
13330 (vn_nary_build_or_lookup): Likewise. Update call to
13331 vn_nary_build_or_lookup_1.
13332 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13333 gimple_match_op *. Update calls to the gimple_resimplify routines
13334 and to gimple_simplified_result_is_gimple_val.
13335 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
13336 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
13337 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
13338 (visit_nary_op): Likewise.
13339 (visit_reference_op_load): Likewise.
13341 2018-05-23 Luis Machado <luis.machado@linaro.org>
13343 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
13344 modifier for printing the step amount.
13346 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
13349 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
13352 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
13354 * doc/sourcebuild.texi (Endianness): New subsubsection.
13356 2018-05-23 Luis Machado <luis.machado@linaro.org>
13358 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13359 <prefetch_dynamic_strides>: New const bool field.
13360 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13361 prefetch_dynamic_strides.
13362 (exynosm1_prefetch_tune): Likewise.
13363 (thunderxt88_prefetch_tune): Likewise.
13364 (thunderx_prefetch_tune): Likewise.
13365 (thunderx2t99_prefetch_tune): Likewise.
13366 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
13368 (aarch64_override_options_internal): Update to set
13369 PARAM_PREFETCH_DYNAMIC_STRIDES.
13370 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13371 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13372 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13373 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13374 prefetch-dynamic-strides setting.
13376 2018-05-23 Luis Machado <luis.machado@linaro.org>
13378 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13379 <minimum_stride>: New const int field.
13380 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13381 minimum_stride field defaulting to -1.
13382 (exynosm1_prefetch_tune): Likewise.
13383 (thunderxt88_prefetch_tune): Likewise.
13384 (thunderx_prefetch_tune): Likewise.
13385 (thunderx2t99_prefetch_tune): Likewise.
13386 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
13387 <default_opt_level>: Set to 3.
13388 (aarch64_override_options_internal): Update to set
13389 PARAM_PREFETCH_MINIMUM_STRIDE.
13390 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13391 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13392 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13393 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13394 stride is constant and is below the minimum stride threshold.
13396 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13398 * config/arm/arm-cpus.in (mode26): Delete.
13399 (armv4): Delete mode26 reference.
13400 * config/arm/arm.c (arm_configure_build_target): Delete use of
13403 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
13405 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
13407 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
13408 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
13410 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
13411 (floatunsdidf2): Ditto.
13413 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
13415 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
13416 (fixuns_trunc<mode>si2_avx512f): Ditto.
13417 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
13418 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
13419 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
13421 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
13423 PR rtl-optimization/79985
13424 * df-scan.c (df_insn_refs_collect): Remove special case for
13425 global registers and asm statements.
13427 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
13429 * extend.texi (Global Register Variables): Rewrite the bullet list.
13430 Note that the register is available for allocation. Note that access
13431 via inline asm must use constraints. Add note about async-signal
13432 handlers. Remove paragraph about automagic register selection.
13434 2018-05-23 Richard Biener <rguenther@suse.de>
13436 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
13437 of fixed offset from memset VN.
13439 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
13441 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
13442 first_interp field.
13443 (alloc_cand_and_find_basis): Initialize first_interp field.
13444 (slsr_process_mul): Modify first_interp field.
13445 (slsr_process_add): Likewise.
13446 (slsr_process_cast): Modify first_interp field for each new
13448 (slsr_process_copy): Likewise.
13449 (dump_candidate): Dump first_interp field.
13450 (replace_mult_candidate): Process all interpretations, not just
13452 (replace_rhs_if_not_dup): Likewise.
13453 (replace_one_candidate): Likewise.
13455 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
13457 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
13459 (aarch64_needs_frame_chain): New function.
13460 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
13462 2018-05-23 Sudakshina Das <sudi.das@arm.com>
13465 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
13466 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
13467 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
13468 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
13469 as true for strict-align.
13470 (aarch64_can_inline_p): Perform checks even when callee has no
13471 attributes to check for strict alignment.
13472 * doc/extend.texi (AArch64 Function Attributes): Document
13474 * doc/invoke.texi: (AArch64 Options): Likewise.
13476 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
13478 PR tree-optimization/85853
13479 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
13480 the handling of the root of the node to...
13481 (vect_slp_analyze_node_operations_1): ...this new function,
13482 and run the whole thing with the child nodes' def types
13483 set according to their SLP node's def type.
13485 2018-05-23 Richard Biener <rguenther@suse.de>
13487 PR middle-end/85874
13488 * tree-data-ref.c (create_runtime_alias_checks): Defer
13489 and ignore overflow warnings.
13491 2018-05-23 Yury Gribov <tetra2005@gmail.com>
13493 PR tree-optimization/85822
13494 * tree-vrp.c (is_masked_range_test): Fix handling of negative
13497 2018-05-23 Richard Biener <rguenther@suse.de>
13499 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13500 memset constants via native_interpret_expr.
13502 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
13505 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
13507 (cgraph_node::create_alias): Likewise.
13508 (cgraph_node::get_availability): Check ifunc_resolver instead
13509 of looking up ifunc attribute.
13510 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
13511 * varasm.c (do_assemble_alias): Likewise.
13512 (assemble_alias): Likewise.
13513 (default_binds_local_p_3): Likewise.
13514 * cgraph.h (cgraph_node): Add ifunc_resolver.
13515 (cgraph_node::only_called_directly_or_aliased_p): Return false
13516 for IFUNC resolver.
13517 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
13519 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
13520 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
13521 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
13522 instead of looking up ifunc attribute.
13524 2018-05-22 Luis Machado <luis.machado@linaro.org>
13526 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
13528 2018-05-22 Martin Sebor <msebor@redhat.com>
13530 PR middle-end/85359
13531 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
13532 only when expasion succeeds.
13533 (expand_builtin_strcmp): Same.
13534 (expand_builtin_strncmp): Same.
13536 2018-05-22 Martin Sebor <msebor@redhat.com>
13538 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
13540 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
13541 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13543 * config/aarch64/aarch64-ldpstp.md: Replace uses of
13544 aarch64_mem_pair_operand with memory_operand and delete operand swapping
13546 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13547 Add check for legitimate_address.
13548 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
13549 (aarch64_swap_ldrstr_operands): New.
13550 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
13553 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
13554 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13556 * config/aarch64/aarch64.md: New patterns to generate stp
13558 (store_pair_sw, store_pair_dw): New patterns to generate stp for
13559 single words and double words.
13560 (load_pair_sw, load_pair_dw): Likewise.
13561 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
13563 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
13565 * config/aarch64/aarch64-ldpstp.md: Modify peephole
13566 for different mode ldpstp and add peephole for merged zero stores.
13567 Likewise for loads.
13568 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13570 (aarch64_gen_store_pair): Rename calls to match new patterns.
13571 (aarch64_gen_load_pair): Rename calls to match new patterns.
13572 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
13573 (load_pair<DREG:mode><DREG2:mode>): ... This.
13574 (store_pair<mode>): Rename to...
13575 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
13576 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
13577 New mode iterators.
13578 (V_INT_EQUIV): Handle SImode.
13579 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
13582 2018-05-22 Martin Sebor <msebor@redhat.com>
13585 * calls.c (maybe_warn_nonstring_arg): Use string length to set
13586 or ajust the presumed bound on an operation to avoid unnecessary
13589 2018-05-22 Martin Sebor <msebor@redhat.com>
13591 PR tree-optimization/85826
13592 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
13593 assuming that a DECL necesarily has a constant size.
13595 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
13597 PR middle-end/85862
13598 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
13600 2018-05-22 Richard Biener <rguenther@suse.de>
13602 PR tree-optimization/85834
13603 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
13604 non-constant and non-zero memset arguments.
13606 2018-05-22 Martin Liska <mliska@suse.cz>
13609 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
13611 2018-05-22 Richard Biener <rguenther@suse.de>
13613 PR tree-optimization/85863
13614 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
13615 comparisons when vectype is specified.
13616 (vectorizable_condition): Do not specify vectype for
13617 vect_is_simple_cond when SLP vectorizing.
13619 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
13622 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
13623 define __ibm128 as long double.
13624 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
13625 as a distinct type when IEEE 128-bit support is enabled.
13626 (init_float128_ieee): Fix up conversions between IFmode and IEEE
13627 128-bit types to use the correct functions.
13628 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
13629 convert between 128-bit floating point types that have different
13630 modes but the same representation, instead of using gen_lowpart to
13632 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
13634 (IFKF_reg): New attributes to give the register constraints for
13636 (extend<mode>tf2_internal): New insns to mark an explicit
13637 conversion between 128-bit floating point types that have a
13638 different mode but share the same representation.
13640 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
13642 PR tree-optimization/85814
13643 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
13644 a null return from get_strinfo when unsharing the next
13645 strinfo in the chain.
13647 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
13650 * varasm.c (weak_finish): Clean up weak_decls.
13652 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13654 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
13655 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
13656 UNSPEC_UADALP values.
13657 * config/aarch64/iterators.md (ABAL): New int iterator.
13660 (sur): Add mappings for the above.
13661 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
13663 (aarch64_<sur>abal<mode>_4): Likewise.
13664 (aarch64_<sur>adalp<mode>_3): Likewise.
13665 (<sur>sadv16qi): New define_expand.
13667 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
13669 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
13670 (*movdf_internal): Ditto.
13671 (*rcpsf2_sse): Ditto.
13672 (*rsqrtsf2_sse): Ditto.
13673 (*sqrt<mode>2_sse): Ditto.
13675 2018-05-21 Tamar Christina <tamar.christina@arm.com>
13677 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
13679 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
13680 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
13681 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
13682 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
13684 * config/aarch64/arm_neon.h: Likewise.
13685 * config/aarch64/iterators.md (VQ_I): New.
13687 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
13689 * config.gcc: Add arc/t-multilib-linux to tmake_file for
13691 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
13694 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13696 * config/nds32/constraints.md (S): New constraint.
13697 * config/nds32/nds32.md (call_internal): Use constraint S.
13698 (call_value_internal): Likewise.
13699 (sibcall_internal): Likewise.
13700 (sibcall_value_internal): Likewise.
13702 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
13703 Chung-Ju Wu <jasonwucj@gmail.com>
13705 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
13706 into consideration.
13708 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
13709 Chung-Ju Wu <jasonwucj@gmail.com>
13711 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
13712 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
13713 (nds32_rtx_costs_impl): Simplify.
13714 (nds32_address_cost_impl): Simplify.
13715 (nds32_init_rtx_costs): New function.
13716 (nds32_rtx_costs_speed_prefer): Likewise.
13717 (nds32_rtx_costs_size_prefer): Likewise.
13718 (nds32_address_cost_speed_prefer): Likewise.
13719 (nds32_address_cost_speed_fwprop): Likewise.
13720 (nds32_address_cost_size_prefer): Likewise.
13721 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
13722 * config/nds32/nds32.c (nds32_option_override): Use
13723 nds32_init_rtx_costs function.
13725 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13727 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
13728 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
13729 (TARGET_PIPELINE_N8): Likewise.
13730 (TARGET_PIPELINE_N10): Likewise.
13731 (TARGET_PIPELINE_N13): Likewise.
13732 (TARGET_PIPELINE_GRAYWOLF): Likewise.
13734 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
13736 * config/nds32/nds32-fpu.md: Update copyright year.
13738 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13740 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
13742 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13744 * config/nds32/nds32.c
13745 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
13746 * config/nds32/nds32.opt (minline-asm-r15): New option.
13748 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13750 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
13752 * config/nds32/nds32.md (abssi2): New pattern.
13754 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
13756 * config/i386/i386.md (rex64namesuffix): New mode attribute.
13757 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
13758 Merge insn pattern from sse_cvtsi2ss<round_name> and
13759 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
13760 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
13761 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
13762 using SWI48 mode iterator.
13763 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
13764 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
13765 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
13766 pattern from sse_cvttss2si<round_saeonly_name>
13767 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
13768 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
13769 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
13770 using SWI48 mode iterator.
13771 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
13772 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
13773 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
13774 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
13775 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
13776 using SWI48 mode iterator.
13777 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
13778 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
13779 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
13780 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
13781 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
13782 SWI48 mode iterator.
13783 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
13784 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
13785 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
13786 pattern from sse_cvttsd2si<round_saeonly_name>
13787 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
13789 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13791 * config/nds32/nds32-md-auxiliary.c
13792 (nds32_valid_smw_lwm_base_p): Refine.
13793 (nds32_output_smw_single_word): Refine.
13794 (nds32_output_smw_double_word): New.
13795 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
13797 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13799 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
13800 (nds32_output_stack_pop): Refine.
13801 (nds32_expand_unaligned_load): Refine.
13802 (nds32_expand_unaligned_store): Refine.
13804 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13805 Chung-Ju Wu <jasonwucj@gmail.com>
13807 * config/nds32/constants.md: Add TP_REGNUM constant.
13808 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
13809 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
13811 * config/nds32/nds32-doubleword.md: Consider flag_pic.
13812 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
13813 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
13814 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
13815 and PIC code generation.
13816 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
13818 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
13820 * config/nds32/nds32.md: Support TLS and PIC.
13821 * config/nds32/nds32.c: Support TLS and PIC.
13822 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
13823 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
13826 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13828 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
13829 mode with E_ prefix.
13831 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13832 Chung-Ju Wu <jasonwucj@gmail.com>
13834 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
13835 * config/nds32/nds32-md-auxiliary.c
13836 (symbolic_reference_mentioned_p): New.
13837 (nds32_legitimize_ict_address): New.
13838 (nds32_expand_ict_move): New.
13839 (nds32_indirect_call_referenced_p): New.
13840 (nds32_symbol_binds_local_p): Delete.
13841 (nds32_long_call_p): Modify.
13842 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
13843 * config/nds32/nds32-protos.h
13844 (symbolic_reference_mentioned_p): Declare.
13845 (nds32_legitimize_ict_address): Declare.
13846 (nds32_expand_ict_move): Declare.
13847 (nds32_indirect_call_referenced_p): Declare.
13848 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
13849 (nds32_relax_group): Use nds32_ict_const_p as condition.
13850 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
13851 (nds32_asm_file_start): Output ict_model directive in asm code.
13852 (nds32_legitimate_address_p): Consider indirect call.
13853 (nds32_print_operand): Consider indirect call.
13854 (nds32_print_operand_address): Consider indirect call.
13855 (nds32_insert_attributes): Handle "indirect_call" attribute.
13856 (TARGET_LEGITIMATE_ADDRESS_P): Define.
13857 (TARGET_LEGITIMATE_CONSTANT_P): Define.
13858 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
13859 (TARGET_DELEGITIMIZE_ADDRESS): Define.
13860 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
13861 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
13862 (TARGET_ICT_MODEL_SMALL): Define.
13863 (TARGET_ICT_MODEL_LARGE): Define.
13864 * config/nds32/nds32.md (movsi): Consider ict model.
13865 (call, call_value): Consider ict model.
13866 (sibcall, sibcall_value): Consider ict model.
13867 * config/nds32/nds32.opt (mict-model): New option.
13868 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
13871 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
13872 Monk Chiang <sh.chiang04@gmail.com>
13873 Jim Wilson <jimw@sifive.com>
13875 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
13876 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
13877 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
13878 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
13879 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
13880 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
13881 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
13882 compute save_libcall_adjustment properly.
13883 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
13884 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
13885 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
13886 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
13887 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
13888 (ABI_SPEC): Handle mabi=ilp32e.
13889 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
13890 (RVE): Add RVE mask.
13891 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
13892 <-march>: Add rv32e as an example.
13894 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
13897 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
13898 (intra_create_variable_infos): Handle C++ constructors.
13900 2018-05-18 Martin Liska <mliska@suse.cz>
13902 * passes.def: Remove a redundant pass.
13904 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
13907 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
13909 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13911 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
13913 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
13915 (armv2, armv2a, armv3, armv3m): Delete architectures.
13916 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
13917 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
13918 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
13920 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
13921 (*mulsidi3adddi): Likewise.
13922 (mulsidi3): Likewise.
13923 (*mulsidi3_nov6): Likewise.
13924 (umulsidi3): Likewise.
13925 (umulsidi3_nov6): Likewise.
13926 (umaddsidi4): Likewise.
13927 (*umulsidi3adddi): Likewise.
13928 (smulsi3_highpart): Likewise.
13929 (*smulsi3_highpart_nov6): Likewise.
13930 (umulsi3_highpart): Likewise.
13931 (*umulsi3_highpart_nov6): Likewise.
13932 * config/arm/arm.h (arm_arch3m): Delete.
13933 * config/arm/arm.c (arm_arch3m): Delete.
13934 (arm_option_override_internal): Update armv3-related comment.
13935 (arm_configure_build_target): Delete use of isa_bit_mode32.
13936 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
13937 (arm_rtx_costs_internal): Delete check of arm_arch3m.
13938 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
13939 (mulsa3): Likewise.
13940 (mulusa3): Likewise.
13941 * config/arm/arm-protos.h (arm_arch3m): Delete.
13942 * config/arm/arm-tables.opt: Regenerate.
13943 * config/arm/arm-tune.md: Likewise.
13944 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
13945 deleted architectures.
13947 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13949 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
13950 (armv5t, armv5te): New features.
13951 (ARMv5, ARMv5e): Delete fgroups.
13952 (ARMv5t, ARMv5te): Adjust for above changes.
13953 (ARMv6m): Likewise.
13954 (armv5, armv5e): Delete arches.
13955 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
13957 (*call_reg_arm): Likewise.
13958 (*call_value_reg_armv5): Likewise.
13959 (*call_value_reg_arm): Likewise.
13960 (*call_symbol): Likewise.
13961 (*call_value_symbol): Likewise.
13962 (*sibcall_insn): Likewise.
13963 (*sibcall_value_insn): Likewise.
13964 (clzsi2): Likewise.
13965 (prefetch): Likewise.
13966 (define_split and define_peephole2 dependent on arm_arch5):
13968 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
13970 (TARGET_ARM_QBIT): Likewise.
13971 (TARGET_DSP_MULTIPLY): Likewise.
13972 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
13973 (arm_arch5, arm_arch5e): Delete.
13974 (arm_arch5t, arm_arch5te): Declare.
13975 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
13976 (arm_arch5t): Declare.
13977 (arm_option_reconfigure_globals): Update for the above.
13978 (arm_options_perform_arch_sanity_checks): Update comment, replace
13979 use of arm_arch5 with arm_arch5t.
13980 (use_return_insn): Likewise.
13981 (arm_emit_call_insn): Likewise.
13982 (output_return_instruction): Likewise.
13983 (arm_final_prescan_insn): Likewise.
13984 (arm_coproc_builtin_available): Likewise.
13985 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
13986 arm_arch5e with arm_arch5t and arm_arch5te.
13987 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
13988 (arm_arch5t, arm_arch5te): Declare.
13989 * config/arm/arm-tables.opt: Regenerate.
13990 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
13991 * config/arm/t-multilib: Likewise.
13992 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
13993 instead of arm_arch5.
13994 (*call_reg_thumb1): Likewise.
13995 (*call_value_reg_thumb1_v5): Likewise.
13996 (*call_value_reg_thumb1): Likewise.
13997 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
13999 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14001 2018-05-18 Martin Liska <mliska@suse.cz>
14003 PR gcov-profile/84846
14004 * doc/gcov.texi: Document -t option of gcov tool.
14006 2018-05-18 Martin Liska <mliska@suse.cz>
14008 PR gcov-profile/84846
14009 * gcov.c (print_usage): Add new -t option.
14010 (process_args): Handle the option.
14011 (generate_results): Use stdout as output when requested by
14014 2018-05-18 Martin Liska <mliska@suse.cz>
14016 PR gcov-profile/84846
14017 * coverage.c (coverage_init): Write PWD to .gcno file.
14018 * doc/gcov.texi: Document how working directory is printed.
14019 * gcov-dump.c (dump_gcov_file): Print PWD.
14020 * gcov.c (output_intermediate_file): Likewise.
14021 (read_graph_file): Read PWD string.
14022 (output_lines): Print PWD.
14024 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14026 PR middle-end/85817
14027 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14028 for retval and return false if all args to phi are zero.
14030 2018-05-18 Richard Biener <rguenther@suse.de>
14032 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14034 (evrp_dom_walker::before_dom_children): Call it.
14036 2018-05-18 Richard Biener <rguenther@suse.de>
14038 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14039 results when processing array refs with variable index.
14041 2018-05-18 Toon Moene <toon@moene.org>
14043 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14044 directly after that of -floop-interchange. Indicate that both
14045 options are enabled by default when specifying -O3.
14047 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14049 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14050 iterator. Delete separate integer-mode vec_set<mode> expander.
14051 (aarch64_simd_vec_setv2di): Delete.
14052 (vec_setv2di): Delete.
14053 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14054 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
14055 the "w, r" alternative.
14057 2018-05-18 Martin Liska <mliska@suse.cz>
14059 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14060 * tree-pass.h (make_pass_lower_switch_O0): New function.
14061 * tree-switch-conversion.c (node_has_low_bound): Remove.
14062 (node_has_high_bound): Likewise.
14063 (node_is_bounded): Likewise.
14064 (class pass_lower_switch): Make it a template type and create
14066 (pass_lower_switch::execute): Add template argument.
14067 (make_pass_lower_switch): New function.
14068 (make_pass_lower_switch_O0): New function.
14069 (do_jump_if_equal): Remove.
14070 (emit_case_nodes): Simplify to just handle all 3 cases and leave
14071 all the hard work to tree optimization passes.
14073 2018-05-18 Martin Liska <mliska@suse.cz>
14075 * dbgcnt.c (limit_low): Renamed from limit.
14076 (limit_high): New variable.
14077 (dbg_cnt_is_enabled): Check for upper limit.
14078 (dbg_cnt): Adjust dumping.
14079 (dbg_cnt_set_limit_by_index): Add new argument for high
14081 (dbg_cnt_set_limit_by_name): Likewise.
14082 (dbg_cnt_process_single_pair): Parse new format.
14083 (dbg_cnt_process_opt): Use strtok.
14084 (dbg_cnt_list_all_counters): Remove 'value' and add
14086 * doc/invoke.texi: Document changes.
14088 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
14090 * doc/sourcebuild.texi (scalar_all_fma): Document.
14091 * tree.def (FMA_EXPR): Delete.
14092 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14093 * internal-fn.c (ternary_direct): New macro.
14094 (expand_ternary_optab_fn): Likewise.
14095 (direct_ternary_optab_supported_p): Likewise.
14096 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14097 * builtins.c (fold_builtin_fma): Delete.
14098 (fold_builtin_3): Don't call it.
14099 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14100 * expr.c (expand_expr_real_2): Likewise.
14101 * fold-const.c (operand_equal_p): Likewise.
14102 (fold_ternary_loc): Likewise.
14103 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14104 * gimple.c (DEFTREECODE): Likewise.
14105 * gimplify.c (gimplify_expr): Likewise.
14106 * optabs-tree.c (optab_for_tree_code): Likewise.
14107 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14108 * tree-eh.c (operation_could_trap_p): Likewise.
14109 (stmt_could_throw_1_p): Likewise.
14110 * tree-inline.c (estimate_operator_cost): Likewise.
14111 * tree-pretty-print.c (dump_generic_node): Likewise.
14112 (op_code_prio): Likewise.
14113 * tree-ssa-loop-im.c (stmt_cost): Likewise.
14114 * tree-ssa-operands.c (get_expr_operands): Likewise.
14115 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14116 * fold-const-call.h (fold_fma): Delete.
14117 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14118 CFN_FNMA and CFN_FNMS.
14119 (fold_fma): Delete.
14120 * genmatch.c (combined_fn): New enum.
14121 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14122 (commutative_op): New function.
14123 (commutate): Use it. Handle more than 2 operands.
14124 (dt_operand::gen_gimple_expr): Use commutative_op.
14125 (parser::parse_expr): Allow :c to be used with non-binary
14126 operators if the commutative operand is known.
14127 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14128 CFN_FMS, CFN_FNMA and CFN_FNMS.
14129 (backprop::process_assign_use): Remove FMA_EXPR handling.
14130 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14131 (gen_hsa_fma): New function.
14132 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14133 IFN_FNMA and IFN_FNMS.
14134 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14135 * gimple-fold.h (follow_all_ssa_edges): Declare.
14136 * gimple-fold.c (follow_all_ssa_edges): New function.
14137 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14138 gimple_build interface and use follow_all_ssa_edges to fold the result.
14139 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14140 instead of checking for optabs directly.
14141 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14142 rather than FMA_EXPRs.
14143 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14144 call to IFN_FMA instead of an FMA_EXPR.
14146 2018-05-17 Jim Wilson <jimw@sifive.com>
14148 * expr.c (do_tablejump): When converting index to Pmode, if we have a
14149 sign extended promoted subreg, and the range does not have the sign bit
14150 set, then do a sign extend.
14152 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14153 test, check for sign extended subreg and/or constant operands, and
14154 do a sign extend in that case.
14156 2018-05-17 Steve Ellcey <sellcey@cavium.com>
14158 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14159 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14161 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14162 Change logics_shift_reg to logics_shift_imm.
14163 (thunderx2t99_fp_loadpair_basic): Delete.
14164 (thunderx2t99_fp_storepair_basic): Delete.
14165 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14166 (thunderx2t99_asimd_polynomial): Delete.
14167 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14168 and neon_fp_mul_d_scalar_q.
14169 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14170 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14171 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14172 (thunderx2t99_asimd_lut): Add missing tbl types.
14173 (thunderx2t99_asimd_ext): Delete.
14174 (thunderx2t99_asimd_load1_1_mult): Delete.
14175 (thunderx2t99_asimd_load1_2_mult): Delete.
14176 (thunderx2t99_asimd_load1_ldp): New.
14177 (thunderx2t99_asimd_load1): New.
14178 (thunderx2t99_asimd_load2): Add missing *load2* types.
14179 (thunderx2t99_asimd_load3): New.
14180 (thunderx2t99_asimd_load4): New.
14181 (thunderx2t99_asimd_store1_1_mult): Delete.
14182 (thunderx2t99_asimd_store1_2_mult): Delete.
14183 (thunderx2t99_asimd_store2_mult): Delete.
14184 (thunderx2t99_asimd_store2_onelane): Delete.
14185 (thunderx2t99_asimd_store_stp): New.
14186 (thunderx2t99_asimd_store1): New.
14187 (thunderx2t99_asimd_store2): New.
14188 (thunderx2t99_asimd_store3): New.
14189 (thunderx2t99_asimd_store4): New.
14191 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
14193 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14194 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
14196 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
14197 Segher Boessenkool <segher@kernel.crashing.org>
14200 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14203 2018-05-17 Richard Biener <rguenther@suse.de>
14205 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14206 for pruning loop and prune defs feeding only already visited PHIs.
14208 2018-05-17 Richard Biener <rguenther@suse.de>
14210 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14212 2018-05-17 Bin Cheng <bin.cheng@arm.com>
14213 Richard Biener <rguenther@suse.de>
14215 PR tree-optimization/85793
14216 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14217 for VMAT_ELEMENTWISE.
14219 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14221 * internal-fn.h (lookup_internal_fn): Declare
14222 * internal-fn.c (lookup_internal_fn): New function.
14223 * gimple.c (gimple_build_call_from_tree): Handle calls to
14224 internal functions.
14225 * gimple-pretty-print.c (dump_gimple_call): Print "." before
14226 internal function names.
14227 * tree-pretty-print.c (dump_generic_node): Likewise.
14228 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14230 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14232 * gimple-fold.h (gimple_build): Make the function forms take
14233 combined_fn rather than built_in_function.
14234 (gimple_simplify): Likewise.
14235 * gimple-match-head.c (gimple_simplify): Likewise.
14236 * gimple-fold.c (gimple_build): Likewise.
14237 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14238 rather than gimple_build_call_internal.
14239 (get_initial_defs_for_reduction): Likewise.
14240 (vect_create_epilog_for_reduction): Likewise.
14241 (vectorizable_live_operation): Likewise.
14243 2018-05-17 Martin Liska <mliska@suse.cz>
14245 * gimple-ssa-sprintf.c (format_directive): Do not use
14246 space in between 'G_' and '('.
14248 2018-05-17 Jakub Jelinek <jakub@redhat.com>
14251 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14252 even if the mask is not all ones.
14255 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14257 (ix86_gimple_fold_builtin): Likewise.
14260 * config/i386/i386.c: Include tree-vector-builder.h.
14261 (ix86_vector_shift_count): New function.
14262 (ix86_fold_builtin): Fold shift builtins by scalar count.
14263 (ix86_gimple_fold_builtin): Likewise.
14265 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14266 _mm512_setzero): New intrinsics.
14268 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
14269 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14271 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14272 code generation for cases where splatting a value is not useful.
14273 * simplify-rtx.c (simplify_ternary_operation): Simplify
14274 vec_merge across a vec_duplicate and a paradoxical subreg forming
14275 a vector mode to a vec_concat.
14277 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
14279 * config.gcc: Support "goldmont-plus".
14280 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14282 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14283 PROCESSOR_GOLDMONT_PLUS.
14284 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14285 (processor_target_table): Add "goldmont-plus".
14286 (PTA_GOLDMONT_PLUS): Define.
14287 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14288 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14289 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14290 (fold_builtin_cpu): Add "goldmont-plus".
14291 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14292 (ix86_option_override_internal): Add "goldmont-plus".
14293 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14294 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14295 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14296 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14298 2018-05-17 Richard Biener <rguenther@suse.de>
14300 PR tree-optimization/85757
14301 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14302 remove defs that only feed that PHI from further processing.
14304 2018-05-16 Jim Wilson <jimw@sifive.com>
14306 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14308 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14310 2018-05-16 Mark Wielaard <mark@klomp.org>
14312 * dwarf2out.c (count_index_strings): New function.
14313 (output_indirect_strings): Call count_index_strings and generate
14314 header for dwarf_version >= 5.
14316 2018-05-16 Mark Wielaard <mark@klomp.org>
14318 * dwarf2out.c (dwarf_FORM): New function.
14319 (set_indirect_string): Use dwarf_FORM.
14320 (reset_indirect_string): Likewise.
14321 (size_of_die): Likewise.
14322 (value_format): Likewise.
14323 (output_die): Likewise.
14324 (add_skeleton_AT_string): Likewise.
14325 (output_macinfo_op): Likewise.
14326 (index_string): Likewise.
14327 (output_index_string_offset): Likewise.
14328 (output_index_string): Likewise.
14329 (count_index_strings): Likewise.
14331 2018-05-16 Carl Love <cel@us.ibm.com>
14333 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14334 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
14336 2018-05-16 Martin Jambor <mjambor@suse.cz>
14338 * ipa-prop.c (ipa_free_all_edge_args): Remove.
14339 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
14341 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
14343 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
14344 (fnma<mode>4): Likewise.
14345 (fms<mode>4): Likewise.
14346 (fnms<mode>4): Likewise.
14347 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
14348 (aarch64_fnma<mode>4): Likewise.
14349 (aarch64_fms<mode>4): Likewise.
14350 (aarch64_fnms<mode>4): Likewise.
14351 (aarch64_fnmadd<mode>4): Likewise.
14353 2018-05-16 Jason Merrill <jason@redhat.com>
14355 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
14357 2018-05-16 Richard Biener <rguenther@suse.de>
14359 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
14360 (dump_stmt_cost): Declare.
14361 (add_stmt_cost): Dump cost we add.
14362 (add_stmt_costs): New function.
14363 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
14364 No longer exported.
14365 (vect_analyze_stmt): Adjust prototype.
14366 (vectorizable_condition): Likewise.
14367 (vectorizable_live_operation): Likewise.
14368 (vectorizable_reduction): Likewise.
14369 (vectorizable_induction): Likewise.
14370 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
14371 cost vector to pass to vectorizable_ and record afterwards.
14372 (vect_model_reduction_cost): Take cost vector argument and adjust.
14373 (vect_model_induction_cost): Likewise.
14374 (vectorizable_reduction): Likewise.
14375 (vectorizable_induction): Likewise.
14376 (vectorizable_live_operation): Likewise.
14377 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
14378 SLP_TREE_NUMBER_OF_VEC_STMTS.
14379 (vect_analyze_slp_cost_1): Remove.
14380 (vect_analyze_slp_cost): Likewise.
14381 (vect_slp_analyze_node_operations): Take visited args and
14382 a target cost vector. Avoid processing already visited stmt sets.
14383 (vect_slp_analyze_operations): Use a local cost vector to gather
14384 costs and register those of non-discarded instances.
14385 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
14386 (vect_schedule_slp_instance): Remove copying of
14387 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
14389 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
14390 adding cost. Record cost entry location.
14391 (vect_prologue_cost_for_slp_op): Function to compute cost of
14392 a constant or invariant generated for SLP vect in the prologue,
14393 split out from vect_analyze_slp_cost_1.
14394 (vect_model_simple_cost): Make static. Adjust for SLP costing.
14395 (vect_model_promotion_demotion_cost): Likewise.
14396 (vect_model_store_cost): Likewise, make static.
14397 (vect_model_load_cost): Likewise.
14398 (vectorizable_bswap): Add cost vector arg and adjust.
14399 (vectorizable_call): Likewise.
14400 (vectorizable_simd_clone_call): Likewise.
14401 (vectorizable_conversion): Likewise.
14402 (vectorizable_assignment): Likewise.
14403 (vectorizable_shift): Likewise.
14404 (vectorizable_operation): Likewise.
14405 (vectorizable_store): Likewise.
14406 (vectorizable_load): Likewise.
14407 (vectorizable_condition): Likewise.
14408 (vectorizable_comparison): Likewise.
14409 (can_vectorize_live_stmts): Likewise.
14410 (vect_analyze_stmt): Likewise.
14411 (vect_transform_stmt): Adjust calls to vectorizable_*.
14412 * tree-vectorizer.c: Include gimple-pretty-print.h.
14413 (dump_stmt_cost): New function.
14415 2018-05-16 Richard Biener <rguenther@suse.de>
14417 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
14418 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
14419 * tree-ssa-dse.c: Include tree-ssa-loop.h.
14420 (check_name): New callback.
14421 (dse_classify_store): Track cycles via a visited bitmap of PHI
14422 defs and simplify handling of in-loop and across loop dead stores
14423 and properly fail for loop-variant refs. Handle byte-tracking with
14424 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
14427 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
14429 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
14430 (vect_get_mask_type_for_stmt): Likewise.
14431 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
14433 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
14434 to determine the statement's vector type and the vector type that
14435 should be used for calculating nunits. Deal with cases in which
14436 the type has to be deferred.
14437 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
14438 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
14439 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
14440 (vect_determine_vf_for_stmt): New functions, split out from...
14441 (vect_determine_vectorization_factor): ...here.
14442 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
14443 (vect_get_mask_type_for_stmt): New functions, split out from
14444 vect_determine_vectorization_factor.
14446 2018-05-16 Richard Biener <rguenther@suse.de>
14448 * tree-cfg.c (verify_gimple_assign_ternary): Properly
14449 verify the [VEC_]COND_EXPR embedded comparison.
14451 2018-05-15 Martin Sebor <msebor@redhat.com>
14453 PR tree-optimization/85753
14454 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
14455 RECORD_TYPE in addition to ARRAY_TYPE.
14457 2018-05-15 Martin Sebor <msebor@redhat.com>
14459 PR middle-end/85643
14460 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
14462 2018-05-15 Richard Biener <rguenther@suse.de>
14464 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
14465 add by_clobber_p one. Change algorithm to collect all defs
14466 representing uses we need to walk and try reducing them to
14467 a single one before failing.
14468 (dse_dom_walker::dse_optimize_stmt): Adjust.
14470 2018-05-13 Mark Wielaard <mark@klomp.org>
14472 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
14473 (size_of_loc_descr): Likewise.
14474 (output_loc_operands): Likewise.
14475 (output_loc_operands_raw): Likewise.
14476 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
14477 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
14478 (hash_loc_operands): Likewise.
14479 (compare_loc_operands): Likewise.
14481 2018-05-14 Mark Wielaard <mark@klomp.org>
14483 * dwarf2out.c (count_index_addrs): New function.
14484 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
14486 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14488 PR tree-optimization/83648
14489 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
14490 return value as malloc candidate.
14492 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14495 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
14496 param as true in call to suggest_attribute.
14498 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
14500 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
14501 -mreadonly-in-sdata.
14503 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14505 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
14507 (aarch64_crypto_aesd_fused): Likewise.
14509 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
14511 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
14512 (movsi_aarch64): Likewise.
14513 (load_pairsi): Likewise.
14514 (load_pairdi): Likewise.
14515 (store_pairsi): Likewise.
14516 (store_pairdi): Likewise.
14517 (load_pairsf): Likewise.
14518 (load_pairdf): Likewise.
14519 (store_pairsf): Likewise.
14520 (store_pairdf): Likewise.
14521 (zero_extend): Likewise.
14522 (trunc): Swap alternatives.
14523 (fcvt_target): Add '?' to prefer w over r.
14525 2018-05-14 Jakub Jelinek <jakub@redhat.com>
14528 * config/i386/i386.md: Disallow non-commutative arithmetics in
14529 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
14530 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
14531 in the peephole2 before it.
14533 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
14535 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
14536 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
14537 (ix86_handle_option): Handle -mcldemote.
14538 * config.gcc: New header.
14539 * config/i386/cldemoteintrin.h: New file.
14540 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
14541 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14543 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14544 OPTION_MASK_ISA_CLDEMOTE.
14545 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
14546 (ix86_valid_target_attribute_inner_p): Ditto.
14547 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
14548 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
14549 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
14550 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
14551 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
14553 * config/i386/i386.opt: Add -mcldemote.
14554 * config/i386/x86intrin.h: New header.
14555 * doc/invoke.texi: Add -mcldemote.
14557 2018-05-14 Richard Biener <rguenther@suse.de>
14559 * doc/match-and-simplify.texi: Adjust :s documentation.
14561 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
14563 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
14564 intended memcpy size.
14565 (REORDER_45): Likewise.
14567 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
14569 * sort.cc: New file.
14570 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
14571 * vec.c (qsort_chk): Use gcc_qsort.
14572 * Makefile.in (OBJS-libcommon): Add sort.o.
14573 (build/sort.o): New target. Use it...
14574 (BUILD_RTL): ... here, and...
14575 (build/gencfn-macros): ... here, and...
14576 (build/genmatch): ... here.
14578 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14579 Chung-Ju Wu <jasonwucj@gmail.com>
14581 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
14582 * config/nds32/nds32-graywolf.md: New file.
14583 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
14584 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
14586 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
14587 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
14588 * config/nds32/nds32.md (pipeline_model): Add graywolf.
14589 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
14590 * config/nds32/pipelines.md: Include n15 settings.
14592 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14593 Chung-Ju Wu <jasonwucj@gmail.com>
14595 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
14596 * config/nds32/nds32-n13.md: New file.
14597 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
14598 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
14600 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
14601 * config/nds32/nds32.md (pipeline_model): Add n13.
14602 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
14603 * config/nds32/pipelines.md: Include n13 settings.
14605 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14606 Chung-Ju Wu <jasonwucj@gmail.com>
14608 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
14609 * config/nds32/nds32-n10.md: New file.
14610 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
14611 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
14613 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
14614 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
14615 * config/nds32/nds32.md (pipeline_model): Add n10.
14616 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
14617 * config/nds32/pipelines.md: Include n10 settings.
14619 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
14620 Kito Cheng <kito.cheng@gmail.com>
14621 Chung-Ju Wu <jasonwucj@gmail.com>
14623 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
14624 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14625 Add enum values for DSP extension instructions.
14626 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
14628 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
14629 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
14630 New code iterators.
14631 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
14632 * config/nds32/nds32-dspext.md: New file for DSP implementation.
14633 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
14634 * config/nds32/nds32-intrinsic.md: Likewise.
14635 * config/nds32/nds32_intrinsic.h: Likewise.
14636 * config/nds32/nds32-md-auxiliary.c: Likewise.
14637 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
14638 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
14639 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
14640 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
14641 * config/nds32/nds32-protos.h: New declarations for DSP extension.
14642 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
14643 TYPE_DMAC in switch statement.
14644 * config/nds32/nds32.c: New checking and implementation for DSP
14645 extension instructions.
14646 * config/nds32/nds32.h: Likewise.
14647 * config/nds32/nds32.md: Likewise.
14648 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
14649 * config/nds32/predicates.md: Implement new predicates for DSP
14652 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14654 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
14655 Reformat alternatives and attributes so it is easier to identify
14656 which constraints/attributes go with which instruction.
14657 (mov<mode>_hardfloat32, FMOVE64): Likewise.
14658 (mov<mode>_softfloat32, FMOVE64): Likewise.
14659 (mov<mode>_hardfloat64, FMOVE64): Likewise.
14660 (mov<mode>_softfloat64, FMOVE64): Likewise.
14662 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
14664 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14666 (Basic PowerPC Built-in Functions): The new name of the
14667 subsection previously known as "PowerPC Built-in Functions".
14668 (Basic PowerPC Built-in Functions Available on all Configurations):
14670 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
14671 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
14672 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
14673 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
14675 2018-05-11 Martin Jambor <mjambor@suse.cz>
14678 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
14681 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
14684 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
14686 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
14688 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
14689 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
14690 (ix86_handle_option): Handle -mwaitpkg.
14691 * config.gcc: New header.
14692 * config/i386/cpuid.h (bit_WAITPKG): New bit.
14693 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
14694 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
14696 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14697 OPTION_MASK_ISA_WAITPKG.
14698 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
14699 (ix86_option_override_internal): Add PTA_WAITPKG.
14700 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
14701 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
14702 IX86_BUILTIN_TPAUSE.
14703 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
14704 __builtin_ia32_umwait and __builtin_ia32_tpause.
14705 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
14706 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
14707 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
14708 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
14709 UNSPECV_TPAUSE): New.
14710 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
14711 * config/i386/i386.opt: Add -mwaitpkg.
14712 * config/i386/waitpkgintrin.h: New file.
14713 * config/i386/x86intrin.h: New header.
14714 * doc/invoke.texi: Add -mwaitpkg.
14716 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
14719 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
14721 (cortex-m0): Use armv6s-m isa.
14722 (cortex-m0plus): Likewise.
14723 (cortex-m1): Likewise.
14724 (cortex-m0.small-multiply): Likewise.
14725 (cortex-m0plus.small-multiply): Likewise.
14726 (cortex-m1.small-multiply): Likewise.
14728 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
14729 Jakub Jelinek <jakub@redhat.com>
14731 PR tree-optimization/85692
14732 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
14733 source permute as well.
14735 2018-05-11 Martin Liska <mliska@suse.cz>
14738 * doc/extend.texi: Document LLVM style format for no_sanitize
14741 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14743 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
14744 mode_supports_vsx_dform_quad to mode_supports_dq_form.
14745 (mode_supports_vsx_dform_quad): Likewise.
14746 (mode_supports_vmx_dform): Move these functions to be next to the
14747 other mode_supports functions.
14748 (mode_supports_dq_form): Likewise.
14749 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
14750 mode_supports_dq_form.
14751 (reg_offset_addressing_ok_p): Likewise.
14752 (offsettable_ok_by_alignment): Likewise.
14753 (rs6000_legitimate_offset_address_p): Likewise.
14754 (legitimate_lo_sum_address_p): Likewise.
14755 (rs6000_legitimize_address): Likewise.
14756 (rs6000_legitimize_reload_address): Likewise.
14757 (rs6000_secondary_reload_inner): Likewise.
14758 (rs6000_preferred_reload_class): Likewise.
14759 (rs6000_output_move_128bit): Likewise.
14761 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14763 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
14764 Generate SImode target register for null target.
14765 <case IX86_BUILTIN_XGETBV>: Ditto.
14766 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
14767 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
14769 2018-05-10 Carl Love <cel@us.ibm.com>
14771 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14772 dcbtt and dcbtstt if operands[2] is 0.
14774 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14777 * config/i386/sse.md (usadv64qi): New expander.
14779 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
14781 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
14782 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
14783 -maltivec=be support.
14784 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
14785 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
14786 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
14787 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
14788 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
14789 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
14790 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
14791 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
14792 altivec_vsumsws): Adjust.
14793 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
14794 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
14795 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
14797 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
14798 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
14799 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
14800 (altivec_lve<VI_char>x): Delete expand.
14801 (*altivec_lve<VI_char>x_internal): Rename to...
14802 (altivec_lve<VI_char>x): ... this.
14803 (altivec_lvxl_<mode>): Delete expand.
14804 (*altivec_lvxl_<mode>_internal): Rename to ...
14805 (altivec_lvxl_<mode>): ... this.
14806 (altivec_stvxl_<mode>): Delete expand.
14807 (*altivec_stvxl_<mode>_internal): Rename to ...
14808 (altivec_stvxl_<mode>): ... this.
14809 (altivec_stve<VI_char>x): Delete expand.
14810 (*altivec_stve<VI_char>x_internal): Rename to ...
14811 (altivec_stve<VI_char>x): ... this.
14812 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
14813 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
14814 reduc_plus_scal_<mode>): Adjust.
14815 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
14817 (rs6000_cpu_cpp_builtins): Adjust.
14818 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
14819 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
14820 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
14821 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
14822 -maltivec=be support.
14823 (rs6000_split_vec_extract_var): Adjust.
14824 (rs6000_split_v4si_init): Adjust.
14825 (swap_selector_for_mode): Delete.
14826 (altivec_expand_lvx_be, altivec_expand_stvx_be,
14827 altivec_expand_stvex_be): Delete.
14828 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
14829 -maltivec=be support.
14830 (rs6000_gimple_fold_builtin): Ditto.
14831 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
14833 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
14834 (TARGET_DIRECT_MOVE_64BIT): Adjust.
14835 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
14836 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
14837 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
14838 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
14839 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
14840 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
14841 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
14842 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
14843 anonymous split): Adjust.
14844 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
14845 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
14847 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
14849 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
14850 when --with-gxx-include-dir is also specified.
14851 * configure: Regenerate.
14853 2018-05-09 Jim Wilson <jimw@sifive.com>
14856 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
14857 * config/riscv/t-withmultilib: New.
14858 * config/riscv/withmultilib.h: New.
14859 * doc/install.texi: Document RISC-V --with-multilib-list support.
14861 2018-05-09 Richard Biener <rguenther@suse.de>
14863 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
14865 (vect_bb_vectorization_profitable_p): Adjust. Compute
14866 actual scalar cost using the cost vector and the add_stmt_cost
14869 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14871 PR rtl-optimization/85645
14872 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
14873 in the REG_CFA_REGISTER note for LR, don't leave it empty.
14875 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14877 PR rtl-optimization/85645
14878 * shrink-wrap.c (spread_components): Return a boolean saying if
14879 anything was changed.
14880 (try_shrink_wrapping_separate): Iterate spread_components until
14881 nothing changes anymore.
14883 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14885 PR rtl-optimization/85645
14886 * regrename.c (build_def_use): Also kill the chains that include the
14887 destination of a REG_CFA_REGISTER note.
14889 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14891 PR rtl-optimization/85645
14892 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
14893 insn that has a REG_CFA_REGISTER note.
14895 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
14897 * cfgexpand.c (expand_clobber): New function.
14898 (expand_gimple_stmt_1): Use it.
14899 * tree-vect-stmts.c (vect_clobber_variable): New function,
14901 (vectorizable_simd_clone_call): ...here.
14902 (vectorizable_store): Emit a clobber either side of an
14903 IFN_STORE_LANES sequence.
14904 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
14906 2018-05-09 Tom de Vries <tom@codesourcery.com>
14909 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
14910 (define_insn "trap_if_false"): Add exit after trap.
14912 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
14914 PR rtl-optimization/85638
14915 * bb-reorder.c: Include common/common-target.h.
14916 (create_forwarder_block): New function extracted from...
14917 (fix_up_crossing_landing_pad): ...here. Rename into...
14918 (dw2_fix_up_crossing_landing_pad): ...this.
14919 (sjlj_fix_up_crossing_landing_pad): New function.
14920 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
14921 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
14922 from both partitions and exit the loop after one iteration.
14924 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14927 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14929 (Basic PowerPC Built-in Functions): The new name of the
14930 subsection previously known as "PowerPC Built-in Functions".
14931 (Basic PowerPC Built-in Functions Available on all Configurations):
14933 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14935 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14937 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14939 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14942 2018-05-08 Jim Wilson <jimw@sifive.com>
14944 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
14945 (LD_EMUL_SUFFIX): New.
14946 (LINK_SPEC): Use it.
14948 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14950 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14952 (Basic PowerPC Built-in Functions): The new name of the
14953 subsection previously known as "PowerPC Built-in Functions".
14954 (Basic PowerPC Built-in Functions Available on all Configurations):
14956 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14958 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14960 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14962 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14965 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14968 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
14969 after cmpelim optimization.
14971 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
14973 * config.gcc: Support "goldmont".
14974 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
14975 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14976 PROCESSOR_GOLDMONT.
14977 * config/i386/i386.c (m_GOLDMONT): Define.
14978 (processor_target_table): Add "goldmont".
14979 (PTA_GOLDMONT): Define.
14980 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
14981 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
14982 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
14983 (fold_builtin_cpu): Add "goldmont".
14984 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
14985 (ix86_option_override_internal): Add "goldmont".
14986 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
14987 (processor_type): Add PROCESSOR_GOLDMONT.
14988 * config/i386/i386.md: Add CPU "glm".
14989 * config/i386/glm.md: New file.
14990 * config/i386/x86-tune.def: Add m_GOLDMONT.
14991 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
14993 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14996 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
14998 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
14999 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
15000 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
15003 * config/i386/i386.c (ix86_fold_builtin): Handle
15004 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15007 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15008 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15010 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
15013 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15014 (check_arch): Likewise.
15015 (check_fpu): Return the result rather than printing it.
15016 (end arch): Fix operator precedence.
15017 (end cpu): Likewise.
15018 (END): Print the result from check_fpu.
15020 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15021 Alan Hayward <alan.hayward@arm.com>
15022 David Sherwood <david.sherwood@arm.com>
15024 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15025 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15026 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15027 (*fcmuo<mode>_and): New patterns.
15029 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15031 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15032 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15033 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15034 (cmp_op, sve_imm_con): New code attributes.
15035 (SVE_COND_INT_CMP, imm_con): Delete.
15036 (cmp_op): Remove above unspecs from int attribute.
15037 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15039 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
15040 comparison-specific unspecs.
15041 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15042 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15043 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15044 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15045 (*vec_fcm<cmp_op><mode>): Rename to...
15046 (*fcm<cmp_op><mode>): ...this and adjust likewise.
15047 (*vec_fcmuo<mode>): Rename to...
15048 (*fcmuo<mode>): ...this and adjust likewise.
15049 (*pred_fcm<cmp_op><mode>): New pattern.
15050 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15051 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15053 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15055 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15056 (aarch64_emit_sve_predicated_cond): New function.
15057 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15058 (aarch64_emit_unspec_cond_or): Replace with...
15059 (aarch64_emit_sve_or_conds): ...this new function. Use
15060 aarch64_emit_sve_ptrue_op for the individual comparisons and
15061 aarch64_emit_binop to OR them together.
15062 (aarch64_emit_inverted_unspec_cond): Replace with...
15063 (aarch64_emit_sve_inverted_cond): ...this new function. Use
15064 aarch64_emit_sve_ptrue_op for the comparison and
15065 aarch64_emit_unop to invert the result.
15066 (aarch64_expand_sve_vec_cmp_float): Update after the above
15067 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
15069 2018-05-07 Nathan Sidwell <nathan@acm.org>
15071 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15072 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15073 (Backwards Compatibility): Likewise.
15075 2018-05-07 Luis Machado <luis.machado@linaro.org>
15079 2018-05-07 Luis Machado <luis.machado@linaro.org>
15081 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15082 <prefetch_dynamic_strides>: New const bool field.
15083 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15084 prefetch_dynamic_strides.
15085 (exynosm1_prefetch_tune): Likewise.
15086 (thunderxt88_prefetch_tune): Likewise.
15087 (thunderx_prefetch_tune): Likewise.
15088 (thunderx2t99_prefetch_tune): Likewise.
15089 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15091 (aarch64_override_options_internal): Update to set
15092 PARAM_PREFETCH_DYNAMIC_STRIDES.
15093 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15094 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15095 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15096 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15097 prefetch-dynamic-strides setting.
15099 2018-05-07 Luis Machado <luis.machado@linaro.org>
15101 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15102 <minimum_stride>: New const int field.
15103 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15104 minimum_stride field.
15105 (exynosm1_prefetch_tune): Likewise.
15106 (thunderxt88_prefetch_tune): Likewise.
15107 (thunderx_prefetch_tune): Likewise.
15108 (thunderx2t99_prefetch_tune): Likewise.
15109 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15110 (aarch64_override_options_internal): Update to set
15111 PARAM_PREFETCH_MINIMUM_STRIDE.
15112 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15113 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15114 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15115 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15116 stride is constant and is below the minimum stride threshold.
15118 2018-05-07 Luis Machado <luis.machado@linaro.org>
15120 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15123 2018-05-07 Luis Machado <luis.machado@linaro.org>
15125 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15126 <prefetch_dynamic_strides>: New const bool field.
15127 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15128 prefetch_dynamic_strides.
15129 (exynosm1_prefetch_tune): Likewise.
15130 (thunderxt88_prefetch_tune): Likewise.
15131 (thunderx_prefetch_tune): Likewise.
15132 (thunderx2t99_prefetch_tune): Likewise.
15133 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15135 (aarch64_override_options_internal): Update to set
15136 PARAM_PREFETCH_DYNAMIC_STRIDES.
15137 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15138 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15139 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15140 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15141 prefetch-dynamic-strides setting.
15143 2018-05-07 Luis Machado <luis.machado@linaro.org>
15145 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15146 <minimum_stride>: New const int field.
15147 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15148 minimum_stride field.
15149 (exynosm1_prefetch_tune): Likewise.
15150 (thunderxt88_prefetch_tune): Likewise.
15151 (thunderx_prefetch_tune): Likewise.
15152 (thunderx2t99_prefetch_tune): Likewise.
15153 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15154 (aarch64_override_options_internal): Update to set
15155 PARAM_PREFETCH_MINIMUM_STRIDE.
15156 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15157 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15158 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15159 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15160 stride is constant and is below the minimum stride threshold.
15162 2018-05-06 Jakub Jelinek <jakub@redhat.com>
15165 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15166 the type is addressable. Don't force op into register if it has
15169 2018-05-05 Roland McGrath <mcgrathr@google.com>
15172 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15173 any section for which we don't know a specific type it should have,
15174 regardless of name. Previously this was done only for the exact
15175 names ".init_array", ".fini_array", and ".preinit_array".
15176 (default_elf_asm_named_section): Add comment about
15177 relationship with default_section_type_flags and SECTION_NOTYPE.
15178 (get_section): Don't consider it a type conflict if one side has
15179 SECTION_NOTYPE and the other doesn't, as long as neither has the
15180 SECTION_BSS et al used in the default_section_type_flags logic.
15182 2018-05-05 Tom de Vries <tom@codesourcery.com>
15185 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15186 (workaround_barsyncs): New function.
15187 (nvptx_reorg): Use workaround_barsyncs.
15188 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15189 (define_expand "nvptx_membar_cta"): New define_expand.
15190 (define_insn "*nvptx_membar_cta"): New insn.
15192 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
15194 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15195 To improve optimization opportunities.
15196 * builtin-types.def: The new needed builtin types for the above.
15198 2018-05-04 Richard Biener <rguenther@suse.de>
15200 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15201 * gimple-ssa-store-merging.c
15202 (imm_store_chain_info::output_merged_store): Remove redundant create,
15203 release split_store vector contents on failure.
15204 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15205 scalar stmt vector on cache hit.
15207 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
15209 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15211 * config.gcc (powerpc-xilinx-eabi*): Remove.
15212 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15214 (fusion_addis_mem_combo_load): Ditto.
15215 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15217 (rs6000_cpu_cpp_builtins): Ditto.
15218 * config/rs6000/rs6000-linux.c
15219 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15220 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15221 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15223 (rs6000_setup_reg_addr_masks): Ditto.
15224 (rs6000_init_hard_regno_mode_ok): Ditto.
15225 (rs6000_option_override_internal): Ditto.
15226 (legitimate_lo_sum_address_p): Ditto.
15227 (rs6000_legitimize_address): Ditto.
15228 (rs6000_legitimize_reload_address): Ditto.
15229 (rs6000_legitimate_address_p): Ditto.
15230 (abi_v4_pass_in_fpr): Ditto.
15231 (setup_incoming_varargs): Ditto.
15232 (rs6000_gimplify_va_arg): Ditto.
15233 (rs6000_split_multireg_move): Ditto.
15234 (rs6000_savres_strategy): Ditto.
15235 (rs6000_emit_prologue_components): Ditto.
15236 (rs6000_emit_epilogue_components): Ditto.
15237 (rs6000_emit_prologue): Ditto.
15238 (rs6000_emit_epilogue): Ditto.
15239 (rs6000_elf_file_end): Ditto.
15240 (rs6000_function_value): Ditto.
15241 (rs6000_libcall_value): Ditto.
15242 * config/rs6000/rs6000.h: Ditto.
15243 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15244 (TARGET_MINMAX): ... this. New.
15245 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15246 * config/rs6000/rs6000.md: Remove Xilinx FP support.
15247 (*movsi_internal1_single): Delete.
15248 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15249 mfpu=, mxilinx-fpu): Delete.
15250 * config/rs6000/singlefp.h: Delete.
15251 * config/rs6000/sysv4.h: Remove Xilinx FP support.
15252 * config/rs6000/t-rs6000: Ditto.
15253 * config/rs6000/t-xilinx: Delete.
15254 * config/rs6000/titan.md: Adjust for fp_type removal.
15255 * config/rs6000/vsx.md: Remove Xilinx FP support.
15256 (VStype_simple): Delete.
15257 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15258 * config/rs6000/xfpu.h: Delete.
15259 * config/rs6000/xfpu.md: Delete.
15260 * config/rs6000/xilinx.h: Delete.
15261 * config/rs6000/xilinx.opt: Delete.
15262 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15263 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15265 2018-05-04 Tom de Vries <tom@codesourcery.com>
15268 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15271 2018-05-04 Richard Biener <rguenther@suse.de>
15273 PR middle-end/85627
15274 * tree-complex.c (update_complex_assignment): We are always in SSA form.
15275 (expand_complex_div_wide): Likewise.
15276 (expand_complex_operations_1): Likewise.
15277 (expand_complex_libcall): Preserve EH info of the original stmt.
15278 (tree_lower_complex): Handle removed blocks.
15279 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15280 on complex multiplication and division libcall builtins.
15282 2018-05-04 Richard Biener <rguenther@suse.de>
15284 PR middle-end/85574
15285 * fold-const.c (negate_expr_p): Restrict negation of operand
15286 zero of a division to when we know that can happen without
15288 (fold_negate_expr_1): Likewise.
15290 2018-05-04 Jakub Jelinek <jakub@redhat.com>
15293 * real.h (real_nextafter): Declare.
15294 * real.c (real_nextafter): New function.
15295 * fold-const-call.c (fold_const_nextafter): New function.
15296 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15297 CASE_CFN_NEXTTOWARD.
15298 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15299 even when arg1_mode is different from arg0_mode.
15301 2018-05-03 Nathan Sidwell <nathan@acm.org>
15303 * doc/extend.texi (Deprecated Features): Remove
15304 -ffriend-injection.
15305 (Backwards Compatibility): Likewise.
15306 * doc/invoke.texi (C++ Language Options): Likewise.
15307 (C++ Dialect Options): Likewise.
15309 2018-05-03 Jakub Jelinek <jakub@redhat.com>
15312 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15313 _mm512_mask_mullox_epi64): New intrinsics.
15315 2018-05-03 Tom de Vries <tom@codesourcery.com>
15318 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15319 dump files): Add offload-tree.
15321 2018-05-03 Richard Biener <rguenther@suse.de>
15323 PR tree-optimization/85615
15324 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15325 to loops not nested in BBs loop father to avoid creating multi-entry
15328 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15330 PR tree-optimization/70291
15331 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15332 arguments. Change return type to tree. Emit libcall as a new
15333 statement rather than replacing existing one when inplace_p is true.
15334 (expand_complex_multiplication_components): New function.
15335 (expand_complex_multiplication): Expand floating-point complex
15336 multiplication using the above.
15337 (expand_complex_division): Rename inner_type parameter to type.
15338 Update expand_complex_libcall call-site.
15339 (expand_complex_operations_1): Update expand_complex_multiplication
15340 and expand_complex_division call-sites.
15342 2018-05-02 Jakub Jelinek <jakub@redhat.com>
15345 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15346 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15347 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
15348 the highest significant bit of the shift count mask is clear. In
15349 check whether and[sq]i3 is needed verify that all significant bits
15350 of the shift count other than the highest are set.
15352 2018-05-02 Tom de Vries <tom@codesourcery.com>
15355 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
15356 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
15357 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
15358 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
15359 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
15360 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
15361 __builtin_goacc_parlevel_size.
15363 2018-05-02 Richard Biener <rguenther@suse.de>
15365 PR tree-optimization/85597
15366 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
15367 do not use split vect_get_vec_defs call but call vect_get_slp_defs
15370 2018-05-02 Tom de Vries <tom@codesourcery.com>
15373 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15374 dump files): Add ltrans-tree.
15376 2018-05-02 Tom de Vries <tom@codesourcery.com>
15379 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15380 dump files): Add wpa-ipa.
15382 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
15384 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
15385 powerpc*-*-linux*paired* target.
15386 * config/rs6000/750cl.h: Delete.
15387 * config/rs6000/paired.h: Delete.
15388 * config/rs6000/paired.md: Delete.
15389 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
15391 * config/rs6000/rs6000-builtin.def: Remove paired float support.
15392 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
15393 comment. Remove paired float support.
15394 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
15395 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
15397 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
15398 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
15400 * config/rs6000/rs6000.c: Remove paired float support.
15401 (paired_expand_vector_init, paired_expand_vector_move,
15402 paired_emit_vector_compare, paired_emit_vector_cond_expr,
15403 (paired_expand_lv_builtin, paired_expand_stv_builtin,
15404 paired_expand_builtin, paired_expand_predicate_builtin,
15405 paired_init_builtins): Delete.
15406 * config/rs6000/rs6000.h: Remove paired float support.
15407 * config/rs6000/rs6000.md: Remove paired float support.
15408 (move_from_CR_ov_bit): Delete.
15409 * config/rs6000/rs6000.opt (mpaired): Delete.
15410 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
15411 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
15413 2018-05-02 Richard Biener <rguenther@suse.de>
15415 PR middle-end/85567
15416 * gimplify.c (gimplify_save_expr): When in SSA form allow
15417 SAVE_EXPRs to compute to SSA vars.
15419 2018-05-02 Jakub Jelinek <jakub@redhat.com>
15422 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15423 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15424 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
15425 clobber operands[2], instead use a new pseudo. Formatting fixes.
15427 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
15429 PR tree-optimization/85586
15430 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
15431 exit early for statements in the same group if the accesses are
15434 2018-05-02 Tom de Vries <tom@codesourcery.com>
15437 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
15440 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
15442 PR tree-optimization/85143
15443 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
15445 2018-05-01 Tom de Vries <tom@codesourcery.com>
15448 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
15449 not found" error message.
15451 2018-05-01 Tom de Vries <tom@codesourcery.com>
15454 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
15455 * vec.c (test_ordered_remove_if): New function.
15456 (vec_c_tests): Call test_ordered_remove_if.
15457 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
15458 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
15459 * tree-vect-patterns.c (vect_pattern_recog_1): Use
15460 VEC_ORDERED_REMOVE_IF.
15462 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15464 PR tree-optimization/82665
15465 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
15466 pointer subtraction where arguments come from a memchr call.
15468 2018-05-01 Jakub Jelinek <jakub@redhat.com>
15470 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
15471 --push-state --as-needed and --pop-state instead of --as-needed and
15472 --no-as-needed if ld supports it.
15473 * configure: Regenerated.
15476 * doc/install.texi2html: Replace _002d with - and _002a with * in
15477 generated html files using sed.
15479 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15482 * gcc-rich-location.c (blank_line_before_p): New function.
15483 (use_new_line): New function.
15484 (gcc_rich_location::add_fixit_insert_formatted): New function.
15485 * gcc-rich-location.h
15486 (gcc_rich_location::add_fixit_insert_formatted): New function.
15488 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15490 * selftest.c (assert_streq): Rename "expected" and "actual" to
15491 "val1" and "val2". Extend NULL-handling to cover both inputs
15492 symmetrically, while still requiring both to be non-NULL for a pass.
15493 * selftest.h (assert_streq): Rename "expected" and "actual" to
15495 (ASSERT_EQ): Likewise.
15496 (ASSERT_EQ_AT): Likewise.
15497 (ASSERT_KNOWN_EQ): Likewise.
15498 (ASSERT_KNOWN_EQ_AT): Likewise.
15499 (ASSERT_NE): Likewise.
15500 (ASSERT_MAYBE_NE): Likewise.
15501 (ASSERT_MAYBE_NE_AT): Likewise.
15502 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
15503 the assertion to pass.
15504 (ASSERT_STREQ_AT): Likewise.
15506 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
15508 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
15509 interaction with -pie.
15511 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15513 * selftest.h: Fix alphabetization of per-source-file selftest
15516 2018-04-30 Jason Merrill <jason@redhat.com>
15518 PR c++/61982 - dead stores to destroyed objects.
15519 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
15522 2018-04-30 Jason Merrill <jason@redhat.com>
15524 * tree.c (build_clobber): New.
15525 * tree.h: Declare it.
15526 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
15528 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15530 * diagnostic-show-locus.c (layout::layout): Update for
15531 location_get_source_line returning a char_span.
15532 (struct char_span): Move to input.h.
15533 (struct correction): Update for fields in char_span becoming
15535 (struct source_line): Update for location_get_source_line
15536 returning a char_span.
15537 (layout::print_line): Likewise.
15538 * edit-context.c (edited_file::print_content): Likewise.
15539 (edited_file::print_diff_hunk): Likewise.
15540 (edited_file::print_run_of_changed_lines): Likewise.
15541 (edited_file::get_num_lines): Likewise.
15542 (edited_line::edited_line): Likewise.
15543 * final.c (asm_show_source): Likewise.
15544 * input.c (location_get_source_line): Convert return type
15545 from const char * to char_span, losing the final "line_len"
15547 (dump_location_info): Update for the above.
15548 (get_substring_ranges_for_loc): Likewise. Use a char_span
15549 when handling the literal within the line.
15550 (test_reading_source_line): Update for location_get_source_line
15551 returning a char_span.
15552 * input.h (class char_span): Move here from
15553 diagnostic-show-locus.c, converting from a struct to a class.
15554 Make data members private.
15555 (char_span::operator bool): New.
15556 (char_span::length): New.
15557 (char_span::get_buffer): New.
15558 (char_span::operator[]): New.
15559 (char_span::subspan): Make const.
15560 (char_span::xstrdup): New.
15561 (location_get_source_line): Convert return type from const char *
15562 to char_span, losing the final "line_size" param.
15564 2018-04-30 Jan Hubicka <jh@suse.cz>
15566 * lto-wrapper.c (ltrans_priorities): New static var.
15567 (cmp_priority): New.
15568 (run_gcc): Read priorities and if doing parallel build order
15569 the Makefile by them.
15571 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15573 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
15575 2018-04-30 Richard Biener <rguenther@suse.de>
15577 * tree-cfg.c (verify_address): Remove base argument, add
15578 flag whether to check TREE_ADDRESSABLE and do that.
15579 (verify_expr): Remove.
15580 (verify_types_in_gimple_reference): Add pieces from verify_expr.
15581 (verify_gimple_assign_single): Likewise.
15582 (verify_gimple_switch): Likewise.
15583 (verify_expr_location_1): Dereference tp once. Add (disabled)
15584 piece from verify_expr.
15585 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
15587 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15589 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
15591 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15593 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
15594 (small_data_pattern): Likewise.
15595 (arc_rewrite_small_data): Likewise.
15596 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
15597 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
15598 (get_symbol_alignment): New function.
15599 (legitimate_small_data_address_p): Likewise.
15600 (legitimate_scaled_address): Update, call
15601 legitimate_small_data_address_p.
15602 (output_sdata): New static variable.
15603 (arc_print_operand): Update how we handle small data operands.
15604 (arc_print_operand_address): Likewise.
15605 (arc_legitimate_address_p): Update, use
15606 legitimate_small_data_address_p.
15607 (arc_rewrite_small_data_p): Remove.
15608 (arc_rewrite_small_data_1): Likewise.
15609 (arc_rewrite_small_data): Likewise.
15610 (small_data_pattern): Likewise.
15611 (compact_sda_memory_operand): Update to use
15612 legitimate_small_data_address_p and get_symbol_alignment.
15613 (prepare_move_operands): Don't rewite sdata pattern.
15614 (prepare_extend_operands): Remove.
15615 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
15617 (zero_extendqisi2): Likewise.
15618 (zero_extendhisi2): Likewise.
15619 (extendqihi2): Likewise.
15620 (extendqisi2): Likewise.
15621 (extendhisi2): Likewise.
15622 (addsi3): Likewise.
15623 (subsi3): Likewise.
15624 (andsi3): Likewise.
15625 * config/arc/constraints.md (Usd): Change it to memory constraint.
15627 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15629 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
15630 as source of std instructions.
15631 * config/arc/arc.md (movsi_insn): Update pattern predicate to
15632 allow 6-bit constants as source for store instructions.
15633 (movdi_insn): Update instruction pattern to allow 6-bit constants
15634 as source for store instructions.
15636 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
15638 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
15640 2018-04-30 Nathan Sidwell <nathan@acm.org>
15641 Sandra Loosemore <sandra@codesourcery.com>
15643 * dumpfile.c (dump_open): Allow '-' for stdout.
15644 * doc/invoke.texi (Developer Options): Document dump filename
15645 determination early. Document stdin/stdout selection.
15647 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
15649 Microblaze Target: PIC data text relative
15651 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
15652 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
15654 * config/microblaze/microblaze.h (microblaze_constant_address_p):
15655 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
15656 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
15657 New addressing mode for data-text relative position indepenedent code.
15658 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
15659 'ADDRESS_SYMBOLIC_TXT_REL'.
15660 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
15661 (microblaze_legitimate_pic_operand): Exclude function calls from
15662 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
15663 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
15665 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15666 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15667 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
15668 for 'address + offset'.
15669 (microblaze_expand_prologue): Add new function prologue call for
15671 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
15672 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
15673 table in case of TARGET_PIC_DATA_TEXT_REL.
15674 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
15675 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
15676 Add new macros 'UNSPEC_TEXT',
15677 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
15678 + exclude function calls from 'UNSPEC_PLT' in case of data text
15680 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15681 new target hook for generating address diff vector tables in case of
15683 * doc/tm.texi : Regenerate.
15684 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
15685 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
15686 of addr diff vector generation.
15687 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15688 target hook definition.
15689 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
15690 Add default function for generate_pic_addr_diff_vec -> flag_pic.
15691 * doc/invoke.texi (Add new pic option): Add new microblaze pic
15692 option for data text relative.
15694 2018-04-30 Richard Biener <rguenther@suse.de>
15696 * tree-chrec.h (evolution_function_is_constant_p): Remove
15698 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
15700 2018-04-30 Richard Biener <rguenther@suse.de>
15703 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
15705 2018-04-30 Richard Biener <rguenther@suse.de>
15707 PR tree-optimization/28364
15708 PR tree-optimization/85275
15709 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
15710 copying first exit test.
15712 2018-04-28 Mark Wielaard <mark@klomp.org>
15714 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
15715 dwarf_version >= 5.
15716 (dwarf_AT): Handle DW_AT_addr_base.
15717 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
15719 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
15722 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
15723 (*ashl<dwi>3_doubleword_mask_1): Ditto.
15724 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
15725 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
15727 2018-04-28 Richard Biener <rguenther@suse.de>
15729 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
15730 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
15731 to reflect use. Only add interesting stmts.
15733 2018-04-27 Martin Jambor <mjambor@suse.cz>
15736 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
15737 the jump function allows for passing through aggregate values.
15739 2018-04-27 David Malcolm <dmalcolm@redhat.com>
15741 * input.h (in_system_header_at): Convert from macro to inline
15743 (from_macro_expansion_at): Likewise.
15744 (from_macro_definition_at): Likewise.
15746 2018-04-27 Jeff Law <law@redhat.com>
15748 * config.gcc: Mark tile* targets as deprecated/obsolete.
15750 2018-04-27 Richard Biener <rguenther@suse.de>
15752 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
15755 2018-04-27 Richard Biener <rguenther@suse.de>
15757 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
15759 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
15761 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
15762 with Yd constraint. Set "preferred_for_speed" attribute from
15763 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
15764 with Yd constraint.
15765 (*movdi_internal): Ditto.
15766 (movti_interunit splitters): Remove
15767 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
15768 (movdi_interunit splitters): Ditto.
15769 * config/i386/constraints.md (Ye): Remove.
15770 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
15772 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15775 * config/aarch64/constraints.md (Usg): Limit to 31.
15776 (Usj): Limit to 63.
15778 2018-04-27 Jakub Jelinek <jakub@redhat.com>
15780 PR tree-optimization/85529
15781 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
15782 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
15783 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
15784 zero extension or masking of the MSB bit.
15785 (optimize_range_tests): Add FIRST_BB argument, pass it through
15786 to optimize_range_tests_var_bound.
15787 (maybe_optimize_range_tests, reassociate_bb): Adjust
15788 optimize_range_tests callers.
15790 2018-04-26 Richard Biener <rguenther@suse.de>
15791 Jakub Jelinek <jakub@redhat.com>
15793 * cgraph.h (symbol_table): Just declare debug method here.
15794 * symtab.c (symbol_table::debug): Define.
15796 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
15798 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
15800 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
15802 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
15803 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
15804 (*movdi_internal): Substitute Yi and Yj constraint with x
15805 and Ym and Yn constraint with y constraint. Update "isa"
15806 attribute and set "preferred_for_speed" attribute from
15807 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
15808 (*movsi_internal): Ditto.
15809 (*movdf_internal): Ditto.
15810 (*movsf_internal): Ditto.
15811 (*zero_extendsidi2): Ditto.
15812 * config/i386/sse.md (vec_set<mode>_0): Ditto.
15813 (sse2_loadld): Ditto.
15814 (*vec_extract<ssevecmodelower>_0): Ditto.
15815 (*vec_extractv4si_0_zext_sse4): Ditto.
15816 (vec_concatv2di): Ditto.
15817 (*vec_dup<mode>): Ditto.
15818 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
15819 * config/i386/constraints.md (Yi): Remove.
15824 2018-04-26 Nathan Sidwell <nathan@acm.org>
15826 * dumpfile.c (dump_open): New.
15827 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
15828 (dump_finish): Detect stdio/stderr by value not name.
15830 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
15832 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
15834 2018-04-26 Tom de Vries <tom@codesourcery.com>
15837 * config/nvptx/nvptx.c (verify_neutering_jumps)
15838 (verify_neutering_labels): New function
15839 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
15841 2018-04-26 Tom de Vries <tom@codesourcery.com>
15844 * config/nvptx/nvptx.c (needs_neutering_p): New function.
15845 (nvptx_single): Use needs_neutering_p to skip over insns that do not
15848 2018-04-26 Richard Biener <rguenther@suse.de>
15849 Tom de Vries <tom@codesourcery.com>
15852 * lto-streamer-out.c (output_function): Fixup loops if required to match
15853 discovery done in the reader.
15855 2018-04-26 Richard Biener <rguenther@suse.de>
15857 PR tree-optimization/85116
15858 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
15859 have a loop exit from the single latch predecessor. Remove
15860 case of header with just condition.
15861 (ch_base::copy_headers): Exclude infinite loops from any
15863 (pass_ch::execute): Record exits.
15865 2018-04-26 Richard Biener <rguenther@suse.de>
15867 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
15868 prologue cost vector and pass it to vect_get_load_cost.
15869 (vect_get_peeling_costs_all_drs): Likewise.
15870 (vect_peeling_hash_get_lowest_cost): Likewise.
15871 (vect_enhance_data_refs_alignment): Likewise.
15873 2018-04-26 Richard Biener <rguenther@suse.de>
15875 PR middle-end/85450
15876 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
15877 checking of integer<->pointer conversions.
15878 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
15879 sign-/zero-extending pointer types.
15880 (expand_omp_for_static_chunk): Likewise.
15882 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
15883 Jean Lee <xiaoyur347@gmail.com>
15885 * config/mips/mips.c (mips_asan_shadow_offset): New function.
15886 (TARGET_ASAN_SHADOW_OFFSET): Define.
15887 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
15888 true for -fsanitize=address.
15890 2018-04-25 Mark Wielaard <mark@klomp.org>
15892 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
15895 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15897 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
15898 than "alu", remove explicit "memory" and "imm_disp" attributes.
15899 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
15901 PR middle-end/85414
15902 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
15903 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
15904 gen_lowpart_no_emit.
15906 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
15909 * config/i386/i386.c (ix86_expand_builtin): Change memory
15910 operand to XI, extend p0 to Pmode.
15911 * config/i386/i386.md: Change unspec volatile and operand
15912 1 mode to XI, change operand 0 mode to P.
15914 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15916 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
15917 GET_MODE_MASK before any checking.
15918 (nds32_can_use_bset_p): Likewise.
15919 (nds32_can_use_btgl_p): Likewise.
15921 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15923 * config/nds32/nds32-doubleword.md: New define_split pattern for
15924 illegal register number.
15926 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15928 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
15930 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15932 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
15934 2018-04-25 Richard Biener <rguenther@suse.de>
15936 * lto-streamer.h (LTO_major_version): Bump to 8.
15938 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15940 * BASE-VER: Set to 9.0.0.
15942 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
15944 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
15945 in __abskf2 and __powikf2.
15947 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15950 * config/aarch64/constraints.md (Usg, Usj): New constraints.
15951 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
15952 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
15953 Use the above on operand 2. Reindent.
15954 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
15956 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15959 * common/config/i386/i386-common.c (ix86_handle_option): Don't
15961 * config/i386/i386.opt (mcet): Removed.
15962 * doc/install.texi: Remove -mcet documentation.
15963 * doc/invoke.texi: Likewise.
15965 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15968 * doc/install.texi: Remove -mcet from bootstrap-cet.
15970 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15973 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
15974 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
15978 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
15979 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
15980 containing a CONST_VECTOR.
15982 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
15984 * doc/install.texi: Update newlib dependency for nvptx.
15986 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15989 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
15990 instead of INTVAL when shifting x left.
15992 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
15994 PR tree-optimization/85478
15995 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
15996 vect_grouped_store_supported for single element vectors.
15998 2018-04-24 Richard Biener <rguenther@suse.de>
16001 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16002 load cost increase to the case of non-constant step.
16004 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16007 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16008 destination if any_malformed_asm.
16010 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
16012 PR middle-end/85496
16013 * expr.c (store_field): In the bitfield case, if the value comes from
16014 a function call and is returned in registers by means of a PARALLEL,
16015 do not change the mode of the temporary unless BLKmode and VOIDmode.
16017 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
16019 PR rtl-optimization/85423
16020 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16021 dependencies to debug insns when the previous insn is non-debug.
16023 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16025 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16026 enums into a single definition.
16027 (fls): Fix predicates and printing.
16030 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16032 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16033 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16034 and short u6 immediate.
16035 (check_if_valid_sleep_operand): Remove.
16036 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16038 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16040 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16041 flag_always_save_lp condition.
16042 * config/nds32/nds32.opt (malways-save-lp): New option.
16044 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16046 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16047 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16048 * config/nds32/nds32.h
16049 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16050 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16052 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16054 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16055 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16057 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16058 Chung-Ju Wu <jasonwucj@gmail.com>
16060 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16062 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16063 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16065 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16067 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16069 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16071 * config/nds32/nds32-protos.h (nds32_data_alignment,
16072 nds32_local_alignment): Declare.
16073 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16074 nds32_local_alignment): New functions.
16075 (TARGET_CONSTANT_ALIGNMENT): Define.
16076 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16078 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16080 * config/nds32/nds32.c
16081 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16082 (TARGET_MODES_TIEABLE_P): Likewise.
16084 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16086 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16087 level Ofast and Og.
16089 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
16090 Chung-Ju Wu <jasonwucj@gmail.com>
16092 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16093 for unaligned access.
16094 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16096 * config/nds32/nds32-intrinsic.md: Likewise.
16097 * config/nds32/nds32_intrinsic.h: Likewise.
16098 * config/nds32/nds32.h (nds32_builtins): Likewise.
16099 * config/nds32/nds32.opt (munaligned-access): New option.
16100 * config/nds32/nds32.c (nds32_asm_file_start): Display
16101 flag_unaligned_access status.
16103 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16105 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16106 -mno-relax is present.
16107 * config/riscv/linux.h (LINK_SPEC): Ditto.
16109 2018-04-20 Martin Sebor <msebor@redhat.com>
16112 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16114 (gimple_fold_builtin_stxcpy_chk): Same.
16115 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16117 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
16120 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16121 __powikf2 when long double is IEEE 128-bit.
16123 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16125 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16126 step to make sure stack always aligned.
16128 2018-04-20 Carl Love <cel@us.ibm.com>
16131 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16132 size check for arg0.
16134 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
16135 Tom de Vries <tom@codesourcery.com>
16138 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16139 Emit insns for calls too.
16140 (nvptx_find_par): Always look for worker-level predecessor insn.
16141 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
16143 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16144 (nvptx_process_pars): Propagate frames for calls.
16146 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
16149 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16151 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16152 (ix86_handle_option): Don't handle OPT_mibt.
16153 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16155 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16157 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16158 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16159 (ix86_target_macros): Define __CET__ with flag_cf_protection
16160 for -fcf-protection.
16161 * config/i386/i386.c (isa2_opts): Remove -mibt.
16162 * config/i386/i386.h (TARGET_IBT): Removed.
16163 (TARGET_IBT_P): Likewise.
16164 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16165 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16166 * config/i386/i386.opt (mcet): Update help message.
16167 (mshstk): Likewise.
16169 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
16170 -mcet as an alias for -mshstk.
16172 2018-04-20 Richard Biener <rguenther@suse.de>
16174 PR middle-end/85475
16175 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16176 complexity by forcing a single use of the multiply operand.
16178 2018-04-20 Martin Jambor <mjambor@suse.cz>
16181 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16182 recursion dependency to only apply to non-clones.
16184 2018-04-20 Martin Jambor <mjambor@suse.cz>
16187 * ipa-cp.c (create_specialized_node): Check that clones of
16188 self-recursive edges exist during IPA-CP.
16190 2018-04-19 Toon Moene <toon@moene.org>
16192 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16195 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16197 PR tree-optimization/85467
16198 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16199 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
16200 VECTOR_CST element to type.
16202 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16205 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16206 * config/i386/i386.md (builtin_setjmp_setup): Removed.
16207 (builtin_longjmp): Likewise.
16208 (save_stack_nonlocal): New pattern.
16209 (restore_stack_nonlocal): Likewise.
16211 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16214 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16215 Replace ASM_OUTPUT_LABEL with fprintf.
16217 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16220 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16221 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16222 * config/i386/i386-c.c (ix86_target_macros_internal): Also
16223 define __IBT__ and __SHSTK__ for -fcf-protection.
16224 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16226 (ix86_trampoline_init): Likewise.
16227 (x86_output_mi_thunk): Likewise.
16228 (ix86_notrack_prefixed_insn_p): Likewise.
16229 (ix86_option_override_internal): Don't disallow -fcf-protection.
16230 * config/i386/i386.md (rdssp<mode>): Also enable for
16232 (incssp<mode>): Likewise.
16233 (nop_endbr): Likewise.
16234 * config/i386/i386.opt (mcet): Change help message to built-in
16237 (mshstk): Likewise.
16238 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16239 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
16240 enable CET built-in functions.
16242 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
16244 * common/config/i386/i386-common.c
16245 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16246 OPTION_MASK_ISA_MOVDIRI_UNSET,
16247 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16248 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16249 * config.gcc (movdirintrin.h): New header.
16250 * config/i386/cpuid.h (bit_MOVDIRI,
16251 bit_MOVDIR64B): New bits.
16252 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16254 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16255 (VOID, PVOID, PCVOID)): New function types.
16256 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16257 __builtin_ia32_directstoreu_u64,
16258 __builtin_ia32_movdir64b): New builtins.
16259 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16260 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16262 (ix86_valid_target_attribute_inner_p): Ditto.
16263 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16264 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16265 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16266 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16267 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16268 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16269 (movdiri<mode>, movdir64b_<mode>): New.
16270 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16271 * config/i386/immintrin.h: Include movdirintrin.h.
16272 * config/i386/movdirintrin.h: New file.
16273 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16275 2018-04-19 Richard Biener <rguenther@suse.de>
16277 PR middle-end/85455
16278 * cfg.c (clear_bb_flags): When loop state says we have
16279 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16281 2018-04-19 Richard Biener <rguenther@suse.de>
16283 PR tree-optimization/84737
16284 * tree-vect-data-refs.c (vect_copy_ref_info): New function
16285 copying restrict info.
16286 (vect_setup_realignment): Use it.
16287 * tree-vectorizer.h (vect_copy_ref_info): Declare.
16288 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16289 the first DR to all generated stores.
16290 (vectorizable_load): Likewise for loads.
16292 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16294 PR tree-optimization/85446
16295 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16296 the integral and pointer types to have the same precision.
16298 * doc/install.texi: Document --disable-cet being the default and
16301 2018-04-18 Martin Liska <mliska@suse.cz>
16303 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16306 2018-04-18 Martin Liska <mliska@suse.cz>
16309 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16312 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16313 arguments if they are comparable.
16315 2018-04-18 Martin Liska <mliska@suse.cz>
16318 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16321 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16324 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
16327 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16328 ENDBR after calling __morestack.
16330 2018-04-18 David Malcolm <dmalcolm@redhat.com>
16333 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
16334 by using gcc_base_ver to generate a gcc_driver_version, and use
16335 it when generating GCC_DRIVER_NAME.
16336 * configure: Regenerate.
16338 2018-04-18 Jakub Jelinek <jakub@redhat.com>
16341 * config.gcc: Obsolete powerpc*-*-*spe*.
16343 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16346 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
16347 (stabstr_D): Change type of unum from unsigned int to
16348 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
16351 2018-04-17 Jim Wilson <jimw@sifive.com>
16354 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
16355 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
16356 Set arg_pointer_offset after using pretend_args_size.
16358 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16360 PR rtl-optimization/85431
16361 * dse.c (record_store): Ignore zero width stores.
16364 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
16365 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
16366 __builtin_stack_restore rather than after it.
16367 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
16368 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
16369 argument instead of virtual_dynamic_stack_rtx.
16371 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
16373 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
16375 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
16376 Add note to error message to explain internal mapping of overloaded
16377 built-in function name to non-overloaded built-in function name.
16378 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
16381 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16384 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
16385 where the inputs overlap with the output.
16387 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16389 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
16390 (=v, v) alternative and explicit "memory" attribute.
16391 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
16392 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16394 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
16395 "sselog1" type instead of "sselog".
16396 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
16397 "sselog". Remove explicit "memory" attribute.
16398 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
16399 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16401 (vec_extract_hi_v32hi): Merge all alternatives into one, use
16402 "sselog1" type instead of "sselog". Remove explicit "memory"
16404 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
16405 use "sselog1" type instead of "sselog". Remove explicit "memory"
16407 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
16408 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16410 (vec_extract_hi_v64qi): Merge all alternatives into one, use
16411 "sselog1" type instead of "sselog". Remove explicit "memory"
16413 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
16414 use "sselog1" type instead of "sselog". Remove explicit "memory"
16418 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
16420 PR middle-end/85414
16421 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
16424 2018-04-17 Martin Jambor <mjambor@suse.cz>
16427 * ipa-cp.c (create_specialized_node): Call
16428 expand_all_artificial_thunks if necessary.
16430 2018-04-17 Martin Liska <mliska@suse.cz>
16433 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
16434 in message, remote space in between '_G' and '('.
16436 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16439 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
16440 avx512f_vmcmp<mode>3<round_saeonly_name>,
16441 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
16442 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
16443 avx512f_rndscale<mode><round_saeonly_name>,
16444 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
16445 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
16446 Use %<iptr>2 instead of %2 for -masm=intel.
16447 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
16448 avx512f_vcvttss2usi<round_saeonly_name>,
16449 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
16451 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
16452 avx512f_vcvttsd2usi<round_saeonly_name>,
16453 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
16454 Use %q1 instead of %1 for -masm=intel.
16455 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
16456 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
16457 of %3 for -masm=intel.
16458 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
16460 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
16462 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
16464 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
16465 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
16467 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
16469 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
16470 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
16471 %g1 and one with %0 and %1.
16472 (avx512er_vmrcp28<mode><round_saeonly_name>,
16473 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
16474 %1 for -masm=intel.
16475 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
16476 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
16477 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
16478 of %0 and %{%4%} for -masm=intel.
16479 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
16480 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
16481 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
16482 order of %0 and %{%5%}%{z%} for -masm=intel.
16484 2018-04-17 Jan Hubicka <jh@suse.cz>
16487 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
16489 2018-04-17 Martin Liska <mliska@suse.cz>
16492 * multiple_target.c (create_dispatcher_calls): Set apostrophes
16493 for target_clone error message. Make default implementation
16494 clone to be a local declaration.
16495 (separate_attrs): Add new argument and check for an empty
16497 (expand_target_clones): Handle it.
16498 (ipa_target_clone): Make redirection just for target_clones
16501 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
16502 Tom de Vries <tom@codesourcery.com>
16504 PR middle-end/84955
16505 * omp-expand.c (expand_oacc_for): Add dummy false branch for
16506 tiled basic blocks without omp continue statements.
16508 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
16511 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
16512 vec_extract expression as having side effects to make sure it gets
16515 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
16518 * config/i386/i386.c (get_builtin_code_for_version): Check
16521 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
16524 * config.gcc: Support "skylake".
16525 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16527 * config/i386/i386.c (m_SKYLAKE): Define.
16528 (processor_target_table): Add "skylake".
16529 (ix86_option_override_internal): Add "skylake".
16530 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
16531 PROCESSOR_CANNONLAKE.
16532 (get_builtin_code_for_version): Fix priority for
16533 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
16534 PROCESSOR_SKYLAKE-AVX512.
16535 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
16536 (processor_type): Add PROCESSOR_SKYLAKE.
16538 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
16539 Jason Merrill <jason@redhat.com>
16542 * convert.c (convert_to_integer_1): Use direct recursion for
16543 enumeral types and types with a precision less than the number
16544 of bits in their mode.
16546 2018-04-16 Julia Koval <julia.koval@intel.com>
16549 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
16550 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
16552 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
16555 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
16556 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
16557 and -mno-direct-move.
16559 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
16562 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
16563 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
16564 Ensure negative shifts result in {0}.
16566 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
16568 PR rtl-optimization/79916
16569 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
16570 regs (if any) to define how to gnerate SD moves when LRA is in
16573 2018-04-13 Jakub Jelinek <jakub@redhat.com>
16575 PR rtl-optimization/85393
16576 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
16577 * except.c (expand_dw2_landing_pad_for_region): Make static.
16578 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
16579 a label and unconditional jump to old_bb, rather than
16580 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
16583 PR rtl-optimization/85376
16584 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
16585 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
16586 instead of a specific value.
16588 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
16589 Bin Cheng <bin.cheng@arm.com>
16591 PR tree-optimization/82965
16592 PR tree-optimization/83991
16593 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
16594 by_profile_only parameter.
16595 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
16596 information if the loop was predicted to iterate too many times.
16597 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
16599 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
16602 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
16605 2018-04-13 Martin Liska <mliska@suse.cz>
16606 Jakub Jelinek <jakub@redhat.com>
16608 PR middle-end/81657
16609 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
16610 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
16611 * builtins.c (expand_builtin_memory_copy_args): Use
16612 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
16613 handle dest_addr == pc_rtx.
16615 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
16618 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
16619 asked to not generate direct moves.
16620 (fix_trunc<mode>si2_stfiwx): Similar.
16621 (fix_trunc<mode>si2_internal): Similar.
16623 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16626 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
16627 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
16628 lookup if dest in some wider mode is known to be const0_rtx and
16629 if so, record permanent equivalence for it to be ZERO_EXTEND of
16630 the narrower mode destination.
16632 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
16634 * lto-streamer-out.c (output_function): Revert 259346.
16635 * omp-expand.c (expand_oacc_for): Likewise.
16637 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
16639 PR rtl-optimization/85354
16640 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
16641 * sel-sched.c (sel_global_init): ... here.
16643 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
16646 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
16647 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
16648 mode for PE-COFF targets.
16649 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
16650 (i386_pe_asm_lto_end): Likewise.
16651 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
16652 (TARGET_ASM_LTO_END): Likewise.
16653 * config/i386/winnt.c (saved_debug_info_level): New static variable.
16654 (i386_pe_asm_lto_start): New function.
16655 (i386_pe_asm_lto_end): Likewise.
16657 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
16658 Richard Biener <rguenther@suse.de>
16660 PR middle-end/84955
16661 * lto-streamer-out.c (output_function): Fix CFG loop state before
16663 * omp-expand.c (expand_oacc_for): Handle calls to internal
16664 functions like regular functions.
16666 2018-04-12 Richard Biener <rguenther@suse.de>
16669 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
16670 for the early LTO debug to properly generate references to it
16671 during DIE emission. Do not re-use that for the skeleton for
16673 (dwarf2out_early_finish): Likewise.
16675 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16678 * config/i386/sse.md
16679 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
16680 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
16681 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
16682 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
16683 and output is a reg, avoid creating invalid lowpart subreg, but
16684 instead split into a 512-bit move. Don't split if not AVX512VL,
16685 input is xmm16+ reg and output is a mem.
16686 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
16687 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
16688 xmm16+ reg and output is a mem.
16690 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16692 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
16693 also for flag_dwarf2_cfi_asm.
16695 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16697 PR rtl-optimization/85342
16698 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
16699 a bool scalar var inside of the loop instead. Don't try to update
16700 recog_data.operand after failed apply_change_group.
16702 2018-04-12 Tom de Vries <tom@codesourcery.com>
16705 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
16706 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
16707 array with flexible array member as array without given dimension.
16708 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
16709 argument for undefined param to true.
16711 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
16714 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16715 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
16716 from PowerPC section.
16717 * config/rs6000/sysv4.opt (mcall-): Improve help text.
16718 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
16719 help text that is too long.
16720 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
16721 help text that is too long.
16722 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
16723 help text that is too long.
16725 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
16727 * config/alpha/alpha.md (stack_probe_internal): Rename
16728 from "probe_stack". Update all callers.
16730 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16732 PR rtl-optimization/84566
16733 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
16734 sched_macro_fuse_insns.
16736 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16739 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
16740 (compute_block_dependences): ... from here.
16742 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16744 PR tree-optimization/85331
16745 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
16746 from int to HOST_WIDE_INT.
16748 2018-04-11 Martin Jambor <mjambor@suse.cz>
16751 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
16752 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
16753 not the same as the source val.
16754 (cgraph_edge_brings_value_p): New parameter.
16755 (gather_edges_for_value): Pass destination value to
16756 cgraph_edge_brings_value_p.
16757 (perhaps_add_new_callers): Likewise.
16758 (get_info_about_necessary_edges): Likewise and exclude values brought
16759 only by self-recursive edges.
16760 (create_specialized_node): Redirect only clones of self-calling edges.
16761 (+self_recursive_pass_through_p): New function.
16762 (find_more_scalar_values_for_callers_subset): Use it.
16763 (find_aggregate_values_for_callers_subset): Likewise.
16764 (known_aggs_to_agg_replacement_list): Removed.
16765 (decide_whether_version_node): Re-calculate known constants for all
16766 remaining context clones.
16768 2018-04-11 Richard Biener <rguenther@suse.de>
16771 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
16772 from early DWARF output.
16773 (dwarf2out_early_finish): Output line info unconditionally into
16774 early DWARF and add reference to it.
16776 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16779 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
16780 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
16781 other than V2DFmode using iptr mode attribute.
16782 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
16784 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16786 PR rtl-optimization/84659
16787 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
16789 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16792 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
16794 (output_loc_list): Pass address of a dummy size variable even in the
16795 locview handling loop.
16796 (index_location_lists): Add comment on why skip_loc_list_entry can't
16799 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
16802 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
16805 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
16808 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16809 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
16810 and -mstring-compare-inline-limit.
16812 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16815 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
16816 for stack clash protection in a register whenever we need it to be in
16819 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16821 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
16822 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
16824 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16827 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
16829 (mlong-double-): Ditto.
16830 * config/rs6000/sysv4.opt (msdata=): Ditto.
16831 (mtls-size=): Ditto.
16833 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16835 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16836 erroneous entries for
16837 "vector int vec_ldl (int, long int *)", and
16838 "vector unsigned int vec_ldl (int, unsigned long int *)".
16839 Add comments and entries for
16840 "vector bool char vec_ldl (int, bool char *)",
16841 "vector bool short vec_ldl (int, bool short *)",
16842 "vector bool int vec_ldl (int, bool int *)",
16843 "vector bool long long vec_ldl (int, bool long long *)",
16844 "vector pixel vec_ldl (int, pixel *)",
16845 "vector long long vec_ldl (int, long long *)",
16846 "vector unsigned long long vec_ldl (int, unsigned long long *)".
16847 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
16848 type tree bool_long_long_type_node and correct definition of
16849 bool_V2DI_type_node to make reference to this new type tree.
16850 (rs6000_mangle_type): Replace erroneous reference to
16851 bool_long_type_node with bool_long_long_type_node.
16852 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
16853 comments to emphasize sign distinctions for char and int types and
16854 replace RS6000_BTI_bool_long constant with
16855 RS6000_BTI_bool_long_long constant. Also add comment to restrict
16856 use of RS6000_BTI_pixel.
16857 (bool_long_type_node): Remove this macro definition.
16858 (bool_long_long_type_node): New macro definition
16860 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16862 PR rtl-optimization/85300
16863 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
16864 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
16865 simplify_unary_operation fails.
16867 2018-04-10 Martin Liska <mliska@suse.cz>
16869 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
16870 cgraph_edge and ipa_ref.
16872 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16876 * config/i386/sse.md
16877 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
16878 computation of the VEC_MERGE selector from mask.
16879 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
16880 Fix decoding of the VEC_MERGE selector into mask.
16882 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16884 PR tree-optimization/85286
16885 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
16887 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16889 * final.c (final_1): Set insn_last_address as well as
16890 insn_current_address.
16892 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16895 * explow.c (emit_stack_probe): Call validize_mem on memory location
16896 before passing it to gen_probe_stack. Create address operand and
16897 legitimize it for the probe_stack_address case.
16899 2018-04-09 Jan Hubicka <jh@suse.cz>
16902 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
16903 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
16904 * tree.c (free_lang_data_in_type): Fix handling of binfos;
16906 (free_lang_data): Rebuild type inheritance graph.
16908 2018-04-09 Martin Sebor <msebor@redhat.com>
16910 * invoke.texi (-finline-small-functions): Mention other optimization
16912 (-findirect-inlining, -fpartial-inlining): Same.
16913 (-finline-functions-called-once): Same.
16914 (-freorder-blocks-and-partition): Same.
16916 2018-04-09 Jan Hubicka <jh@suse.cz>
16919 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
16920 jumps; choose last target that matches the criteria (i.e.
16921 no partition changes for non-crossing jumps).
16922 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
16923 support for redirecting crossing jumps to non-crossing.
16925 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
16927 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
16928 also for naked functions.
16930 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
16932 * config/arc/arc.md (add_shift): New pattern.
16933 (add_shift2): Likewise.
16934 (sub_shift): Likewise.
16935 (sub_shift_cmp0_noout): Likewise.
16936 (compare_si_ashiftsi): Likewise.
16937 (xbfu_cmp0_noout): New combine pattern.
16938 (xbfu_cmp0"): Likewise.
16939 (movsi_set_cc_insn): Place the predicable variant first.
16940 (commutative_binary_cmp0_noout): Remove clobber.
16941 (commutative_binary_cmp0): New pattern.
16942 (noncommutative_binary_cmp0): Likewise.
16943 (noncommutative_binary_cmp0_noout): Likewise.
16944 (noncommutative_binary_comparison_result_used): Removed.
16945 (rsub_cmp0): New pattern.
16946 (rsub_cmp0_noout): Likewise.
16947 (extzvsi): Changed, keep only meaningful variants.
16948 (SQH, SEZ): New iterators.
16949 (SQH_postfix): New mode attribute.
16950 (SEZ_prefix): New code attribute.
16951 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
16952 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
16953 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
16954 of numerical value.
16955 (noncommutative_operator): Check the availability of barrel
16958 2018-04-09 Richard Biener <rguenther@suse.de>
16960 PR tree-optimization/85284
16961 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
16962 Only use the niter constraining form of simple_iv when the exit
16963 is always executed.
16965 2018-04-09 Tom de Vries <tom@codesourcery.com>
16968 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16969 (define_expand "*memory_barrier"): New define_expand.
16970 (define_insn "memory_barrier"): New insn.
16972 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16974 PR rtl-optimization/80463
16975 PR rtl-optimization/83972
16976 PR rtl-optimization/83480
16978 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
16979 correct producer for the insn.
16980 (tidy_control_flow): Fixup seqnos in case of debug insns.
16982 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16984 PR rtl-optimization/83913
16986 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
16987 different sched-times when merging exprs.
16989 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16991 PR rtl-optimization/83962
16993 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
16994 tidy_fallthru_edge and tidy_control_flow.
16996 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16998 PR rtl-optimization/83530
17000 * sel-sched.c (force_next_insn): New global variable.
17001 (remove_insn_for_debug): When force_next_insn is true, also leave only
17002 next insn in the ready list.
17003 (sel_sched_region): When the region wasn't scheduled, make another pass
17004 over it with force_next_insn set to 1.
17006 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17008 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17010 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17011 for interrupt control.
17012 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17013 functions for interrupt control.
17014 * config/nds32/nds32-intrinsic.md: Likewise.
17015 * config/nds32/nds32_intrinsic.h: Likewise.
17016 * config/nds32/nds32.h (nds32_builtins): Likewise.
17018 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17020 * config/nds32/nds32.c (nds32_init_machine_status,
17021 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17022 strict_aligned_p field.
17023 (nds32_expand_to_rtl_hook): New function.
17024 (TARGET_EXPAND_TO_RTL_HOOK): Define.
17025 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17027 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17028 Chung-Ju Wu <jasonwucj@gmail.com>
17030 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17031 * config/nds32/nds32-n7.md: New file.
17032 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17033 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17035 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17036 * config/nds32/nds32.md (pipeline_model): Add n7.
17037 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17038 * config/nds32/pipelines.md: Include n7 settings.
17040 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17041 Chung-Ju Wu <jasonwucj@gmail.com>
17043 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17044 * config/nds32/nds32-e8.md: New file.
17045 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17046 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17048 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17049 * config/nds32/nds32.md (pipeline_model): Add e8.
17050 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17051 * config/nds32/pipelines.md: Include e8 settings.
17053 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17054 Chung-Ju Wu <jasonwucj@gmail.com>
17056 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17057 * config/nds32/nds32-n8.md: New file.
17058 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17059 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17061 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17062 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17063 * config/nds32/nds32.md (pipeline_model): Add n8.
17064 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17065 * config/nds32/pipelines.md: Include n8 settings.
17067 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17068 Chung-Ju Wu <jasonwucj@gmail.com>
17070 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17071 * config/nds32/nds32-n9-2r1w.md: New file.
17072 * config/nds32/nds32-n9-3r2w.md: New file.
17073 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17074 nds32_register_ports): New or modify for cpu n9.
17075 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17077 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17078 * config/nds32/nds32-utils.c: New file.
17079 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17080 TARGET_MUL_SLOW): Define.
17081 * config/nds32/nds32.md (pipeline_model): New attribute.
17082 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17083 New options that support cpu n9.
17084 * config/nds32/pipelines.md: Include n9 settings.
17085 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17087 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17089 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17090 information if necessary.
17091 (output_cond_branch_compare_zero): Likewise.
17092 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17093 (nds32_target_alignment): Refine for alignment.
17094 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17095 (FUNCTION_BOUNDARY): Modify.
17096 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17098 * config/nds32/nds32.opt (malways-align, malign-functions): New.
17100 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17102 * config/nds32/constants.md (unspec_volatile_element): Add values for
17103 TLB operation and data prefetch.
17104 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17105 functions for TLB operation and data prefetch.
17106 * config/nds32/nds32-intrinsic.md: Likewise.
17107 * config/nds32/nds32_intrinsic.h: Likewise.
17108 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17109 (nds32_print_operand): Likewise.
17110 * config/nds32/nds32.h (nds32_builtins): Likewise.
17112 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
17113 Andrew Pinski <pinsika@gcc.gnu.org>
17115 PR middle-end/82976
17116 * match.pd: Use constant_boolean_node of correct type instead of
17117 boolean_true_node or boolean_false_node for simplifying
17118 pointer comparisons to zero.
17120 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17122 PR tree-optimization/80021
17123 * tree.c (verify_type_variant): Make error call in verify_variant_match
17124 translatable and remove final full stop.
17126 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17128 * config/nds32/constants.md (unspec_volatile_element): Add
17129 UNSPEC_VOLATILE_EH_RETURN.
17130 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17131 nds32_output_stack_pop): Support dwarf exception handling process.
17132 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17133 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17134 exception handling process.
17135 (nds32_compute_stack_frame): Likewise.
17136 (nds32_return_addr_rtx): Likewise.
17137 (nds32_initial_elimination_offset): Likewise.
17138 (nds32_expand_prologue): Likewise.
17139 (nds32_expand_epilogue): Likewise.
17140 (nds32_dynamic_chain_address): New function.
17141 * config/nds32/nds32.h (machine_function): Add fields for dwarf
17142 exception handling.
17143 (DYNAMIC_CHAIN_ADDRESS): Define.
17144 (EH_RETURN_DATA_REGNO): Define.
17145 (EH_RETURN_STACKADJ_RTX): Define.
17146 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17147 patterns for dwarf exception handling.
17149 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17151 * config/nds32/nds32.h: Clean up obsolete macros.
17153 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17155 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17156 Add enum values for particular instructions.
17157 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17158 particular intrinsic functions.
17159 * config/nds32/nds32-intrinsic.md: Likewise.
17160 * config/nds32/nds32_intrinsic.h: Likewise.
17161 * config/nds32/nds32.h (nds32_builtins): Likewise.
17162 * config/nds32/nds32.md (type): Add pbsad and pbsada.
17163 (btst, ave): New patterns for particular instructions.
17165 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17167 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17168 Add enum values for atomic load/store and memory sync.
17169 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17171 * config/nds32/nds32-intrinsic.md: Likewise.
17172 * config/nds32/nds32_intrinsic.h: Likewise.
17173 * config/nds32/nds32.h (nds32_builtins): Likewise.
17175 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17177 PR tree-optimization/85257
17178 * fold-const.c (native_encode_vector): If not all elts could fit
17179 and off is -1, return 0 rather than offset.
17180 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17181 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17182 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
17183 adjust buffer in native_interpret_expr call.
17185 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17187 * config/nds32/constants.md (unspec_volatile_element): Add cache
17188 control enum values.
17189 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17190 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17191 * config/nds32/nds32.c (nds32_cctl_names): New.
17192 (nds32_print_operand): Handle cache control register names.
17193 * config/nds32/nds32.h (nds32_builtins): New enum values.
17194 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17196 * config/nds32/nds32.md (type): Add mmu.
17197 * config/nds32/pipelines.md (simple_insn): Add mmu.
17199 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17201 * config/nds32/nds32.md (type): Remove call.
17202 * config/nds32/pipelines.md (simple_insn): Likewise.
17204 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17206 * config/nds32/constants.md (unspec_volatile_element): Add
17207 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17208 UNSPEC_VOLATILE_FMFCFG.
17209 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17210 description for fmfcfg and fmfcsr.
17211 (bdesc_1arg): Add fmtcsr.
17212 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17213 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17214 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17215 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17216 unspec_fmfcfg): New patterns.
17217 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17218 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17219 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17220 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17221 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17222 __nds32__fmfcfg): Define.
17224 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17226 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17227 intrinsic register names.
17228 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17229 intrinsic register enum values and macros.
17231 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17233 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17234 for load/store addressing form.
17235 (nds32_print_operand_address): Likewise.
17237 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
17240 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17241 based on LABEL_REF. Remove useless assertion.
17242 (pic_address_needs_scratch): Fix formatting.
17243 (sparc_legitimize_pic_address): Minor tweaks.
17244 (sparc_delegitimize_address): Adjust assertion accordingly.
17245 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17246 into symbolic_operand.
17247 (movsi_high_pic_label_ref): Likewise.
17248 (movsi_lo_sum_pic_label_ref): Likewise.
17249 (movdi_pic_label_ref): Likewise.
17250 (movdi_high_pic_label_ref): Likewise.
17251 (movdi_lo_sum_pic_label_ref): Likewise.
17253 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
17255 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17256 custom LIB_SPEC setup.
17258 2018-04-06 Ruslan Bukin <br@bsdpad.com>
17259 Kito Cheng <kito.cheng@gmail.com>
17261 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17262 * config/riscv/freebsd.h: New.
17264 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17266 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17267 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17270 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17271 Kito Cheng <kito.cheng@gmail.com>
17273 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17274 nds32_output_call, nds32_symbol_binds_local_p): New functions.
17275 * config/nds32/nds32-protos.h (nds32_output_call,
17276 nds32_output_return): Declare.
17277 * config/nds32/nds32.md: Refine all the call and return patterns.
17279 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17282 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17283 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17285 PR rtl-optimization/84872
17286 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17287 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17288 EDGE_CROSSING edge.
17290 2018-04-06 Tamar Christina <tamar.christina@arm.com>
17292 * expr.c (copy_blkmode_to_reg): Revert 254862.
17293 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17295 2018-04-06 Richard Biener <rguenther@suse.de>
17297 PR middle-end/85244
17298 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17299 after seeing a component reference with an adjacent field. Treat
17300 refs to arrays at struct end of external decls similar to
17301 refs to unconstrained commons.
17303 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17306 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17307 look through SAVE_EXPRs with non-side-effects argument. Adjust
17309 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17312 2018-04-06 Richard Biener <rguenther@suse.de>
17314 PR middle-end/85180
17315 * alias.c (find_base_term): New wrapper around find_base_term
17316 unwinding CSELIB_VAL_PTR changes.
17317 (find_base_term): Do not restore CSELIB_VAL_PTR during the
17320 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17322 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17324 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17325 constant definitions.
17326 ("nop"): lr 0,0 -> nopr r0
17327 ("nop_lr0", "nop_lr1"): New insn definitions.
17329 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17331 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17332 NDS32_V3PUSH_AVAILABLE_P macro.
17334 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
17335 Chung-Ju Wu <jasonwucj@gmail.com>
17337 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
17338 (nds32*-*-*): Add float and fpu_config into supported_defaults.
17339 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
17340 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
17341 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
17342 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
17343 * config/nds32/constraints.md: New constraints and checking for hard
17344 float configuration.
17345 * config/nds32/iterators.md: New mode iterator and attribute for hard
17346 float configuration.
17347 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
17349 * config/nds32/nds32-fpu.md: New file.
17350 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
17351 deal with hard float code generation.
17352 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
17354 (abi_type, float_reg_number): New enum type.
17355 * config/nds32/nds32-predicates.c: New predicates for hard float.
17356 * config/nds32/nds32-protos.h: Declare functions for hard float.
17357 * config/nds32/nds32.c: Implementation for hard float configuration.
17358 * config/nds32/nds32.h: Definitions for hard float configuration.
17359 * config/nds32/nds32.md: Include hard float machine description and
17360 modify patterns for hard float configuration.
17361 * config/nds32/nds32.opt: New options for hard float configuration.
17362 * config/nds32/predicates.md: New predicates for hard float
17365 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17367 * common/config/nds32/nds32-common.c
17368 (nds32_option_optimization_table): Enable -mreleax-hint by default.
17370 2018-04-05 Jakub Jelinek <jakub@redhat.com>
17372 PR middle-end/85195
17373 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
17374 CONSTRUCTOR_ELT (ctor, ...)->value.
17376 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
17379 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
17381 2018-04-05 Tom de Vries <tom@codesourcery.com>
17384 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
17387 2018-04-05 Shiva Chen <shiva0217@gmail.com>
17388 Kito Cheng <kito.cheng@gmail.com>
17390 * config/nds32/constraints.md (U33): Fine-tune checking condition.
17391 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
17392 * config/nds32/nds32.h (nds32_16bit_address_type): Add
17393 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
17395 2018-04-05 Shiva Chen <shiva0217@gmail.com>
17396 Kito Cheng <kito.cheng@gmail.com>
17398 * config/nds32/constraints.md (Ufe): New memory constraint.
17399 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
17400 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
17401 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
17403 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
17404 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
17406 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17408 * config/nds32/nds32.md: Use optimize_size in the condition for
17409 alu-shift instructions.
17411 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17413 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
17415 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17417 * config/nds32/nds32.md (negsi2): Refine pattern.
17419 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
17420 Chung-Ju Wu <jasonwucj@gmail.com>
17422 * config/nds32/iterators.md (shift_rotate): New code iterator.
17423 (shift): New code attribute.
17424 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
17425 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
17426 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
17427 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
17428 bit-wise operations.
17429 (andsi3, *andsi3): Ditto.
17430 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
17431 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
17432 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
17433 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
17434 nds32_ior_operand, nds32_xor_operand): New predicates.
17436 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17438 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
17439 (addsi3, subsi3): ... this.
17441 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17443 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
17445 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17447 * config/nds32/nds32.md: Adjust indention.
17449 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
17451 * config/nds32/nds32.md (feature): New attribute.
17453 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17455 * config/nds32/nds32.md (subtype): New attribute.
17457 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
17460 * config/arm/arm-builtins.c (arm_expand_builtin): Change
17461 expansion to perform a bitwise AND of the argument followed by a
17462 boolean negation of the result.
17464 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
17466 PR rtl-optimization/84878
17467 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
17468 the basic block. Assert the use reference is not artificial and that
17469 it has an associated insn.
17471 2018-04-04 Michael Matz <matz@suse.de>
17473 * builtins.c (compute_objsize): Pass correct operand
17474 to array_at_struct_end_p.
17476 2018-04-04 Richard Biener <rguenther@suse.de>
17479 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
17480 from contexts for DINFO_LEVEL_TERSE and below.
17482 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17484 * config/nds32/nds32-doubleword.md (move_<mode>): Require
17485 resiter_operand condition.
17486 * config/nds32/nds32.md (*move<mode>): Ditto.
17488 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17489 Monk Chiang <sh.chiang04@gmail.com>
17491 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
17493 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17495 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
17497 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17498 Kito Cheng <kito.cheng@gmail.com>
17500 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
17501 nds32_cond_code_str, output_cond_branch,
17502 output_cond_branch_compare_zero, nds32_expand_cbranch,
17503 nds32_expand_cstore, nds32_expand_movcc,
17504 nds32_output_cbranchsi4_equality_zero,
17505 nds32_output_cbranchsi4_equality_reg,
17506 nds32_output_cbranchsi4_equality_reg_or_const_int,
17507 nds32_output_cbranchsi4_greater_less_zero: New functions.
17508 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
17509 nds32_expand_cstore, nds32_expand_movcc,
17510 nds32_output_cbranchsi4_equality_zero,
17511 nds32_output_cbranchsi4_equality_reg,
17512 nds32_output_cbranchsi4_equality_reg_or_const_int,
17513 nds32_output_cbranchsi4_greater_less_zero): Declare.
17514 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
17515 nds32_rimm11s_operand): New predicates.
17516 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
17517 * config/nds32/nds32.md: Rewrite all the branch and conditional move
17520 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17522 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
17523 * config/nds32/nds32.md: Ditto.
17524 * config/nds32/pipelines.md: Ditto.
17526 2018-04-04 Richard Biener <rguenther@suse.de>
17528 PR tree-optimization/85168
17529 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
17530 propagating abnormals.
17532 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17534 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
17536 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17537 Kito Cheng <kito.cheng@gmail.com>
17539 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
17540 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
17541 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
17542 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
17543 * config/nds32/nds32.md (sibcall_internal): New.
17544 (sibcall_register): Remove.
17545 (sibcall_immediate): Remove.
17546 (sibcall_value_internal): New.
17547 (sibcall_value_register): Remove.
17548 (sibcall_value_immediate): Remove.
17549 * config/nds32/predicates.md (nds32_general_register_operand): New.
17550 (nds32_call_address_operand): New.
17552 2018-04-03 Jakub Jelinek <jakub@redhat.com>
17554 PR rtl-optimization/85167
17555 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
17556 bb_defs if *split_p, instead preinitialize it to NULL.
17558 PR tree-optimization/85156
17559 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
17560 evaluating the argument multiple times.
17562 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
17564 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
17566 (_mm_cvtpd_ps): Likewise.
17567 (_mm_cvttpd_epi32): Likewise.
17568 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
17569 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
17570 vector, pixel, and bool following altivec.h include.
17572 2018-04-03 Martin Sebor <msebor@redhat.com>
17574 * doc/extend.texi (Common Function Attributes): Clarify.
17575 (const attribute): Likewise.
17576 (pure attribute): Likewise.
17578 2018-04-03 Jakub Jelinek <jakub@redhat.com>
17581 * config/i386/i386.c (ix86_expand_vector_set): Use
17582 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
17584 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
17586 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
17587 instructions when changing rounding bits to preserve precision bits
17588 in the x87 control word.
17590 2018-04-03 Martin Liska <mliska@suse.cz>
17592 PR tree-optimization/82491
17593 * rtl.h (strip_offset_and_add): Replace += suboffset with
17594 poly_uint64 () + suboffset.
17596 2018-03-29 Martin Liska <mliska@suse.cz>
17597 Martin Jambor <mjambor@suse.cz>
17600 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
17601 param_type is not an integral or pointer type.
17603 2018-04-03 Richard Biener <rguenther@suse.de>
17605 * sese.h (recompute_all_dominators): Remove.
17607 2018-04-02 Martin Sebor <msebor@redhat.com>
17609 * doc/invoke.texi (-Wrestrict): Fix typos.
17611 2018-04-02 Jim Wilson <jimw@sifive.com>
17613 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
17614 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
17615 (<optab>di3, <optab>si3_extend): Likewise.
17616 (<optab>si3_mask, <optab>si3_mask_1): New.
17617 (<optab>di3_mask, <optab>di3_mask_1): New.
17618 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
17619 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
17620 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
17622 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
17624 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
17627 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
17629 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
17630 (nds32_canonicalize_comparison): New function.
17632 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17633 Kito Cheng <kito.cheng@gmail.com>
17634 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17636 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
17637 * config/nds32/constants.md (unspec_volatile_element): Add
17638 UNSPEC_VOLATILE_RELAX_GROUP.
17639 * config/nds32/nds32-relax-opt.c: New file.
17640 * config/nds32/nds32-predicates.c
17641 (nds32_symbol_load_store_p): New function.
17642 * config/nds32/nds32-protos.h
17643 (nds32_symbol_load_store_p): Declare function.
17644 (make_pass_nds32_relax_opt): Declare new rtl pass function.
17645 * config/nds32/nds32.c
17646 (nds32_register_pass): New function to register pass.
17647 (nds32_register_passes): New function to register passes.
17648 * config/nds32/nds32.md (relax_group): New pattern.
17649 * config/nds32/nds32.opt (mrelax-hint): New option.
17650 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
17652 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
17654 * config/nds32/t-nds32: Modify files dependency.
17656 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17658 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
17659 (PROFILE_HOOK): Define its implementation.
17661 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17663 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
17664 type and 32-bit size.
17666 2018-04-01 Jakub Jelinek <jakub@redhat.com>
17668 PR middle-end/85090
17669 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
17670 (V_128_256): New mode iterator.
17671 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
17672 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
17673 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
17675 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
17676 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
17678 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
17681 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
17682 NaN inputs correctly.
17684 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
17687 * config/rs6000/vsx.md (??r): New mode attribute.
17688 (*vsx_mov<mode>_64bit): Use it.
17689 (*vsx_mov<mode>_32bit): Likewise.
17691 2018-03-30 Martin Sebor <msebor@redhat.com>
17693 PR tree-optimization/84818
17694 * builtins.c (check_access): Use warning_n.
17696 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17699 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
17701 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
17704 2018-03-30 Julia Koval <julia.koval@intel.com>
17707 * x86-tune.def (movx, partial_reg_dependency): Enable for
17710 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
17712 PR inline-asm/84985
17713 * lra-constraints.c (process_alt_operands): Move setting
17714 this_alternative_matches below.
17716 2018-03-29 Martin Liska <mliska@suse.cz>
17719 * doc/invoke.texi: Document how LTO works with debug info.
17720 Describe auto-load support of binutils. Mention 'x86-64'
17721 as valid option value of -march option.
17723 2018-03-29 Jakub Jelinek <jakub@redhat.com>
17725 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
17728 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
17729 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
17730 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
17733 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17736 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
17737 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
17738 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
17739 for RS6000_BTM_POWERPC64.
17740 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
17741 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
17742 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
17745 (DIVDEU): Likewise.
17747 2018-03-28 Carl Love <cel@us.ibm.com>
17750 2017-09-27 Carl Love <cel@us.ibm.com>
17752 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
17753 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
17754 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
17757 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17759 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
17760 instead of __vector bool.
17761 (_mm_max_pu8): Likewise.
17762 (_mm_min_pi16): Likewise.
17764 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17767 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
17768 (DIVWEUO): Likewise.
17769 (DIVDEO): Likewise.
17770 (DIVDEUO): Likewise.
17771 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
17772 DIVWEUO and DIVDEUO.
17773 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
17774 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
17775 (div_extend): Likewise.
17776 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
17778 (__builtin_divweuo): Likewise.
17779 (__builtin_divdeo): Likewise.
17780 (__builtin_divdeuo): Likewise.
17782 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17785 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
17786 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
17788 PR tree-optimization/82004
17789 * gimple-match-head.c (optimize_pow_to_exp): New function.
17790 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
17791 Don't fold to exp if optimize_pow_to_exp is false.
17793 2018-03-28 Martin Liska <mliska@suse.cz>
17796 * calls.c (initialize_argument_information): Fix trailing space.
17797 * common.opt: Fix typo and provide better explanation for
17798 -fsanitize-coverage option.
17799 * config/i386/i386.opt: Fix typo.
17801 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17802 Martin Liska <mliska@suse.cz>
17805 * gimplify.c (asan_poison_variable): Don't do the check for
17806 gimplify_omp_ctxp here.
17807 (gimplify_decl_expr): Do it here.
17808 (gimplify_target_expr): Likewise.
17810 2018-03-28 Martin Liska <mliska@suse.cz>
17813 * config/i386/i386.c (ix86_function_arg_advance): Do not call
17814 chkp_type_bounds_count if MPX is not enabled.
17816 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
17818 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
17820 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
17823 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
17824 function to create the function decl for complex long double
17825 multiply and divide for -mabi=ieeelongdouble.
17826 (init_float128_ieee): Call it.
17828 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
17831 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
17832 -fcf-protection=branch -mibt.
17833 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
17835 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17838 * config/arm/arm.c (arm_valid_symbolic_address): Handle
17839 arm_word_relocations.
17841 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
17844 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
17845 extern array declarations.
17847 2018-03-27 Richard Biener <rguenther@suse.de>
17849 PR middle-end/84067
17850 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
17851 explicit single_use checks.
17853 2018-03-27 Richard Biener <rguenther@suse.de>
17855 PR tree-optimization/85082
17856 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
17859 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17861 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
17862 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
17863 Turn on fasynchronous-unwind-tables and funwind-tables.
17865 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
17868 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
17869 (*bmi_blsr_<mode>_ccz): Ditto.
17871 2018-03-26 Tom de Vries <tom@codesourcery.com>
17873 PR tree-optimization/85063
17874 * omp-general.c (offloading_function_p): New function. Factor out
17876 * omp-offload.c (pass_omp_target_link::gate): ... here.
17877 * omp-general.h (offloading_function_p): Declare.
17878 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
17879 with attribute omp declare target for offloading functions.
17881 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
17883 PR tree-optimization/84005
17884 * tree-data-ref.h (get_base_for_alignment): Declare.
17885 * tree-data-ref.c (get_base_for_alignment_1): New function.
17886 (get_base_for_alignment): Likewise.
17887 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
17888 get_base_for_alignment to find a suitable base object, instead
17889 of always using drb->base_address.
17891 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17893 PR inline-asm/85022
17894 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
17895 known size by default.
17897 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
17899 PR inline-asm/85030
17900 * lra-constraints.c (process_alt_operands): Don't match BLKmode
17901 and non BLKmode operands.
17903 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17906 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
17907 Clean up attributes.
17909 2018-03-23 Richard Biener <rguenther@suse.de>
17912 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
17913 we are going to emit early debug for LTO.
17915 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17917 PR inline-asm/85034
17918 * function.c (match_asm_constraints_1): Don't optimize if input
17919 doesn't satisfy general_operand predicate for output's mode.
17921 PR inline-asm/85022
17922 * alias.c (write_dependence_p): Don't require for x_canonicalized
17923 non-VOIDmode if x has VOIDmode.
17926 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
17927 just don't try to optimize it rather than assert it never happens.
17929 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17931 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
17932 macro expansions for definition of ST_INTERNAL_<mode> and
17933 LD_INTERNAL_<mode> builtins.
17934 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
17936 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
17938 (altivec_expand_st_builtin): Likewise.
17939 (altivec_expand_builtin): Remove calls to deleted functions.
17940 (rs6000_address_for_altivec): Delete this function.
17941 * config/rs6000/vector.md: Remove expands for
17942 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
17944 2018-03-22 Sudakshina Das <sudi.das@arm.com>
17947 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
17948 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
17949 re-computing once computed.
17950 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
17951 (arm_init_machine_status): Initialize
17952 machine->static_chain_stack_bytes.
17954 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17957 * doc/extend.texi: Add four new prototypes for vec_ld.
17958 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
17959 definitions for more logical presentation.
17960 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
17961 entries for V1TI variants of __builtin_altivec_ld builtin.
17962 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
17963 handling of V1TI variant of LVX icode pattern.
17964 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
17965 (rs6000_gimple_fold_builtin): Likewise.
17966 (altivec_init_builtins): Add code to define
17967 __builtin_altivec_lvx_v1ti function.
17969 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17971 PR inline-asm/84941
17972 * function.c (match_asm_constraints_1): Don't do the optimization
17973 if input isn't a REG, SUBREG, MEM or constant.
17975 2018-03-22 Tom de Vries <tom@codesourcery.com>
17977 PR tree-optimization/84956
17978 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
17979 bb_has_abnormal_pred.
17981 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17984 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
17985 DECL_INITIAL (decl) to decl at the end.
17986 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
17987 adjust the comment.
17989 2018-03-21 Joseph Myers <joseph@codesourcery.com>
17991 * doc/extend.texi (__builtin_tgmath): Document when complex
17992 integer types are treated as _Complex _Float64.
17994 2018-03-21 Tom de Vries <tom@codesourcery.com>
17996 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
17998 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18000 PR tree-optimization/84960
18001 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18002 if it is ENTRY block, move them into single succ of ENTRY in that case.
18004 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
18006 PR tree-optimization/84811
18007 * poly-int.h (poly_span_traits): Remove the T3 parameter and
18008 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18009 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18010 (known_subrange_p): Update accordingly. Cast each value involved
18011 in the size comparison, rather than casting the result of the
18014 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18016 PR tree-optimization/84982
18017 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18018 by flipping the least significant bit rather than all bits from
18019 bitpos to bitpos + bitsize - 1.
18021 2018-03-21 Nathan Sidwell <nathan@acm.org>
18023 * doc/extend.texi (Deprecated Features): Remove mention of
18024 long-deleted deprecations.
18026 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18029 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18030 * configure: Regenerate.
18032 2018-03-21 Tom de Vries <tom@codesourcery.com>
18034 PR tree-optimization/83126
18035 * tree-parloops.c (num_phis): New function.
18036 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18038 2018-03-21 Nathan Sidwell <nathan@acm.org>
18040 * doc/extend.texi (Deprecated Features): Update deprecated flags,
18041 mention anon-struct/union members and trailing attributes.
18043 2018-03-21 Bin Cheng <bin.cheng@arm.com>
18045 PR tree-optimization/84969
18046 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18047 builtin memset partitions if they set different rhs values.
18049 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18051 PR rtl-optimization/84989
18052 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18053 VEC_DUPLICATE with scalar result mode.
18055 2018-03-21 Martin Liska <mliska@suse.cz>
18058 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18059 not intended return statement.
18061 2018-03-21 Martin Liska <mliska@suse.cz>
18064 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18065 (chkp_find_bound_slots_1): Limit number of iterations.
18067 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
18070 * Minor grammar fixes for x86 options.
18072 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18075 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18076 holding REG_CFA_RESTORE notes, instead turn them into a USE.
18078 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
18081 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18082 (altivec_lvx_<mode>_1op): Likewise.
18083 (altivec_stvx_<mode>_2op): Likewise.
18084 (altivec_stvx_<mode>_1op): Likewise.
18085 (altivec_lvx_<VM2:mode>): New define_expand.
18086 (altivec_stvx_<VM2:mode>): Likewise.
18087 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18088 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18089 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18090 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18091 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18092 (rs6000_gen_lvx): Likewise.
18093 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18094 (altivec_expand_stv_builtin): Likewise.
18095 (altivec_expand_builtin): Likewise.
18096 * config/rs6000/vector.md: Likewise.
18098 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18101 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18104 2018-03-20 Richard Biener <rguenther@suse.de>
18107 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18108 sign-conversions as zero, fall back to standard scalar_stmt
18111 2018-03-20 Martin Liska <mliska@suse.cz>
18114 * predict.c (rebuild_frequencies): Handle case when we have
18115 PROFILE_ABSENT, but flag_guess_branch_prob is false.
18117 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18120 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18121 flag_section_anchors.
18122 * varasm.c (use_blocks_for_decl_p): Remove hack for
18123 dw2_force_const_mem.
18126 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18128 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
18129 be created, use lowpart_subreg of operands[0] rather than operands[0]
18131 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18132 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18133 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18134 and n constraint instead of aarch64_shift_imm_di and Usd.
18135 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18136 (*aarch64_<optab>_reg_minus<mode>3): ... this.
18138 2018-03-20 Sudakshina Das <sudi.das@arm.com>
18141 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18142 to favor GPR over NEON registers.
18143 (<shift>di3_neon): Likewise.
18145 2018-03-20 Tom de Vries <tom@codesourcery.com>
18148 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18149 (nvptx_process_pars): Emit bar.sync asap and alap.
18151 2018-03-20 Tom de Vries <tom@codesourcery.com>
18154 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18155 seen_label if seen_label is already set.
18157 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18160 * config/i386/i386.c (fold_builtin_cpu): For features above 31
18161 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18162 Use 1U instead of 1. Formatting fixes.
18165 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18166 instead of TREE_TYPE (s1) for the return value.
18168 2018-03-19 Jakub Jelinek <jakub@redhat.com>
18170 PR tree-optimization/84946
18171 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18172 bitsize + bitsize in poly_uint64 rather than poly_int64.
18175 * dwarf2asm.c: Include fold-const.c.
18176 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18177 of decl rather than decl itself.
18179 PR rtl-optimization/84643
18180 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18182 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
18185 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18186 calling assemble_variable.
18188 2018-03-19 Sudakshina Das <sudi.das@arm.com>
18191 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18192 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18194 2018-03-19 Jim Wilson <jimw@sifive.com>
18197 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18198 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18199 (riscv_first_stack_step): Likewise.
18200 (riscv_option_override): Use STACK_BOUNDARY instead of
18201 MIN_STACK_BOUNDARY.
18202 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18203 MIN_STACK_BOUNDARY.
18204 (BIGGEST_ALIGNMENT): Set to 128.
18205 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18206 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18209 2018-03-19 Richard Biener <rguenther@suse.de>
18211 PR tree-optimization/84933
18212 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18213 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18215 2018-03-19 Richard Biener <rguenther@suse.de>
18217 PR tree-optimization/84859
18218 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18219 (cond_if_else_store_replacement): Perform sinking operation on
18220 single-store BBs regardless of MAX_STORES_TO_SINK setting.
18221 Generalize what a BB with a single eligible store is.
18223 2018-03-19 Richard Biener <rguenther@suse.de>
18225 PR tree-optimization/84929
18226 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18227 chrec_is_positive against non-chrec arg.
18229 2018-03-19 Tamar Christina <tamar.christina@arm.com>
18232 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18234 2018-03-18 Martin Liska <mliska@suse.cz>
18236 PR rtl-optimization/84635
18237 * regrename.c (build_def_use): Use matches_mode only when
18240 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
18242 PR tree-optimization/84913
18243 * tree-vect-loop.c (vectorizable_reduction): Don't try to
18244 vectorize chains of COND_EXPRs.
18246 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18248 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18250 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18252 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18254 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18256 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18258 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18259 Kito Cheng <kito.cheng@gmail.com>
18261 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18262 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18263 (nds32_adjust_reg_alloc_order): New function.
18264 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18266 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
18268 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18269 nds32_print_operand, nds32_print_operand_address): Use
18270 HOST_WIDE_INT_PRINT_DEC instead.
18272 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18274 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18276 2018-03-17 Jakub Jelinek <jakub@redhat.com>
18279 * config/i386/i386.c (initial_ix86_tune_features,
18280 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18281 unsigned long long.
18282 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18283 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18284 rather than 1u << ix86_tune. Formatting fix.
18285 (ix86_option_override_internal): Change ix86_arch_mask from
18286 unsigned int to unsigned HOST_WIDE_INT, initialize to
18287 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18288 (ix86_function_specific_restore): Likewise.
18290 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18293 * postreload.c (reload_combine_recognize_pattern): Perform
18294 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18295 truncate_int_for_mode the result for the destination's mode.
18298 * hsa-gen.c (mem_type_for_type): Fix comment typo.
18299 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18301 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18304 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
18307 * lra-assigns.c (lra_split_hard_reg_for): Don't use
18308 regno_allocno_class_array and sorted_pseudos.
18309 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18310 insns where regno is used.
18312 2018-03-16 Martin Liska <mliska@suse.cz>
18315 * multiple_target.c (create_dispatcher_calls): Redirect
18316 reference in the symbol table.
18318 2018-03-16 Martin Liska <mliska@suse.cz>
18321 * multiple_target.c (create_dispatcher_calls): Redirect also
18324 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18328 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18329 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18330 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18332 2018-03-16 Julia Koval <julia.koval@intel.com>
18334 * doc/invoke.texi (Skylake Server): Add CLWB.
18335 Cannonlake): Remove CLWB.
18337 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18339 PR tree-optimization/84841
18340 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
18342 (FLOAT_ONE_CONST_TYPE): Define.
18343 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
18344 (sort_by_operand_rank): Put entries with higher constant_type last
18345 rather than first to match comments.
18347 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
18349 * config/nios2/nios2.md (movsi_internal): Fix thinko in
18352 2018-03-15 Jakub Jelinek <jakub@redhat.com>
18355 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
18356 check and use address of target always.
18358 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
18361 * config/i386/i386.c (indirect_thunk_needed): Update comments.
18362 (indirect_thunk_bnd_needed): Likewise.
18363 (indirect_thunks_used): Likewise.
18364 (indirect_thunks_bnd_used): Likewise.
18365 (indirect_return_needed): New.
18366 (indirect_return_bnd_needed): Likewise.
18367 (output_indirect_thunk_function): Add a bool argument for
18369 (output_indirect_thunk_function): Don't generate alias for
18370 function return thunk.
18371 (ix86_code_end): Call output_indirect_thunk_function to generate
18372 function return thunks.
18373 (ix86_output_function_return): Set indirect_return_bnd_needed
18374 and indirect_return_needed instead of indirect_thunk_bnd_needed
18375 and indirect_thunk_needed.
18377 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
18379 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
18380 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
18381 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
18383 2018-03-15 David Malcolm <dmalcolm@redhat.com>
18384 Paul Hua <paul.hua.gm@gmail.com>
18387 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
18389 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
18391 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
18392 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
18393 resp. SFmode cases.
18395 2018-03-15 Tamar Christina <tamar.christina@arm.com>
18398 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
18399 instead of GET_MODE_SIZE when comparing Units.
18401 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18404 * varasm.c (hash_section): Return an unchangeble hash value
18405 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
18406 Return !aarch64_can_use_per_function_literal_pools_p ().
18408 2018-03-15 Jakub Jelinek <jakub@redhat.com>
18411 * optabs.c (emit_conditional_move): Pass address of cmode's copy
18412 rather than address of cmode as last argument to prepare_cmp_insn.
18414 2018-03-15 Julia Koval <julia.koval@intel.com>
18416 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
18417 F_AVX512VNNI, F_AVX512BITALG): New.
18419 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
18422 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
18423 insn for floating-point loads and stores.
18425 2018-03-14 Carl Love <cel@us.ibm.com>
18427 * config/rs6000/rs6000-c.c: Add macro definitions for
18428 ALTIVEC_BUILTIN_VEC_PERMXOR.
18429 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
18430 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
18431 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
18432 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
18434 * config/doc/extend.texi: Add prototypes for vec_permxor.
18436 2018-03-14 David Malcolm <dmalcolm@redhat.com>
18439 * diagnostic-show-locus.c (class layout_point): Convert m_line
18440 from int to linenum_type.
18441 (line_span::comparator): Use linenum "compare" function when
18442 comparing line numbers.
18443 (test_line_span): New function.
18444 (layout_range::contains_point): Convert param "row" from int to
18446 (layout_range::intersects_line_p): Likewise.
18447 (layout::will_show_line_p): Likewise.
18448 (layout::print_source_line): Likewise.
18449 (layout::should_print_annotation_line_p): Likewise.
18450 (layout::print_annotation_line): Likewise.
18451 (layout::print_leading_fixits): Likewise.
18452 (layout::annotation_line_showed_range_p): Likewise.
18453 (struct line_corrections): Likewise for field m_row.
18454 (line_corrections::line_corrections): Likewise for param "row".
18455 (layout::print_trailing_fixits): Likewise.
18456 (layout::get_state_at_point): Likewise.
18457 (layout::get_x_bound_for_row): Likewise.
18458 (layout::print_line): Likewise.
18459 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
18460 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
18461 * input.c (selftest::test_linenum_comparisons): New function.
18462 (selftest::input_c_tests): Call it.
18463 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
18464 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
18465 * selftest.h (ASSERT_GT): New macro.
18466 (ASSERT_GT_AT): New macro.
18467 (ASSERT_LT): New macro.
18468 (ASSERT_LT_AT): New macro.
18470 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
18472 PR rtl-optimization/84780
18473 * combine.c (distribute_links): Don't make a link based on pc_rtx.
18475 2018-03-14 Martin Liska <mliska@suse.cz>
18477 * tree.c (record_node_allocation_statistics): Use
18478 get_stats_node_kind.
18479 (get_stats_node_kind): New function extracted from
18480 record_node_allocation_statistics.
18481 (free_node): Use get_stats_node_kind.
18483 2018-03-14 Richard Biener <rguenther@suse.de>
18485 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
18486 that the value-set of ANTIC_IN doesn't grow.
18489 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18491 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18492 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18494 2018-03-14 Julia Koval <julia.koval@intel.com>
18496 * config.gcc (icelake-client, icelake-server): New.
18498 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
18499 (initial_ix86_arch_features): Ditto.
18500 (PTA_SKYLAKE): Add SGX.
18501 (PTA_ICELAKE): Remove.
18502 (PTA_ICELAKE_CLIENT): New.
18503 (PTA_ICELAKE_SERVER): New.
18504 (ix86_option_override_internal): Split up icelake on icelake client and
18506 (get_builtin_code_for_version): Ditto.
18507 (fold_builtin_cpu): Ditto.
18508 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
18509 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
18510 * config/i386/i386.h (processor_type): Ditto.
18511 * doc/invoke.texi: Ditto.
18513 2018-03-14 Jakub Jelinek <jakub@redhat.com>
18516 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
18517 INTEGER_CST offset, add it together with bitpos / 8 and
18518 sign extend based on POINTER_SIZE.
18522 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
18525 * config/i386/constraints.md (Yc): New register constraint.
18526 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
18527 Use Yc constraint for alternative 2 of operand 0. Remove
18528 preferred_for_speed attribute.
18530 2018-03-14 Richard Biener <rguenther@suse.de>
18532 PR tree-optimization/84830
18533 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
18534 with the old one to avoid oscillations.
18536 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
18539 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
18541 (assign_by_spills): Return a flag of reload assignment failure.
18542 Do not process the reload assignment failures. Do not spill other
18543 reload pseudos if they has the same reg class. Update n if
18545 (lra_assign): Add a return arg. Set up from the result of
18546 assign_by_spills call.
18547 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18548 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18549 usage_insns if it is not NULL.
18550 (spill_hard_reg_in_range): New function.
18551 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18552 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18553 function prototypes.
18554 (lra_assign): Change prototype.
18555 * lra.c (lra): Add code to deal with fails by splitting hard reg
18558 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
18560 * config/riscv/riscv.opt (mrelax): New option.
18561 * config/riscv/riscv.c (riscv_file_start): Emit ".option
18562 "norelax" when riscv_mrelax is disabled.
18563 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
18565 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18568 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
18569 reassociation for int modes.
18571 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18573 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
18574 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
18576 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
18577 * config/aarch64/aarch64-sve.md
18578 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
18579 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
18580 (*extend<mode><Vwide>2): Rename to...
18581 (aarch64_sve_extend<mode><Vwide>2): ...this.
18582 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
18583 renaming the old pattern to...
18584 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
18586 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
18587 define_expand, renaming the old pattern to...
18588 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
18589 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
18590 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
18591 account when deciding which SVE instruction the optab should use.
18592 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
18594 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18596 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
18597 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
18598 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
18599 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
18600 (tlsdesc_small_<mode>): Turn a define_expand and use
18601 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
18602 (tlsdesc_small_advsimd_<mode>): ...this.
18603 (tlsdesc_small_sve_<mode>): New pattern.
18605 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18607 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
18608 (UNSPEC_UMUL_HIGHPART): New constants.
18609 (MUL_HIGHPART): New int iteraor.
18610 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
18611 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
18613 (*<su>mul<mode>3_highpart): New define_insn.
18615 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
18618 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18621 2018-03-13 Martin Liska <mliska@suse.cz>
18624 * (sem_item_optimizer::sem_item_optimizer): Initialize new
18626 (sem_item_optimizer::~sem_item_optimizer): Release it.
18627 (sem_item_optimizer::merge_classes): Register variable aliases.
18628 (sem_item_optimizer::fixup_pt_set): New function.
18629 (sem_item_optimizer::fixup_points_to_sets): Likewise.
18630 * ipa-icf.h: Declare new variables and functions.
18632 2018-03-13 Jakub Jelinek <jakub@redhat.com>
18634 PR middle-end/84834
18635 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
18636 integer_pow2p@2 and test integer_pow2p in condition.
18637 (A < 0 ? C : 0): Similarly for @1.
18639 PR middle-end/84831
18640 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
18641 characters starting at p contain '\0' character, don't look beyond
18645 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
18646 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
18649 * reg-stack.c (change_stack): Change update_end var from int to
18650 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
18651 also call set_block_for_insn on the newly added insns and rescan.
18654 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
18655 on the last operand.
18658 * tree.c (stabilize_reference_1): Return save_expr (e) for
18659 STATEMENT_LIST even if it doesn't have side-effects.
18661 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
18663 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
18665 2018-03-12 Renlin Li <renlin.li@arm.com>
18667 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
18668 aarch64_output_scalar_simd_mov_immediate.
18670 2018-03-12 Martin Sebor <msebor@redhat.com>
18672 PR tree-optimization/83456
18673 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
18674 for perfectly overlapping calls to memcpy.
18675 (gimple_fold_builtin_memory_chk): Same.
18676 (gimple_fold_builtin_strcpy): Handle no-warning.
18677 (gimple_fold_builtin_stxcpy_chk): Same.
18678 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
18680 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
18682 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
18683 parameter. Use it for SFmode.
18684 (rs6000_function_arg_advance_1): Adjust.
18685 (rs6000_function_arg): Adjust.
18686 (rs6000_gimplify_va_arg): Pass false for that new parameter.
18688 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
18690 PR rtl-optimization/84169
18691 PR rtl-optimization/84780
18692 * combine.c (can_combine_p): Check for a 2-insn combination whether
18693 the destination register is used between the two insns, too.
18695 2018-03-12 Richard Biener <rguenther@suse.de>
18697 PR tree-optimization/84803
18698 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
18699 for refs DR analysis didn't process.
18701 2018-03-12 Richard Biener <rguenther@suse.de>
18703 PR tree-optimization/84777
18704 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
18705 force-vectorize loops ignore whether we are optimizing for size.
18707 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
18709 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
18710 (TARGET_MD_ASM_ADJUST): Define.
18712 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
18713 Kito Cheng <kito.cheng@gmail.com>
18714 Chung-Ju Wu <jasonwucj@gmail.com>
18716 * config/nds32/nds32.c (nds32_compute_stack_frame,
18717 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
18718 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
18719 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
18720 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
18721 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
18722 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
18723 * config/nds32/nds32.md (prologue, epilogue): Use macro
18724 NDS32_V3PUSH_AVAILABLE_P to do checking.
18726 2018-03-11 Jakub Jelinek <jakub@redhat.com>
18729 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
18730 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
18731 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
18732 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
18733 addition of most attributes on !orig_type_die or the attribute not
18734 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
18736 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18737 Chung-Ju Wu <jasonwucj@gmail.com>
18739 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
18740 __NDS32_VH__ macro.
18741 * config/nds32/nds32.opt (mvh): New option.
18743 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18744 Chung-Ju Wu <jasonwucj@gmail.com>
18746 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
18748 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
18749 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
18752 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18753 Chung-Ju Wu <jasonwucj@gmail.com>
18755 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
18757 * config/nds32/nds32-multiple.md (strlensi): New pattern.
18758 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
18760 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
18761 Kito Cheng <kito.cheng@gmail.com>
18762 Chung-Ju Wu <jasonwucj@gmail.com>
18764 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
18765 UNSPEC_FFMISM and UNSPEC_FLMISM.
18766 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
18767 for ffb, ffmism and flmism.
18768 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
18769 (unspec_ffmism): Ditto.
18770 (unspec_flmism): Ditto.
18771 (nds32_expand_builtin_impl): Check if string extension is available.
18772 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
18773 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
18775 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
18778 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18781 * lra-assigns.c (assign_by_spills): Return a flag of reload
18782 assignment failure. Do not process the reload assignment
18783 failures. Do not spill other reload pseudos if they has the same
18785 (lra_assign): Add a return arg. Set up from the result of
18786 assign_by_spills call.
18787 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18788 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18789 usage_insns if it is not NULL.
18790 (spill_hard_reg_in_range): New function.
18791 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18792 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18793 function prototypes.
18794 (lra_assign): Change prototype.
18795 * lra.c (lra): Add code to deal with fails by splitting hard reg
18798 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
18801 * config/i386/i386.opt: Replace Enforcment with Enforcement.
18803 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
18806 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
18807 (dw_val_node): Add val_symbolic_view.
18808 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
18809 (symview_upper_bound): New.
18810 (new_line_info_table): Initialize symviews_since_reset.
18811 (dwarf2out_source_line): Count symviews_since_reset and set
18812 symview_upper_bound.
18813 (dw_val_equal_p): Handle symview.
18814 (add_AT_symview): New.
18815 (print_dw_val): Handle symview.
18816 (attr_checksum, attr_checksum_ordered): Likewise.
18817 (same_dw_val_p, size_of_die): Likewise.
18818 (value_format, output_die): Likewise.
18819 (add_high_low_attributes): Use add_AT_symview for entry_view.
18820 (dwarf2out_finish): Reset symview_upper_bound, clear
18823 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
18826 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
18827 Add strict argument and use it.
18828 (rs6000_split_multireg_move): Update for new strict argument.
18829 (mem_operand_gpr): Disallow all non-offsettable addresses.
18830 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
18832 2018-03-09 Jakub Jelinek <jakub@redhat.com>
18835 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
18836 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
18837 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
18840 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
18841 decl, use remap_type if we want to use the type.
18843 2018-03-09 Martin Sebor <msebor@redhat.com>
18845 PR tree-optimization/84526
18846 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18848 (builtin_access::generic_overlap): Be prepared to handle non-array
18851 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
18853 PR rtl-optimization/84682
18854 * lra-constraints.c (process_address_1): Check is_address flag
18855 for address constraints.
18856 (process_alt_operands): Likewise.
18857 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
18858 preprocess_constraints.
18859 * recog.h (preprocess_constraints): Add oploc parameter.
18861 * recog.c (preprocess_constraints): Test address_operand for
18862 CT_ADDRESS constraints.
18864 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18867 * lra-assigns.c (assign_by_spills): Return a flag of reload
18868 assignment failure. Do not process the reload assignment
18869 failures. Do not spill other reload pseudos if they has the same
18871 (lra_assign): Add a return arg. Set up from the result of
18872 assign_by_spills call.
18873 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18874 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18875 usage_insns if it is not NULL.
18876 (spill_hard_reg_in_range): New function.
18877 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18878 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18879 function prototypes.
18880 (lra_assign): Change prototype.
18881 * lra.c (lra): Add code to deal with fails by splitting hard reg
18884 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18887 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
18888 Accept complain bool parameter. Only emit errors if it is true.
18889 (arm_parse_cpu_option_name): Likewise.
18890 (arm_target_thumb_only): Adjust callers of the above.
18891 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
18892 prototype to take a default true bool parameter.
18893 (arm_parse_arch_option_name): Likewise.
18895 2018-03-09 David Malcolm <dmalcolm@redhat.com>
18896 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
18900 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
18901 * configure: Regenerate.
18902 * configure.ac ("linker --version-script option"): New.
18903 ("linker soname option"): New.
18905 2018-03-09 Richard Biener <rguenther@suse.de>
18907 PR tree-optimization/84775
18908 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
18909 immediate uses of predicate stmts and mark them modified.
18912 PR tree-optimization/84178
18913 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18915 (version_loop_for_if_conversion): Delay update_ssa call.
18916 (tree_if_conversion): Delay update_ssa until after predicate
18919 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
18922 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
18923 when the function accesses prior frames.
18925 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18928 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
18929 gen_llsym, otherwise call maybe_gen_llsym.
18931 PR inline-asm/84742
18932 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
18933 has ',' character inside of it.
18935 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18938 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
18939 as clobbering CC_REGNUM.
18941 2018-03-08 Richard Biener <rguenther@suse.de>
18943 PR middle-end/84552
18944 * tree-scalar-evolution.c: Include tree-into-ssa.h.
18945 (follow_copies_to_constant): Do not follow SSA names registered
18948 2018-03-08 Richard Biener <rguenther@suse.de>
18950 PR tree-optimization/84178
18951 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18953 (version_loop_for_if_conversion): Delay update_ssa call.
18954 (tree_if_conversion): Delay update_ssa until after predicate
18957 2018-03-08 David Malcolm <dmalcolm@redhat.com>
18959 PR tree-optimization/84178
18960 * tree-if-conv.c (release_bb_predicate): Remove the
18961 the assertion that the stmts have NULL use_ops.
18962 Discard the statements, asserting that they haven't
18963 yet been added to a BB.
18965 2018-03-08 Richard Biener <rguenther@suse.de>
18967 PR tree-optimization/84746
18968 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
18969 (phi_translate): Pass in destination ANTIC_OUT set.
18970 (phi_translate_1): Likewise. For a simplified result lookup
18971 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
18972 (phi_translate_set): Adjust.
18973 (do_pre_regular_insertion): Likewise.
18974 (do_pre_partial_partial_insertion): Likewise.
18976 2018-03-08 Martin Liska <mliska@suse.cz>
18978 PR gcov-profile/84735
18979 * doc/gcov.texi: Document usage of profile files.
18980 * gcov-io.h: Document changes in the format.
18982 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
18986 * dwarf2out.c (struct dw_line_info_table): Update comments for
18988 (FORCE_RESET_NEXT_VIEW): New.
18989 (FORCE_RESETTING_VIEW_P): New.
18990 (RESETTING_VIEW_P): Check for -1 too.
18991 (ZERO_VIEW_P): Likewise.
18992 (new_line_info_table): Force-reset next view.
18993 (dwarf2out_begin_function): Likewise.
18994 (dwarf2out_source_line): Simplify zero_view_p initialization.
18995 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
18996 view directly. Omit view when omitting .loc at line 0.
18998 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19000 PR tree-optimization/84740
19001 * tree-switch-conversion.c (process_switch): Call build_constructors
19002 only if info.phi_count is non-zero.
19004 PR tree-optimization/84739
19005 * tree-tailcall.c (find_tail_calls): Check call arguments against
19006 DECL_ARGUMENTS (current_function_decl) rather than
19007 DECL_ARGUMENTS (func) when checking for tail recursion.
19009 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19011 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19012 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19013 Volker Reichelt's entry and add entries for people that perform
19014 GCC fuzzy testing and report numerous bugs.
19016 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
19019 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19020 readonly data in sdata, if that is disabled.
19021 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19022 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19023 -mreadonly-in-sdata option.
19025 2018-03-07 Martin Sebor <msebor@redhat.com>
19027 PR tree-optimization/84468
19028 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19029 basic block when looking for nul assignment.
19031 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
19034 * except.h (output_function_exception_table): Adjust prototype.
19035 * except.c (output_function_exception_table): Remove FNNAME parameter
19036 and add SECTION parameter. Ouput one part of the table at a time.
19037 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19038 the first part of the exception table and emit unwind directives.
19039 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19040 (i386_pe_seh_cold_init): Likewise.
19041 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19042 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19043 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19044 (ix86_output_call_insn): Emit a nop in one more case for SEH.
19045 * config/i386/winnt.c: Include except.h.
19046 (struct seh_frame_state): Add reg_offset, after_prologue and
19047 in_cold_section fields.
19048 (i386_pe_seh_end_prologue): Set seh->after_prologue.
19049 (i386_pe_seh_cold_init): New function.
19050 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19051 to seh->in_cold_section.
19052 (seh_emit_push): Record the offset of the push.
19053 (seh_emit_save): Record the offet of the save.
19054 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19055 Test seh->after_prologue to disregard the epilogue.
19056 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19057 (i386_pe_end_cold_function): New function.
19059 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19062 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19063 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19066 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19068 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19069 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19071 PR middle-end/84723
19072 * multiple_target.c: Include tree-inline.h and intl.h.
19073 (expand_target_clones): Diagnose and fail if node->definition and
19074 !tree_versionable_function_p (node->decl).
19076 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
19078 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19080 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19081 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19082 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19084 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19087 * combine.c (try_combine): Use reg_or_subregno instead of handling
19088 just paradoxical SUBREGs and REGs.
19090 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
19092 * config/arc/arc.c (arc_finalize_pic): Remove function.
19093 (arc_must_save_register): We use single base PIC register, remove
19094 checks to save/restore the PIC register.
19095 (arc_expand_prologue): Likewise.
19096 * config/arc/arc-protos.h (arc_set_default_type_attributes):
19098 (arc_verify_short): Likewise.
19099 (arc_attr_type): Likewise.
19100 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19101 (walk_stores): Likewise.
19102 (arc_address_cost): Make it static.
19103 (arc_verify_short): Likewise.
19104 (branch_dest): Likewise.
19105 (arc_attr_type): Likewise.
19106 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19107 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19108 (arc_final_prescan_insn): Remove inserting the nops due to
19109 hardware hazards. It is done in reorg step.
19110 (insn_length_variant_t): Remove.
19111 (insn_length_parameters_t): Likewise.
19112 (arc_insn_length_parameters): Likewise.
19113 (arc_get_insn_variants): Likewise.
19114 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19116 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19118 PR inline-asm/84683
19119 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19122 PR tree-optimization/84687
19123 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19125 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19127 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19129 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19130 Rename to ppc_speculation_barrier.
19131 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19132 __builtin_ppc_speculation_barrier.
19134 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19137 * combine.c (combine_simplify_rtx): Don't try to simplify if
19138 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19141 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
19143 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19144 to 32 bytes when compiling for POWER9.
19146 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19149 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19150 regparm >= 3 with no arg reg available also for calls with
19151 flag_force_indirect_call. Pass decl to ix86_function_regparm.
19154 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19156 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
19158 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
19161 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19163 2018-03-05 Richard Biener <rguenther@suse.de>
19165 PR tree-optimization/84486
19166 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19167 When inserting a __builtin_assume_aligned call set the LHS
19168 SSA name alignment info accordingly.
19170 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
19172 PR tree-optimization/84114
19173 * config/aarch64/aarch64.c (aarch64_reassociation_width)
19174 Avoid reassociation of FLOAT_MODE addition.
19176 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
19178 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19179 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19180 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19181 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19182 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19183 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19184 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19186 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19187 __builtin_ia32_wbinvd): New builtins.
19188 (SPECIAL_ARGS2): New.
19189 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19190 (SPECIAL_ARGS2): New.
19191 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19192 (ix86_valid_target_attribute_inner_p): Ditto.
19193 (ix86_init_mmx_sse_builtins): Add special_args2.
19194 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19195 TARGET_WBNOINVD_P): New.
19196 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19197 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19198 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19199 * config/i386/immintrin.h (_wbinvd): New intrinsic.
19200 * config/i386/pconfigintrin.h: New file.
19201 * config/i386/wbnoinvdintrin.h: Ditto.
19202 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19204 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19206 2018-03-05 Richard Biener <rguenther@suse.de>
19208 PR tree-optimization/84670
19209 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19211 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19212 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19213 (compute_antic_aux): Only assert the number of values in ANTIC_IN
19214 doesn't grow if all successors (recursively) were visited at least
19217 2018-03-05 Richard Biener <rguenther@suse.de>
19219 PR tree-optimization/84650
19220 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19221 if executed in the loop pipeline.
19223 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
19225 * doc/configfiles.texi (Configuration Files): Move info about
19226 conditionalizing $target-protos.h to...
19227 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
19228 differs from $target-protos.h.
19230 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
19231 Chung-Ju Wu <jasonwucj@gmail.com>
19233 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19234 * config/nds32/nds32-multiple.md (setmemsi): Define.
19235 * config/nds32/nds32-memory-manipulation.c
19236 (nds32_gen_dup_4_byte_to_word_value): New.
19237 (emit_setmem_word_loop): New.
19238 (emit_setmem_byte_loop): New.
19239 (nds32_expand_setmem_loop): New.
19240 (nds32_expand_setmem_loop_v3m): New.
19241 (nds32_expand_setmem_unroll): New.
19242 (nds32_expand_setmem): New.
19244 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19245 Chung-Ju Wu <jasonwucj@gmail.com>
19247 * config/nds32/nds32-memory-manipulation.c
19248 (nds32_emit_load_store): New.
19249 (nds32_emit_post_inc_load_store): New.
19250 (nds32_emit_mem_move): New.
19251 (nds32_emit_mem_move_block): New.
19252 (nds32_expand_movmemsi_loop_unknown_size): New.
19253 (nds32_expand_movmemsi_loop_known_size): New.
19254 (nds32_expand_movmemsi_loop): New.
19255 (nds32_expand_movmemsi_unroll): New.
19256 (nds32_expand_movmemqi): Rename ...
19257 (nds32_expand_movmemsi): ... to this.
19258 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19259 (movmemsi): ... to this.
19260 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19261 (nds32_expand_movmemsi): ... to this.
19263 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19264 Monk Chiang <sh.chiang04@gmail.com>
19265 Chung-Ju Wu <jasonwucj@gmail.com>
19267 * config/nds32/nds32-protos.h
19268 (nds32_expand_load_multiple): New arguments.
19269 (nds32_expand_store_multiple): Ditto.
19270 (nds32_valid_multiple_load_store): Rename ...
19271 (nds32_valid_multiple_load_store_p): ... to this.
19272 * config/nds32/nds32-memory-manipulation.c
19273 (nds32_expand_load_multiple): Refine implementation.
19274 (nds32_expand_store_multiple): Ditto.
19275 * config/nds32/nds32-multiple.md
19276 (load_multiple): Update nds32_expand_load_multiple interface.
19277 (store_multiple): Update nds32_expand_store_multiple interface.
19278 * config/nds32/nds32-predicates.c
19279 (nds32_valid_multiple_load_store): Rename ...
19280 (nds32_valid_multiple_load_store_p): ... to this and refine
19282 * config/nds32/predicates.md
19283 (nds32_load_multiple_and_update_address_operation): New predicate.
19284 (nds32_store_multiple_and_update_address_operation): New predicate.
19286 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19287 Chung-Ju Wu <jasonwucj@gmail.com>
19289 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19290 (combo): New attribute.
19291 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19293 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
19295 * config/nds32/nds32.opt: Change -mcmodel= default value.
19297 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
19298 Monk Chiang <sh.chiang04@gmail.com>
19299 Chung-Ju Wu <jasonwucj@gmail.com>
19301 * config/nds32/constants.md (unspec_element): New enum.
19302 * config/nds32/constraints.md (Umw): New constraint.
19303 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19304 * config/nds32/nds32-intrinsic.md: Likewise.
19305 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19306 (nds32_valid_smw_lwm_base_p): New.
19307 (nds32_output_smw_single_word): New.
19308 (nds32_output_lmw_single_word): New.
19309 (nds32_expand_unaligned_load): New.
19310 (nds32_expand_unaligned_store): New.
19311 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19312 (nds32_output_smw_single_word): Declare.
19313 (nds32_output_lmw_single_word): Declare.
19314 (nds32_expand_unaligned_load): Declare.
19315 (nds32_expand_unaligned_store): Declare.
19316 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19317 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19318 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19319 NDS32_BUILTIN_UASTORE_DW.
19320 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19323 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
19324 Kito Cheng <kito.cheng@gmail.com>
19325 Chung-Ju Wu <jasonwucj@gmail.com>
19327 * config/nds32/nds32-intrinsic.c
19328 (nds32_expand_builtin_null_ftype_reg): Delete.
19329 (nds32_expand_builtin_reg_ftype_imm): Ditto.
19330 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19331 (nds32_read_argument): New.
19332 (nds32_legitimize_target): Ditto.
19333 (nds32_legitimize_argument): Ditto.
19334 (nds32_check_constant_argument): Ditto.
19335 (nds32_expand_unop_builtin): Ditto.
19336 (nds32_expand_unopimm_builtin): Ditto.
19337 (nds32_expand_binop_builtin): Ditto.
19338 (nds32_builtin_decl_impl): Ditto.
19339 (builtin_description): Ditto.
19340 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
19341 (nds32_init_builtins_impl): Ditto.
19342 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
19343 (nds32_builtin_decl): New.
19344 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
19345 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
19347 2018-03-02 Jeff Law <law@redhat.com>
19349 * reorg.c (stop_search_p): Handle DEBUG_INSN.
19350 (redundant_insn, fill_simple_delay_slots): Likewise.
19351 (fill_slots_from_thread): Likewise.
19352 * resource.c (mark_referenced_resources): Likewise.
19353 (mark_set_resources, find_dead_or_set_registers): Likewise.
19355 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19357 * substring-locations.h (format_warning_va): Formatting fix for
19358 ATTRIBUTE_GCC_DIAG.
19359 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
19361 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
19362 * substring-locations.c: Include intl.h.
19363 (format_warning_va): Turned into small wrapper around
19364 format_warning_n_va, renamed to ...
19365 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
19366 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
19368 (format_warning_at_substring_n): New function.
19369 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
19370 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
19371 format_warning_at_substring with just a shorter name instead of
19372 const function pointer.
19373 (fmtwarn_n): New function.
19374 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
19375 appropriate, get rid of all the fmtstr temporaries, move conditionals
19376 with G_() wrapped string literals directly into fmtwarn arguments,
19377 cast dir.len to (int), formatting fixes.
19379 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
19381 * doc/invoke.texi: Remove "Cilk Plus" references.
19383 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19384 Richard Biener <rguenther@suse.de>
19387 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
19388 for error or warning attributes if CALL_FROM_THUNK_P is set.
19391 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19394 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
19395 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
19398 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
19399 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
19401 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
19402 instead of -1U in last predictors element's probability member.
19404 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
19407 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19408 arguments if they are comparable.
19410 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
19412 PR tree-optimization/84634
19413 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
19414 masks and masked_loop_p with a single loop_masks, making sure it's
19415 null for bb vectorization.
19417 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
19419 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
19420 (vect_analyze_data_ref_access): Use loop->safe_len rather than
19421 loop->force_vectorize to check whether there is no alias.
19423 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19426 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
19428 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
19430 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
19431 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
19432 instead of a loop around prev_real_insn.
19433 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
19436 PR inline-asm/84625
19437 * config/i386/i386.c (ix86_print_operand): Use conditional
19438 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
19441 2018-03-02 Richard Biener <rguenther@suse.de>
19443 PR tree-optimization/84427
19444 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
19445 (bitmap_set_subtract_values): Rewrite to handle multiple
19448 (prune_clobbered_mems): Likewise.
19449 (phi_translate): Take edge instead of pred/phiblock.
19450 (phi_translate_1): Likewise.
19451 (phi_translate_set): Likewise. Insert all translated
19452 exprs for a value into the set, keeping possibly multiple
19453 expressions per value.
19454 (compute_antic_aux): Adjust for phi_translate changes.
19455 When intersecting union the expressions and prune those
19456 not in the final value set, keeping possibly multiple
19457 expressions per value. Do not use value-insertion
19458 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
19459 all expressions. Add verification that the value-sets
19460 only shrink during iteration.
19461 (compute_partial_antic_aux): Adjust for the phi_translate changes.
19462 (do_pre_regular_insertion): Likewise.
19463 (do_pre_partial_partial_insertion): Likewise.
19465 2018-03-02 Richard Biener <rguenther@suse.de>
19468 * config/darwin.c (saved_debug_info_level): New static global.
19469 (darwin_asm_lto_start): Disable debug info generation for LTO out.
19470 (darwin_asm_lto_end): Restore debug info generation settings.
19472 2018-03-01 Martin Liska <mliska@suse.cz>
19475 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
19476 volatile arguments.
19478 2018-03-01 Richard Biener <rguenther@suse.de>
19481 * dwarf2out.c (gen_variable_die): Properly handle late VLA
19482 type annotation with LTO when debug was disabled at compile-time.
19484 2018-03-01 Matthew Fortune <mfortune@gmail.com>
19486 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
19488 (mips_final_postscan_insn): Likewise.
19490 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
19492 PR rtl-optimization/84528
19493 * alias.c (init_alias_target): Add commentary.
19494 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
19495 a unique base value if the frame pointer is not eliminated
19496 to the stack pointer.
19498 2018-03-01 Tom de Vries <tom@codesourcery.com>
19500 PR rtl-optimization/83327
19501 * lra-int.h (hard_regs_spilled_into): Declare.
19502 * lra.c (hard_regs_spilled_into): Define.
19503 (init_reg_info): Init hard_regs_spilled_into.
19504 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
19505 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
19506 (process_bb_lives): Handle hard_regs_spilled_into.
19507 (lra_create_live_ranges_1): Before doing liveness propagation, clear
19508 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
19510 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
19512 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
19513 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
19514 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
19515 * config/rs6000/aix72.h: New file.
19517 2018-02-28 Jakub Jelinek <jakub@redhat.com>
19519 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
19520 instead of warning_at with conditional singular and plural messages
19524 * stor-layout.c (update_alignment_for_field): For
19525 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
19526 && !DECL_PACKED (field), do the alignment update, just use
19527 only desired_align instead of MAX (type_align, desired_align)
19529 (place_field): Don't do known_align < desired_align handling
19530 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
19531 is non-NULL, instead do it after rli->prev_field handling and
19532 only if not within a bitfield word. For DECL_PACKED (field)
19533 use type_align of BITS_PER_UNIT.
19535 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
19537 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
19538 superfluous parentheses and trailing spaces.
19540 2018-02-28 Richard Biener <rguenther@suse.de>
19542 PR tree-optimization/84584
19543 * graphite-scop-detection.c (scop_detection::add_scop): Discard
19544 SCoPs with fake exit edge.
19546 2018-02-28 Martin Liska <mliska@suse.cz>
19549 * timevar.c (timer::print): Fix format to properly print 100%
19552 2018-02-28 Richard Biener <rguenther@suse.de>
19554 PR middle-end/84607
19555 * genmatch.c (capture_info::walk_match): Do not mark
19556 captured expressions without operands as expr_p given
19557 they act more like predicates and should be subject to
19558 "lost tail" side-effect preserving.
19560 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
19562 PR rtl-optimization/81611
19563 * auto-inc-dec.c (attempt_change): Move dead note from
19564 mem_insn if it's the next use of regno
19565 (find_address): Take address use of reg holding
19566 non-incremented value. Add parm to limit search to the named
19568 (merge_in_block): Attempt to use a mem insn that is the next
19569 use of the original regno.
19571 2018-02-27 Martin Sebor <msebor@redhat.com>
19574 * doc/invoke.texi (-Wmissing-attributes): New option.
19575 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
19577 2018-02-27 Martin Sebor <msebor@redhat.com>
19579 PR translation/84207
19580 * diagnostic-core.h (warning_n, error_n, inform_n): Change
19581 n argument to unsigned HOST_WIDE_INT.
19582 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
19583 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
19584 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
19585 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
19587 2018-02-27 Richard Biener <rguenther@suse.de>
19589 PR tree-optimization/84512
19590 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19591 Do not use the estimate returned from record_stmt_cost for
19592 the scalar iteration cost but sum properly using add_stmt_cost.
19594 2018-02-27 Richard Biener <rguenther@suse.de>
19596 PR tree-optimization/84466
19597 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19598 Adjust last change to less strictly validate use operands.
19600 2018-02-27 Martin Liska <mliska@suse.cz>
19602 PR gcov-profile/84548
19603 * gcov.c (process_file): Allow partial overlap and consider it
19604 also as group functions.
19605 (output_lines): Properly calculate range of lines for a group.
19607 2018-02-27 Martin Liska <mliska@suse.cz>
19609 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
19610 'ggc' suffixes. Change first column width.
19611 (timer::print): Fix formatting of the column.
19613 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
19615 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
19616 preserve inline entry blocks for the sake of debug inline
19617 entry point markers alone.
19618 (remove_unused_locals): Suggest in comments a better place to
19619 force the preservation of inline entry blocks that are
19620 otherwise unused, but do not preserve them.
19622 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19624 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
19626 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19629 * config/i386/constraints.md (Bs): Replace
19630 ix86_indirect_branch_register with
19631 TARGET_INDIRECT_BRANCH_REGISTER.
19633 * config/i386/i386.md (indirect_jump): Likewise.
19634 (tablejump): Likewise.
19635 (*sibcall_memory): Likewise.
19636 (*sibcall_value_memory): Likewise.
19637 Peepholes of indirect call and jump via memory: Likewise.
19638 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
19639 (*sibcall_value_GOT_32): Likewise.
19640 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19641 (GOT_memory_operand): Likewise.
19642 (call_insn_operand): Likewise.
19643 (sibcall_insn_operand): Likewise.
19644 (GOT32_symbol_operand): Likewise.
19645 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
19647 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
19649 PR rtl-optimization/83496
19650 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
19651 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
19652 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
19653 redundant insn, if any.
19654 (relax_delay_slots): Likewise.
19655 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
19657 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
19659 PR tree-optimization/83965
19660 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
19661 that grouped statements are part of a reduction chain. Return
19662 true if the statement is not marked as a reduction itself but
19663 is part of a group.
19664 (vect_recog_dot_prod_pattern): Don't check whether the statement
19665 is part of a group here.
19666 (vect_recog_sad_pattern): Likewise.
19667 (vect_recog_widen_sum_pattern): Likewise.
19669 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
19672 * final.c (rest_of_clean_state): Also look for calls inside sequences.
19674 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19677 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
19679 (ix86_output_indirect_function_return): New prototype.
19680 (ix86_split_simple_return_pop_internal): Likewise.
19681 * config/i386/i386.c (indirect_return_via_cx): New.
19682 (indirect_return_via_cx_bnd): Likewise.
19683 (indirect_thunk_name): Handle return va CX_REG.
19684 (output_indirect_thunk_function): Create alias for
19685 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
19686 (ix86_output_indirect_jmp): Remove the bool argument.
19687 (ix86_output_indirect_function_return): New function.
19688 (ix86_split_simple_return_pop_internal): Likewise.
19689 * config/i386/i386.md (*indirect_jump): Don't pass false
19690 to ix86_output_indirect_jmp.
19691 (*tablejump_1): Likewise.
19692 (simple_return_pop_internal): Change it to define_insn_and_split.
19693 Call ix86_split_simple_return_pop_internal to split it for
19694 -mfunction-return=.
19695 (simple_return_indirect_internal): Call
19696 ix86_output_indirect_function_return instead of
19697 ix86_output_indirect_jmp.
19699 2018-02-26 Jakub Jelinek <jakub@redhat.com>
19702 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
19703 memset and value initialization afterwards.
19705 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
19707 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
19709 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19712 * common/config/aarch64/aarch64-common.c
19713 (aarch_option_optimization_table[]): Switch
19714 off fomit-frame-pointer
19716 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19717 Chung-Ju Wu <jasonwucj@gmail.com>
19719 * config/nds32/nds32-multiple.md (load_multiple): Disallow
19721 (store_multiple): Ditto.
19723 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19725 * config.gcc: Add --with-cpu support for nds32 target.
19726 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
19727 * config/nds32/nds32.opt: Add -mcpu= option.
19729 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
19731 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
19732 isel=yes): Warn for these deprecated options.
19734 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
19736 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
19737 ISA_2_5_MASKS_EMBEDDED.
19739 2018-02-23 Jakub Jelinek <jakub@redhat.com>
19741 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
19742 p->max as pointers rather than using iterative_hash_expr.
19744 2018-02-23 Carl Love <cel@us.ibm.com>
19746 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
19747 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
19749 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
19750 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
19751 P8V_BUILTIN_VEC_VUNSIGNED2.
19753 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
19756 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
19757 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
19759 * lra-constraints.c (curr_insn_transform): Set up
19760 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
19762 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
19763 * lra-eliminations.c (spill_pseudos): Ditto.
19764 (process_insn_for_elimination): Ditto.
19765 * lra-lives.c (reg_early_clobber_p): Use the new macros.
19766 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
19767 LRA_NON_CLOBBERED_ALT.
19769 2018-02-22 Martin Sebor <msebor@redhat.com>
19771 PR tree-optimization/84480
19772 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
19773 to maybe_diag_stxncpy_trunc. Call it.
19774 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
19775 from gimple_fold_builtin_strcpy. Print inlining stack.
19776 (handle_builtin_stxncpy): Print inlining stack.
19777 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
19779 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
19782 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
19783 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
19784 and -fcheck-pointer-bounds are used together.
19785 (indirect_thunk_prefix): New enum.
19786 (indirect_thunk_need_prefix): New function.
19787 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
19788 "_nt" instead of "_bnd" for NOTRACK prefix.
19789 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
19790 (output_indirect_thunk_function): Likewise.
19792 (ix86_code_end): Update output_indirect_thunk_function calls.
19793 (ix86_output_indirect_branch_via_reg): Replace
19794 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
19795 (ix86_output_indirect_branch_via_push): Likewise.
19796 (ix86_output_function_return): Likewise.
19797 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
19798 incompatible with -fcf-protection=branch and
19799 -fcheck-pointer-bounds.
19801 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19804 * config/aarch64/aarch64.c (aarch64_print_address_internal):
19805 Change gcc_assert call to output_operand_lossage.
19807 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19809 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
19811 2018-02-22 DJ Delorie <dj@redhat.com>
19812 Sebastian Perta <sebastian.perta@renesas.com>
19813 Oleg Endo <olegendo@gcc.gnu.org>
19815 * config/rx/rx.c (rx_rtx_costs): New function.
19816 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
19818 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
19820 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
19822 2018-02-22 Martin Liska <mliska@suse.cz>
19825 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
19826 Add "native" as a possible value.
19828 2018-02-22 Martin Liska <mliska@suse.cz>
19831 * config/i386/i386.c (ix86_option_override_internal):
19832 Add "native" as a possible value for -march and -mtune.
19834 2018-02-22 Jakub Jelinek <jakub@redhat.com>
19837 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
19838 to all type variants.
19840 PR tree-optimization/84503
19841 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
19842 width as info->bitpos + info->bitsize - start.
19843 (merged_store_group::merge_overlapping): Simplify width computation.
19844 (check_no_overlap): New function.
19845 (imm_store_chain_info::try_coalesce_bswap): Compute expected
19846 start + width and last_order of the group, fail if check_no_overlap
19848 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
19849 to group if check_no_overlap fails.
19851 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19853 * config/rs6000/altivec.md: Delete contraint arguments to
19854 define_expand, define_split, and define_peephole2, and in
19855 define_insn_and_split if always unused.
19856 * config/rs6000/darwin.md: Ditto.
19857 * config/rs6000/dfp.md: Ditto.
19858 * config/rs6000/rs6000.md: Ditto.
19859 * config/rs6000/sync.md: Ditto.
19860 * config/rs6000/vector.md: Ditto.
19861 * config/rs6000/vsx.md: Ditto.
19863 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19865 * config/rs6000/altivec.md: Write output control strings as braced
19866 blocks instead of double-quoted strings.
19867 * config/rs6000/darwin.md: Ditto.
19868 * config/rs6000/rs6000.md: Ditto.
19869 * config/rs6000/vector.md: Ditto.
19870 * config/rs6000/vsx.md: Ditto.
19872 2018-02-21 Jason Merrill <jason@redhat.com>
19874 PR c++/84314 - ICE with templates and fastcall attribute.
19875 * attribs.c (build_type_attribute_qual_variant): Remove assert.
19877 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19879 * ipa-cp.c (determine_versionability): Fix comment typos.
19881 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19884 * ipa-cp.c (determine_versionability): Do not version functions caling
19887 2018-02-21 Martin Liska <mliska@suse.cz>
19890 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
19891 Add "native" as a possible value.
19892 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
19893 the macro when native cpu detection is available.
19895 2018-02-21 Martin Liska <mliska@suse.cz>
19898 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
19899 Add "native" as a possible value.
19900 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
19901 when native cpu detection is available.
19903 2018-02-21 Jakub Jelinek <jakub@redhat.com>
19904 Martin Sebor <msebor@redhat.com>
19906 PR tree-optimization/84478
19907 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
19909 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
19910 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
19911 support which is conservatively correct, for 2 only stay conservative
19912 for maxlen. Formatting and comment capitalization fixes. Add STRICT
19913 argument to the 2 argument get_range_strlen, adjust 6 arg
19914 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
19916 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
19917 (gimple_fold_builtin_strlen): Pass true as last argument to
19920 2018-02-20 Martin Sebor <msebor@redhat.com>
19922 PR middle-end/84095
19923 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
19924 (builtin_memref::set_base_and_offset): Same. Handle inner references.
19925 (builtin_memref::builtin_memref): Factor out parts into
19926 set_base_and_offset and call it.
19928 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19930 PR middle-end/84406
19931 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
19932 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
19933 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
19934 search at the associated MODE_INT.
19936 2018-02-20 Jeff Law <law@redhat.com>
19938 PR middle-end/82123
19939 PR tree-optimization/81592
19940 PR middle-end/79257
19941 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
19942 for range data rather than using global data.
19943 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
19944 range data rather than using global data.
19945 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
19946 pass it to children as needed.
19947 (struct directive::fmtresult): Similarly.
19948 (struct directive::set_width): Similarly.
19949 (struct directive::set_precision): Similarly.
19950 (format_integer, format_directive, parse_directive): Similarly.
19951 (format_none): Accept unnamed vr_values parameter.
19952 (format_percent, format_floating, format_character): Similarly.
19953 (format_string, format_plain): Similarly.
19954 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
19955 the EVRP range analyzer for range data rather than using global data.
19956 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
19957 gimple-ssa-evrp-analyze.h
19958 (class sprintf_dom_walker): Add after_dom_children member function.
19959 Add evrp_range_analyzer member.
19960 (sprintf_dom_walker::before_dom_children): Call into the EVRP
19961 range analyzer as needed.
19962 (sprintf_dom_walker::after_dom_children): New member function.
19963 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
19965 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
19966 (evrp_range_analyzer::pop_to_marker): Likewise.
19968 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19970 PR tree-optimization/84419
19971 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
19972 with the required type if its current type is compatible but
19975 2018-02-20 Jakub Jelinek <jakub@redhat.com>
19977 PR middle-end/82004
19978 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
19979 after vectorization.
19981 2018-02-20 Martin Liska <mliska@suse.cz>
19984 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
19985 possible values if we don't have a hint.
19987 2018-02-20 Martin Liska <mliska@suse.cz>
19991 * final.c (shorten_branches): Build align_tab array with one
19993 * opts.c (finish_options): Add alignment option limit check.
19994 (MAX_CODE_ALIGN): Likewise.
19995 (MAX_CODE_ALIGN_VALUE): Likewise.
19996 * doc/invoke.texi: Document maximum allowed option value for
19997 all -falign-* options.
19999 2018-02-19 Jakub Jelinek <jakub@redhat.com>
20002 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20003 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20004 * var-tracking.c (emit_note_insn_var_location): Remove all references
20005 to NOTE_INSN_CALL_ARG_LOCATION.
20006 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20007 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20008 Use copy_rtx_if_shared.
20009 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20010 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20011 (dwarf2out_var_location): Remove handling of
20012 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20014 * final.c (final_scan_insn): Remove all references to
20015 NOTE_INSN_CALL_ARG_LOCATION.
20016 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
20017 before dumping final insns.
20018 * except.c (emit_note_eh_region_end): Remove all references to
20019 NOTE_INSN_CALL_ARG_LOCATION.
20020 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20021 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20022 * config/arc/arc.c (hwloop_optimize): Likewise.
20023 * config/arm/arm.c (create_fix_barrier): Likewise.
20024 * config/s390/s390.c (s390_chunkify_start): Likewise.
20025 * config/sh/sh.c (find_barrier): Likewise.
20026 * config/i386/i386.c (rest_of_insert_endbranch,
20027 ix86_seh_fixup_eh_fallthru): Likewise.
20028 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20029 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20030 * config/frv/frv.c (frv_function_prologue): Likewise.
20031 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
20033 (note_outside_basic_block_p): Remove all references to
20034 NOTE_INSN_CALL_ARG_LOCATION.
20035 * gengtype.c (adjust_field_rtx_def): Likewise.
20036 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20038 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20039 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20042 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20045 PR tree-optimization/84452
20046 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20047 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20050 2018-02-19 Martin Liska <mliska@suse.cz>
20053 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20055 2018-02-19 Martin Liska <mliska@suse.cz>
20056 Richard Sandiford <richard.sandiford@linaro.org>
20058 PR tree-optimization/82491
20059 * gimple-fold.c (get_base_constructor): Make earlier bail out
20062 2018-02-19 Carl Love <cel@us.ibm.com>
20064 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20065 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20067 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20068 P8V_BUILTIN_VEC_NEG.
20070 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
20072 * config/rl78/rl78.md (movdf): New define expand.
20074 2018-02-19 Martin Liska <mliska@suse.cz>
20077 * doc/invoke.texi: Fix typo.
20078 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20080 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
20082 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20083 handle rs6000_single_float and rs6000_double_float specially for
20086 2018-02-16 Jeff Law <law@redhat.com>
20088 * config/rx/rx.c (add_pop_cfi_notes): New function.;
20089 (pop_regs): Use it.
20091 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20094 * ipa-inline.c (inline_small_functions): Fix a typo.
20096 2018-02-16 Nathan Sidwell <nathan@acm.org>
20098 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20100 2018-02-16 Carl Love <cel@us.ibm.com>
20102 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20103 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20104 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20105 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20106 expansion to P8V_BUILTIN_VEC_FLOAT2.
20108 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
20110 PR rtl-optimization/70023
20111 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20112 src_regno into account.
20114 2018-02-16 Carl Love <cel@us.ibm.com>
20116 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20117 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20118 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20119 * config/rs6000/rs6000.c: Remove case statements for
20120 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20121 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20122 and P9V_BUILTIN_VEC_VINSERT4B.
20123 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20124 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20125 * config/rs6000/vsx.md:
20126 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20129 2018-02-16 Carl Love <cel@us.ibm.com>
20131 * config/rs6000/altivec.h: Add builtin names vec_extract4b
20133 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20135 * config/rs6000/rs6000-c.c: Add the definitions for
20136 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20137 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20138 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20139 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
20140 definition for insert4b and define insn *insert3b_internal.
20141 * doc/extend.texi: Add documentation for vec_extract4b.
20143 2018-02-16 Nathan Sidwell <nathan@acm.org>
20145 * doc/extend.texi (Backwards Compatibility): Mention friend
20146 injection. Note for-scope is deprecated.
20147 * doc/invoke.texi (-ffriend-injection): Deprecate.
20149 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
20151 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20152 that moved to I2, also allow destinations that are a paradoxical
20153 subreg (instead of a normal reg).
20155 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
20158 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20161 2018-02-16 Richard Biener <rguenther@suse.de>
20163 PR tree-optimization/84037
20164 PR tree-optimization/84016
20166 * config/i386/i386.c (ix86_builtin_vectorization_cost):
20167 Adjust vec_construct for the fact we need additional higher latency
20168 128bit inserts for AVX256 and AVX512 vector builds.
20169 (ix86_add_stmt_cost): Scale vector construction cost for
20172 2018-02-16 Richard Biener <rguenther@suse.de>
20174 PR tree-optimization/84417
20175 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20176 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20177 (non_rewritable_lvalue_p): Likewise, use poly-ints.
20179 2018-02-16 Martin Liska <mliska@suse.cz>
20182 * internal-fn.def (ASAN_CHECK): Set proper flags.
20183 (ASAN_MARK): Likewise.
20185 2018-02-16 Julia Koval <julia.koval@intel.com>
20187 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20188 from PTA_CANNONLAKE.
20190 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20193 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20194 Use ++iter rather than iter++ for std::list iterators.
20195 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
20196 defer deleting them until all nodes in the forest are processed. Do
20197 free even leaf nodes. Change to_process into auto_vec.
20200 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20201 * vec.h (vec_default_construct): Use memset instead of placement new
20202 if BROKEN_VALUE_INITIALIZATION is defined.
20203 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20204 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20207 PR rtl-optimization/83723
20208 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20209 * lra.c (lra_substitute_pseudo): Likewise. If true, use
20210 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
20212 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20214 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20216 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
20218 PR rtl-optimization/81443
20219 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20220 from inner REGs to paradoxical SUBREGs.
20222 2018-02-16 Richard Biener <rguenther@suse.de>
20224 PR tree-optimization/84399
20225 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20226 For operands we can analyze at their definition make sure we can
20227 analyze them at each use as well.
20229 2018-02-16 Richard Biener <rguenther@suse.de>
20231 PR tree-optimization/84190
20232 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20233 volatile accesses if the decl isn't volatile.
20235 2018-02-15 Jason Merrill <jason@redhat.com>
20237 PR c++/84314 - ICE with templates and fastcall attribute.
20238 * attribs.c (build_type_attribute_qual_variant): Don't clobber
20239 TYPE_CANONICAL on an existing type.
20241 2018-02-15 Jakub Jelinek <jakub@redhat.com>
20243 PR tree-optimization/84383
20244 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20245 dstoff nor call operand_equal_p if dstbase is NULL.
20247 PR tree-optimization/84334
20248 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20249 also a CONSTANT_CLASS_P, punt.
20251 2018-02-14 Jim Wilson <jimw@sifive.com>
20253 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20254 first SMALL_OPERAND check. New local min_second_step. Move assert
20255 to where locals are set. Add TARGET_RVC support.
20256 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20258 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
20260 * doc/invoke.texi: Correct -Wformat-overflow code sample.
20262 2018-02-14 Martin Sebor <msebor@redhat.com>
20264 PR tree-optimization/83698
20265 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20266 arrays constrain the offset range to their bounds.
20267 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20268 (builtin_access::overlap): Avoid setting the size of overlap if it's
20270 (maybe_diag_overlap): Also consider arrays when deciding what values
20271 of offsets to include in diagnostics.
20273 2018-02-14 Martin Sebor <msebor@redhat.com>
20276 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20277 that correspond to the kind of a declaration.
20279 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
20282 * config/pa/pa.md: Load address of PIC label using the linkage table
20283 if the label is nonlocal.
20285 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
20287 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20288 warning message if user requests -maltivec=be.
20289 * doc/invoke.texi: Document deprecation of -maltivec=be.
20291 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
20294 * config/rs6000/rs6000-c.c: Update definitions for
20295 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20296 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20298 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20301 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20302 add _get_ssp intrinsics. Remove argument from
20303 __builtin_ia32_rdssp[d|q].
20304 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20305 * config/i386/i386-builtin.def: Remove argument from
20306 __builtin_ia32_rdssp[d|q].
20307 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20308 ix86_expand_special_args_builtin for _rdssp[d|q].
20309 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20310 Clear register before usage.
20311 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20312 Add documentation for new _get_ssp and _inc_ssp intrinsics.
20314 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
20316 PR tree-optimization/84357
20317 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20318 operand 1 of an ARRAY_REF too.
20320 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
20323 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20324 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20326 (set_of_reg): New struct.
20327 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20328 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20329 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20331 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20332 Split into bitclr, bitset, bitinvert patterns if appropriate.
20333 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
20334 use rx_fuse_in_memory_bitop.
20335 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
20336 to named insn, correct maximum insn length.
20338 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
20341 * machmode.def: Define a complex mode for PARTIAL_INT.
20342 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
20344 * doc/rtl.texi: Document CSPImode.
20345 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
20347 (msp430_hard_regno_nregs_with_padding): Likewise.
20349 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
20352 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
20354 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
20356 PR rtl-optimization/84169
20357 * combine.c (try_combine): New variable split_i2i3. Set it to true if
20358 we generated a parallel as new i3 and we split that to new i2 and i3
20359 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
20360 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
20361 those to i2, not i1. Partially rewrite this scan code.
20363 2018-02-13 Jakub Jelinek <jakub@redhat.com>
20366 * stor-layout.c (place_field): For variable length fields, adjust
20367 offset_align afterwards not just based on the field's alignment,
20368 but also on the size.
20370 PR middle-end/84309
20371 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
20372 of exps and logs in the use_exp2 case.
20374 2018-02-13 Jeff Law <law@redhat.com>
20376 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
20377 entry for "vector".
20379 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
20382 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
20386 * common.opt (gas-loc-support, gas-locview-support): New.
20387 (ginline-points, ginternal-reset-location-views): New.
20388 * doc/invoke.texi: Document them. Use @itemx where intended.
20389 (gvariable-location-views): Adjust.
20390 * target.def (reset_location_view): New.
20391 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
20392 (TARGET_RESET_LOCATION_VIEW): New.
20393 * doc/tm.texi: Rebuilt.
20394 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
20395 (dwarf2out_default_as_locview_support): New.
20396 (output_asm_line_debug_info): Use option variables.
20397 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
20398 (output_loc_list): Likewise.
20399 (add_high_low_attributes): Check option variables.
20400 Don't output entry view attribute in strict mode.
20401 (gen_inlined_subroutine_die): Check option variables.
20402 (dwarf2out_inline_entry): Likewise.
20403 (init_sections_and_labels): Likewise.
20404 (dwarf2out_early_finish): Likewise.
20405 (maybe_reset_location_view): New, from...
20406 (dwarf2out_var_location): ... here. Call it.
20407 * debug.h (dwarf2out_default_as_loc_support): Declare.
20408 (dwarf2out_default_as_locview_support): Declare.
20409 * hooks.c (hook_int_rtx_insn_0): New.
20410 * hooks.h (hook_int_rtx_insn_0): Declare.
20411 * toplev.c (process_options): Take -gas-loc-support and
20412 -gas-locview-support from dwarf2out. Enable
20413 -gvariable-location-views by default only with locview
20414 assembler support. Enable -ginternal-reset-location-views by
20415 default only if the target defines the corresponding hook.
20416 Enable -ginline-points by default if location views are
20417 enabled; force it disabled if statement frontiers are
20419 * tree-inline.c (expand_call_inline): Check option variables.
20420 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
20422 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
20424 PR tree-optimization/84321
20425 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
20426 handling. Also check whether the anti-range contains any values
20427 that satisfy the mask; switch to a VR_RANGE if not.
20429 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
20432 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
20434 2018-02-13 Martin Jambor <mjambor@suse.cz>
20437 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
20438 of call statements, also set location of a load to a temporary.
20440 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
20442 * config/rl78/rl78.c (add_vector_labels): New function.
20443 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
20444 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
20445 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
20446 which checks that no arguments are passed.
20447 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
20448 * doc/extend.texi: Documentation for the new attribute.
20450 2018-02-13 Andreas Schwab <schwab@suse.de>
20452 * config/riscv/linux.h (CPP_SPEC): Define.
20454 2018-02-13 Jakub Jelinek <jakub@redhat.com>
20457 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
20458 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
20459 OPTION_MASK_ISA_AES as first argument to def_builtin_const
20460 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
20461 instead of OPTION_MASK_ISA_PCLMUL as first argument to
20462 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
20463 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
20464 temporarily for AES and PCLMUL builtins.
20466 PR tree-optimization/84339
20467 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
20468 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
20471 PR middle-end/84309
20472 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
20473 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
20474 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
20476 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
20478 * omp-simd-clone.h: New file.
20479 * omp-simd-clone.c: Include omp-simd-clone.h.
20480 (expand_simd_clones): No longer static.
20481 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
20482 cgraph.h and omp-simd-clone.h.
20483 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
20484 (vect_recog_widen_shift_pattern): Formatting fix.
20485 (vect_pattern_recog_1): Don't check optab for calls.
20488 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
20489 operands[2] into a REG before using gen_lowpart on it.
20491 2018-02-12 Jeff Law <law@redhat.com>
20494 * config/sh/sh.c (find_barrier): Consider a sibling call
20497 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
20498 successfully back substituting a reg.
20500 2018-02-12 Richard Biener <rguenther@suse.de>
20502 PR tree-optimization/84037
20503 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
20504 parameter, move visited init to caller.
20505 (vect_slp_analyze_operations): Separate cost from validity
20506 check, initialize visited once for all instances.
20507 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
20509 * tree-vect-stmts.c (vect_model_simple_cost): Make early
20511 (vect_model_promotion_demotion_cost): Likewise.
20512 (vectorizable_bswap): Guard cost modeling with !slp_node
20513 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
20515 (vectorizable_call): Likewise.
20516 (vectorizable_conversion): Likewise.
20517 (vectorizable_assignment): Likewise.
20518 (vectorizable_shift): Likewise.
20519 (vectorizable_operation): Likewise.
20520 (vectorizable_store): Likewise.
20521 (vectorizable_load): Likewise.
20522 (vectorizable_condition): Likewise.
20523 (vectorizable_comparison): Likewise.
20525 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
20528 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
20529 (ASAN_MARK): Fix fnspec to account for return value, change pointer
20530 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
20532 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20534 PR middle-end/83665
20535 * params.def (inline-min-speedup): Increase from 8 to 15.
20536 (max-inline-insns-auto): Decrease from 40 to 30.
20537 * ipa-split.c (consider_split): Add some buffer for function to
20538 be considered inlining candidate.
20539 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
20542 2018-02-12 Richard Biener <rguenther@suse.de>
20544 PR tree-optimization/84037
20545 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
20546 matched stmts if we cannot swap the non-matched ones.
20548 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
20550 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
20551 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
20552 _mm_maskz_scalef_round_ss): New intrinsics.
20553 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
20554 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
20555 __builtin_ia32_scalefss_round): Remove.
20556 (__builtin_ia32_scalefsd_mask_round,
20557 __builtin_ia32_scalefss_mask_round): New intrinsics.
20558 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
20559 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
20560 ((match_operand:VF_128 2 "<round_nimm_predicate>"
20561 "<round_constraint>")): Changed to ...
20562 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
20563 "<round_scalar_constraint>")): ... this.
20564 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
20565 %0, %1, %2<round_op3>}"): Changed to ...
20566 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
20567 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
20568 %2<round_scalar_mask_op3>}"): ... this.
20569 * config/i386/subst.md (round_scalar_nimm_predicate): New.
20571 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
20573 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
20574 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
20575 (_mm_maskz_sqrt_round_ss): New intrinsics.
20576 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
20577 (__builtin_ia32_sqrtsd_mask_round)
20578 (__builtin_ia32_sqrtss_mask_round): New builtins.
20579 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
20580 (__builtin_ia32_sqrtss_round): Remove.
20581 (__builtin_ia32_sqrtsd_mask_round)
20582 (__builtin_ia32_sqrtss_mask_round): New builtins.
20583 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
20584 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
20585 ((match_operand:VF_128 1 "vector_operand"
20586 "xBm,<round_constraint>")): Changed to ...
20587 ((match_operand:VF_128 1 "vector_operand"
20588 "xBm,<round_scalar_constraint>")): ... this.
20589 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
20590 %0, %2, %<iptr>1<round_op3>}): Changed to ...
20591 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
20592 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
20593 %<iptr>1<round_scalar_mask_op3>}): ... this.
20594 ((set_attr "prefix" "<round_prefix>")): Changed to ...
20595 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
20597 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
20600 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
20601 Cast vec_cmpeq result to correct type.
20602 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
20603 Cast vec_cmpgt result to correct type.
20605 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
20607 * final.c (final_scan_insn_1): Renamed from...
20608 (final_scan_insn): ... this. New wrapper, to recover
20609 seen from the outermost call in recursive ones.
20610 * config/sparc/sparc.c (output_return): Drop seen from call.
20611 (output_sibcall): Likewise.
20612 * config/visium/visium.c (output_branch): Likewise.
20614 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
20616 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
20619 2018-02-10 Alan Modra <amodra@gmail.com>
20622 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
20623 Specify LR as an input.
20625 2018-02-10 Jakub Jelinek <jakub@redhat.com>
20628 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
20629 remove_member_access_dummy_vars): New functions.
20630 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
20631 lower_omp_1, execute_lower_omp): Use them.
20633 PR rtl-optimization/84308
20634 * shrink-wrap.c (spread_components): Release todo vector.
20636 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
20638 PR rtl-optimization/57193
20639 * ira-color.c (struct allocno_color_data): Add member
20640 conflict_allocno_hard_prefs.
20641 (update_conflict_allocno_hard_prefs): New.
20642 (bucket_allocno_compare_func): Add a preference based on
20643 conflict_allocno_hard_prefs.
20644 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
20645 (color_allocnos): Remove a dead code. Initiate
20646 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
20648 2018-02-09 Jakub Jelinek <jakub@redhat.com>
20651 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
20652 constraint from =wa to wa. Avoid a subreg on the output operand,
20653 instead use a pseudo and subreg it in a move.
20654 (p9_xxbrd_<mode>): Changed to ...
20655 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
20656 (p9_xxbrd_v2df): New expander.
20657 (p9_xxbrw_<mode>): Changed to ...
20658 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
20659 (p9_xxbrw_v4sf): New expander.
20661 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
20663 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
20665 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
20668 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
20669 multiply in 32-bit mode.
20670 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
20671 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
20674 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
20676 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
20677 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
20678 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
20679 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
20681 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
20684 * dwarf2out.c (is_trivial_indirect_ref): New function.
20685 (dwarf2out_late_global_decl): Do not generate a location
20686 attribute for variables that have a non-trivial DECL_VALUE_EXPR
20687 and that are not defined in the current unit.
20689 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
20691 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
20692 instead of a libcall for UNORDERED.
20694 2018-02-09 Tamar Christina <tamar.christina@arm.com>
20697 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
20698 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
20700 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20703 * config/s390/s390.c (s390_set_current_function): Invoke
20704 s390_indirect_branch_settings also if fndecl didn't change.
20706 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20708 * config/rs6000/rs6000.md (blockage): Set length to zero.
20710 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
20712 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
20714 2018-02-09 Jakub Jelinek <jakub@redhat.com>
20717 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
20718 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
20722 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
20723 PARALLEL incoming that failed vt_get_decl_and_offset check.
20725 PR middle-end/84237
20726 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
20727 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
20729 (get_variable_section): For decls in named .bss* sections pass true as
20730 second argument to bss_initializer_p.
20732 2018-02-09 Marek Polacek <polacek@redhat.com>
20733 Jakub Jelinek <jakub@redhat.com>
20736 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
20737 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
20738 Sync some changes from cxx_fold_indirect_ref.
20740 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20742 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
20744 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
20745 (BLOCK_INLINE_ENTRY_LABEL): New.
20746 (dwarf2out_var_location): Disregard inline entry markers.
20747 (inline_entry_data): New struct.
20748 (inline_entry_data_hasher): New hashtable type.
20749 (inline_entry_data_hasher::hash): New.
20750 (inline_entry_data_hasher::equal): New.
20751 (inline_entry_data_table): New variable.
20752 (add_high_low_attributes): Add DW_AT_entry_pc and
20753 DW_AT_GNU_entry_view attributes if a pending entry is found
20754 in inline_entry_data_table. Add old entry_pc attribute only
20755 if debug nonbinding markers are disabled.
20756 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
20757 markers are enabled.
20758 (block_within_block_p, dwarf2out_inline_entry): New.
20759 (dwarf2out_finish): Check that no entries remained in
20760 inline_entry_data_table.
20761 * final.c (reemit_insn_block_notes): Handle inline entry notes.
20762 (final_scan_insn, notice_source_line): Likewise.
20763 (rest_of_clean_state): Skip inline entry markers.
20764 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
20766 * gimple.c (gimple_build_debug_inline_entry): New.
20767 * gimple.h (enum gimple_debug_subcode): Add
20768 GIMPLE_DEBUG_INLINE_ENTRY.
20769 (gimple_build_debug_inline_entry): Declare.
20770 (gimple_debug_inline_entry_p): New.
20771 (gimple_debug_nonbind_marker_p): Adjust.
20772 * insn-notes.def (INLINE_ENTRY): New.
20773 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
20774 inline entry marker notes.
20775 (print_insn): Likewise.
20776 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
20777 (INSN_DEBUG_MARKER_KIND): Likewise.
20778 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
20779 * tree-inline.c (expand_call_inline): Build and insert
20780 debug_inline_entry stmt.
20781 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
20782 inline entry blocks early, if nonbind markers are enabled.
20783 (dump_scope_block): Dump fragment info.
20784 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
20785 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
20786 (gimple_build_debug_inline_entry): New.
20787 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
20788 Enable/disable inline entry points too.
20789 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
20790 (DEBUG_INSN): Describe inline entry markers.
20792 * common.opt (gvariable-location-views): New.
20793 (gvariable-location-views=incompat5): New.
20794 * config.in: Rebuilt.
20795 * configure: Rebuilt.
20796 * configure.ac: Test assembler for view support.
20797 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
20798 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
20799 * dwarf2out.c (var_loc_view): New typedef.
20800 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
20801 (dwarf2out_locviews_in_attribute): New.
20802 (dwarf2out_locviews_in_loclist): New.
20803 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
20804 (enum dw_line_info_opcode): Add LI_adv_address.
20805 (struct dw_line_info_table): Add view.
20806 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
20807 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
20808 (zero_view_p): New variable.
20809 (ZERO_VIEW_P): New macro.
20810 (output_asm_line_debug_info): New.
20811 (struct var_loc_node): Add view.
20812 (add_AT_view_list, AT_loc_list): New.
20813 (add_var_loc_to_decl): Add view param. Test it against last.
20814 (new_loc_list): Add view params. Record them.
20815 (AT_loc_list_ptr): Handle loc and view lists.
20816 (view_list_to_loc_list_val_node): New.
20817 (print_dw_val): Handle dw_val_class_view_list.
20818 (size_of_die): Likewise.
20819 (value_format): Likewise.
20820 (loc_list_has_views): New.
20821 (gen_llsym): Set vl_symbol too.
20822 (maybe_gen_llsym, skip_loc_list_entry): New.
20823 (dwarf2out_maybe_output_loclist_view_pair): New.
20824 (output_loc_list): Output view list or entries too.
20825 (output_view_list_offset): New.
20826 (output_die): Handle dw_val_class_view_list.
20827 (output_dwarf_version): New.
20828 (output_compilation_unit_header): Use it.
20829 (output_skeleton_debug_sections): Likewise.
20830 (output_rnglists, output_line_info): Likewise.
20831 (output_pubnames, output_aranges): Update version comments.
20832 (output_one_line_info_table): Output view numbers in asm comments.
20833 (dw_loc_list): Determine current endview, pass it to new_loc_list.
20834 Call maybe_gen_llsym.
20835 (loc_list_from_tree_1): Adjust.
20836 (add_AT_location_description): Create view list attribute if
20837 needed, check it's absent otherwise.
20838 (convert_cfa_to_fb_loc_list): Adjust.
20839 (maybe_emit_file): Call output_asm_line_debug_info for test.
20840 (dwarf2out_var_location): Reset views as needed. Precompute
20841 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
20842 attribute. Set view.
20843 (new_line_info_table): Reset next view.
20844 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
20845 (dwarf2out_source_line): Likewise. Output view resets and labels to
20846 the assembler, or select appropriate line info opcodes.
20847 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
20848 (optimize_string_length): Catch it. Adjust.
20849 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
20850 dw_val_class_view_list, and remove it if no longer needed.
20851 (hash_loc_list): Hash view numbers.
20852 (loc_list_hasher::equal): Compare them.
20853 (optimize_location_lists): Check whether a view list symbol is
20854 needed, and whether the locview attribute is present, and
20855 whether they match. Remove the locview attribute if no longer
20857 (index_location_lists): Call skip_loc_list_entry for test.
20858 (dwarf2out_finish): Call output_asm_line_debug_info for test.
20859 Use output_dwarf_version.
20860 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
20861 (struct dw_val_node): Add val_view_list.
20862 * final.c (SEEN_NEXT_VIEW): New.
20863 (set_next_view_needed): New.
20864 (clear_next_view_needed): New.
20865 (maybe_output_next_view): New.
20866 (final_start_function): Rename to...
20867 (final_start_function_1): ... this. Take pointer to FIRST,
20868 add SEEN parameter. Emit param bindings in the initial view.
20869 (final_start_function): Reintroduce SEEN-less interface.
20870 (final): Rename to...
20871 (final_1): ... this. Take SEEN parameter. Output final pending
20872 next view at the end.
20873 (final): Reintroduce seen-less interface.
20874 (final_scan_insn): Output pending next view before switching
20875 sections or ending a block. Mark the next view as needed when
20876 outputting variable locations. Notify debug backend of section
20877 changes, and of location view changes.
20878 (rest_of_handle_final): Adjust.
20879 * toplev.c (process_options): Autodetect value for debug variable
20880 location views option. Warn on incompat5 without -gdwarf-5.
20881 * doc/invoke.texi (gvariable-location-views): New.
20882 (gvariable-location-views=incompat5): New.
20883 (gno-variable-location-views): New.
20885 2018-02-08 David Malcolm <dmalcolm@redhat.com>
20887 PR tree-optimization/84136
20888 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
20889 that the result of find_edge is non-NULL.
20891 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20894 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
20895 storing integer register in SImode. Fix cost of 256 and 512
20896 byte aligned SSE register store.
20898 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20900 * config/i386/i386.c (ix86_multiplication_cost): Fix
20901 multiplication cost for TARGET_AVX512DQ.
20903 2018-02-08 Marek Polacek <polacek@redhat.com>
20905 PR tree-optimization/84238
20906 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
20909 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20911 PR tree-optimization/84265
20912 * tree-vect-stmts.c (vectorizable_store): Don't treat
20913 VMAT_CONTIGUOUS accesses as grouped.
20914 (vectorizable_load): Likewise.
20916 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20918 PR tree-optimization/81635
20919 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
20920 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
20921 (test_round_for_mask): New functions.
20922 (wide_int_cc_tests): Call test_round_for_mask.
20923 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
20924 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
20925 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
20926 range returned by get_range_info.
20928 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20931 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
20932 * symtab.c: Include builtins.h
20933 (symtab_node::output_to_lto_symbol_table_p): Move here
20934 from lto-streamer-out.c:output_symbol_p.
20935 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
20936 (output_symbol_p): Move all logic to symtab.c
20937 (produce_symtab): Update.
20939 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20941 * config/s390/s390-opts.h (enum indirect_branch): Define.
20942 * config/s390/s390-protos.h (s390_return_addr_from_memory)
20943 (s390_indirect_branch_via_thunk)
20944 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
20945 (enum s390_indirect_branch_type): Define.
20946 * config/s390/s390.c (struct s390_frame_layout, struct
20947 machine_function): Remove.
20948 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
20949 (indirect_branch_table_label_no, indirect_branch_table_name):
20951 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
20952 (enum s390_indirect_branch_option): Define.
20953 (s390_return_addr_from_memory): New function.
20954 (s390_handle_string_attribute): New function.
20955 (s390_attribute_table): Add new attribute handler.
20956 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
20957 (s390_indirect_branch_via_thunk): New function.
20958 (s390_indirect_branch_via_inline_thunk): New function.
20959 (s390_function_ok_for_sibcall): When jumping via thunk disallow
20960 sibling call optimization for non z10 compiles.
20961 (s390_emit_call): Force indirect branch target to be a single
20962 register. Add r1 clobber for non-z10 compiles.
20963 (s390_emit_epilogue): Emit return jump via return_use expander.
20964 (s390_reorg): Handle JUMP_INSNs as execute targets.
20965 (s390_option_override_internal): Perform validity checks for the
20966 new command line options.
20967 (s390_indirect_branch_attrvalue): New function.
20968 (s390_indirect_branch_settings): New function.
20969 (s390_set_current_function): Invoke s390_indirect_branch_settings.
20970 (s390_output_indirect_thunk_function): New function.
20971 (s390_code_end): Implement target hook.
20972 (s390_case_values_threshold): Implement target hook.
20973 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
20975 * config/s390/s390.h (struct s390_frame_layout)
20976 (struct machine_function): Move here from s390.c.
20977 (TARGET_INDIRECT_BRANCH_NOBP_RET)
20978 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
20979 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
20980 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
20981 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
20982 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
20983 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
20984 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
20985 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
20986 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
20987 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
20988 (mnemonic attribute): Add values which aren't recognized
20990 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
20991 pattern for branch conversion. Fix mnemonic attribute.
20992 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
20993 indirect branch via thunk if requested.
20994 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
20995 ("*indirect_jump"): Disable for branch conversion using out of
20997 ("indirect_jump_via_thunk<mode>_z10")
20998 ("indirect_jump_via_thunk<mode>")
20999 ("indirect_jump_via_inlinethunk<mode>_z10")
21000 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21001 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21002 ("casesi_jump_via_inlinethunk<mode>_z10")
21003 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21004 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21005 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21006 ("*indirect2_jump"): Disable for branch conversion.
21007 ("casesi_jump"): Turn into expander and expand patterns for branch
21009 ("return_use"): New expander.
21010 ("*return"): Emit return via thunk and rename it to ...
21011 ("*return<mode>"): ... this one.
21012 * config/s390/s390.opt: Add new options and and enum for the
21015 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21017 * lra-constraints.c (match_reload): Unconditionally use
21018 gen_lowpart_SUBREG, rather than selecting between that
21019 and equivalent gen_rtx_SUBREG code.
21021 2018-02-08 Richard Biener <rguenther@suse.de>
21023 PR tree-optimization/84233
21024 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21025 changed flag instead of boguously re-using phi_inserted.
21027 2018-02-08 Martin Jambor <mjambor@suse.cz>
21029 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21030 static local variables.
21032 2018-02-08 Richard Biener <rguenther@suse.de>
21034 PR tree-optimization/84278
21035 * tree-vect-stmts.c (vectorizable_store): When looking for
21036 smaller vector types to perform grouped strided loads/stores
21037 make sure the mode is supported by the target.
21038 (vectorizable_load): Likewise.
21040 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21042 * config/aarch64/aarch64.c (aarch64_components_for_bb):
21043 Increase LDP/STP opportunities by adding adjacent callee-saves.
21045 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21047 PR rtl-optimization/84068
21048 PR rtl-optimization/83459
21049 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21051 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
21053 PR tree-optimization/84224
21054 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21055 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21056 non-zero arguments.
21058 2018-02-07 Iain Sandoe <iain@codesourcery.com>
21061 * config/rs6000/altivec.md (*restore_world): Remove LR use.
21062 * config/rs6000/predicates.md (restore_world_operation): Adjust op
21063 count, remove one USE.
21065 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
21067 * doc/install.texi (Configuration): Document the
21068 --with-long-double-format={ibm,ieee} PowerPC configuration
21072 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21073 Convert from define_expand to be define_insn_and_split. Rework
21074 float/double/_Float128 conversions to QI/HI/SImode to work with
21075 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
21076 conversions to QI/HImode types did a store and then a load to
21077 truncate the value. For conversions to VSX registers, don't split
21078 the insn, instead emit the code directly. Use the code iterator
21079 any_fix to combine signed and unsigned conversions.
21080 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21081 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21082 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21083 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21084 (fix_<mode>di2_hw): Likewise.
21085 (fixuns_<mode>di2_hw): Likewise.
21086 (fix_<mode>si2_hw): Likewise.
21087 (fixuns_<mode>si2_hw): Likewise.
21088 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21089 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21090 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21091 fix<uns>_trunc<SFDF:mode>si2_p8.
21092 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21093 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21094 (fix<uns>_<mode>_mem): Likewise.
21095 (fctiw<u>z_<mode>_mem): Likewise.
21096 (fix<uns>_<mode>_mem): Likewise.
21097 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21098 the register allocator from doing a direct move to the GPRs to do
21099 a store, and instead use the ISA 3.0 store byte/half-word from
21100 vector register instruction. For IEEE 128-bit floating point,
21101 also optimize stores of 32-bit ints.
21102 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21104 2018-02-07 Alan Hayward <alan.hayward@arm.com>
21106 * genextract.c (push_pathstr_operand): New function to support
21108 (walk_rtx): Call push_pathstr_operand.
21109 (print_path): Support [a-zA-Z].
21111 2018-02-07 Richard Biener <rguenther@suse.de>
21113 PR tree-optimization/84037
21114 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21115 (cse_and_gimplify_to_preheader): Declare.
21116 (vect_get_place_in_interleaving_chain): Likewise.
21117 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21119 (_loop_vec_info::~_loop_vec_info): Delete it.
21120 (cse_and_gimplify_to_preheader): New function.
21121 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21122 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21123 (vectorizable_load): Likewise. For grouped stores always base
21124 the IV on the first element.
21125 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21126 condition before gimplifying.
21128 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21130 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21131 *DIV_EXPR and *MOD_EXPR.
21133 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
21136 * config/i386/i386.c (ix86_option_override_internal): Mask out
21137 the CF_SET bit when checking -fcf-protection.
21139 2018-02-07 Tom de Vries <tom@codesourcery.com>
21142 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21145 2018-02-07 Richard Biener <rguenther@suse.de>
21147 PR tree-optimization/84204
21148 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21151 PR tree-optimization/84205
21152 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21153 special-case isl_ast_op_zdiv_r.
21155 PR tree-optimization/84223
21156 * graphite-scop-detection.c (gather_bbs::before_dom_children):
21157 Only add conditions from within the region.
21158 (gather_bbs::after_dom_children): Adjust.
21160 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
21163 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21164 * config/avr/avr.md: Only post-reload split REG-REG moves if
21165 either register is GENERAL_REG_P.
21167 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21169 PR tree-optimization/84235
21170 * tree-ssa-scopedtables.c
21171 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21172 if the subtraction is performed in floating point type where NaNs are
21173 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21174 build 1. Formatting fix.
21176 2018-02-06 Jakub Jelinek <jakub@redhat.com>
21179 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21180 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21181 and skip it regardless of bb boundaries. Use CALL_P macro,
21182 don't test INSN_P (insn) together with CALL_P or JUMP_P check
21183 unnecessarily, formatting fix.
21185 2018-02-06 Michael Collison <michael.collison@arm.com>
21187 * config/arm/thumb2.md:
21188 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21189 (*thumb_mov_notscc): Ditto.
21191 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
21194 * config/rs6000/rs6000.md (su code attribute): Use "u" for
21195 unsigned_fix, not "s".
21197 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21199 * configure.ac (gcc_fn_eh_frame_ro): New function.
21200 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21201 correct .eh_frame permissions.
21202 * configure: Regenerate.
21204 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
21206 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21207 irrelevant options.
21209 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21211 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21212 Display warning message for -mno-speculate-indirect-jumps.
21214 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
21216 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21218 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21220 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
21222 PR tree-optimization/84225
21223 * tree-eh.c (find_trapping_overflow): Only call
21224 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21226 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21229 * config/i386/i386.c: Reimplement the check of possible options
21230 -mibt/-mshstk conbination. Change error messages.
21231 * doc/invoke.texi: Fix a typo: remove extra '='.
21233 2018-02-06 Marek Polacek <polacek@redhat.com>
21235 PR tree-optimization/84228
21236 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21238 2018-02-06 Tamar Christina <tamar.christina@arm.com>
21241 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21242 emitted arch directives.
21243 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21244 __ARM_FEATURE_COPROC before changing architectures.
21246 2018-02-06 Richard Biener <rguenther@suse.de>
21248 * config/i386/i386.c (print_reg): Fix typo.
21249 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21251 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
21253 * configure: Regenerate.
21255 2018-02-05 Martin Sebor <msebor@redhat.com>
21257 PR tree-optimization/83369
21258 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21261 2018-02-05 Martin Liska <mliska@suse.cz>
21263 * doc/invoke.texi: Cherry-pick upstream r323995.
21265 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
21267 * ira.c (ira_init_register_move_cost): Adjust comment.
21269 2018-02-05 Martin Liska <mliska@suse.cz>
21271 PR gcov-profile/84137
21272 * doc/gcov.texi: Fix typo in documentation.
21274 2018-02-05 Martin Liska <mliska@suse.cz>
21276 PR gcov-profile/83879
21277 * doc/gcov.texi: Document necessity of --dynamic-list-data when
21278 using dlopen functionality.
21280 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
21282 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21283 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21284 _mm_maskz_range_ss, _mm_mask_range_round_ss,
21285 _mm_maskz_range_round_ss): New intrinsics.
21286 (__builtin_ia32_rangesd128_round)
21287 (__builtin_ia32_rangess128_round): Remove.
21288 (__builtin_ia32_rangesd128_mask_round,
21289 __builtin_ia32_rangess128_mask_round): New builtins.
21290 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21291 __builtin_ia32_rangess128_round): Remove.
21292 (__builtin_ia32_rangesd128_mask_round,
21293 __builtin_ia32_rangess128_mask_round): New builtins.
21294 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21295 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21296 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21297 "<round_saeonly_constraint>")): Changed to ...
21298 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21299 "<round_saeonly_scalar_constraint>")): ... this.
21300 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21301 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21302 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21303 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21304 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21306 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
21308 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21310 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21311 Remove all values except native, 8540 and 8548.
21313 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
21315 * config/i386/i386.c (ix86_output_function_return): Pass
21316 INVALID_REGNUM, instead of -1, as invalid register number to
21317 indirect_thunk_name and output_indirect_thunk.
21319 2018-02-02 Julia Koval <julia.koval@intel.com>
21321 * config.gcc: Add -march=icelake.
21322 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21323 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21324 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21325 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21326 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21327 (processor_target_table): Add icelake.
21328 (ix86_option_override_internal): Handle new PTAs.
21329 (get_builtin_code_for_version): Handle icelake.
21330 (M_INTEL_COREI7_ICELAKE): New.
21331 (fold_builtin_cpu): Handle icelake.
21332 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21333 * doc/invoke.texi: Add -march=icelake.
21335 2018-02-02 Julia Koval <julia.koval@intel.com>
21337 * config/i386/i386.c (ix86_option_override_internal): Change flags type
21338 to wide_int_bitmask.
21339 * wide-int-bitmask.h: New.
21341 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21344 * config/i386/i386.md: Replace Pmode with word_mode in
21345 builtin_setjmp_setup and builtin_longjmp to support x32.
21347 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
21351 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
21353 (rs6000_supported_cpu_names): New static variable.
21354 (linux_cpu_translation_table): Likewise.
21355 (elf_platform) <cpu>: Define new static variable and use it.
21356 Translate kernel AT_PLATFORM name to canonical name if needed.
21357 Error if platform name is unknown.
21359 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
21362 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
21364 2018-02-01 Jeff Law <law@redhat.com>
21367 * config/i386/i386.c (release_scratch_register_on_entry): Add new
21368 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
21369 the scratch if RELEASE_VIA_POP is false.
21370 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
21371 If we have to save a temporary register, decrement SIZE appropriately.
21372 Pass new arguments to release_scratch_register_on_entry.
21373 (ix86_adjust_stack_and_probe): Likewise.
21374 (ix86_emit_probe_stack_range): Pass new arguments to
21375 release_scratch_register_on_entry.
21377 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
21379 PR rtl-optimization/84157
21380 * combine.c (change_zero_ext): Use REG_P predicate in
21381 front of HARD_REGISTER_P predicate.
21383 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
21385 * config/avr/avr.c (avr_option_override): Move disabling of
21386 -fdelete-null-pointer-checks to...
21387 * common/config/avr/avr-common.c (avr_option_optimization_table):
21390 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21392 PR tree-optimization/81635
21393 * tree-data-ref.c (split_constant_offset_1): For types that
21394 wrap on overflow, try to use range info to prove that wrapping
21397 2018-02-01 Renlin Li <renlin.li@arm.com>
21400 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
21401 TAILCALL_ADDR_REGS.
21402 (aarch64_register_move_cost): Likewise.
21403 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
21404 TAILCALL_ADDR_REGS.
21405 (REG_CLASS_NAMES): Likewise.
21406 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
21407 TAILCALL_ADDR_REGS. Remove IP registers.
21408 * config/aarch64/aarch64.md (Ucs): Update register constraint.
21410 2018-02-01 Richard Biener <rguenther@suse.de>
21412 * domwalk.h (dom_walker::dom_walker): Add additional constructor
21413 for specifying RPO order and allow NULL for that.
21414 * domwalk.c (dom_walker::dom_walker): Likewise.
21415 (dom_walker::walk): Handle NULL RPO order.
21416 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
21418 (rewrite_update_dom_walker): Likewise.
21419 (mark_def_dom_walker): Likewise.
21421 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21423 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
21424 (aarch64_maybe_expand_sve_subreg_move): Declare.
21425 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
21426 * config/aarch64/predicates.md (aarch64_any_register_operand): New
21428 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
21429 that are semantically a reverse operation.
21430 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
21431 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
21432 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
21434 (aarch64_can_change_mode_class): For big-endian, forbid changes
21435 between two SVE modes if they have different element sizes.
21437 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21439 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
21440 the TImode handling for big-endian targets.
21442 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21444 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
21445 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
21447 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
21448 Remove BSWAP handing for big-endian targets and use the form of
21449 LD1RQ appropariate for the mode.
21451 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21453 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
21454 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
21455 duplicated element.
21457 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21460 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
21461 check for operands that need to go through aarch64_sve_reload_be.
21463 2018-02-01 Jakub Jelinek <jakub@redhat.com>
21465 PR tree-optimization/81661
21466 PR tree-optimization/84117
21467 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
21468 * tree-eh.c: Include gimplify.h.
21469 (find_trapping_overflow, replace_trapping_overflow,
21470 rewrite_to_non_trapping_overflow): New functions.
21471 * tree-vect-loop.c: Include tree-eh.h.
21472 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
21473 * tree-data-ref.c: Include tree-eh.h.
21474 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
21476 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
21478 PR rtl-optimization/84123
21479 * combine.c (change_zero_ext): Check if hard register satisfies
21480 can_change_dest_mode before calling gen_lowpart_SUBREG.
21482 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
21485 * ira.c (ira_init_register_move_cost): Remove assert.
21487 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21489 PR rtl-optimization/84071
21490 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
21491 * doc/tm.texi: Regenerate.
21493 2018-01-31 Richard Biener <rguenther@suse.de>
21495 PR tree-optimization/84132
21496 * tree-data-ref.c (analyze_miv_subscript): Properly
21497 check whether evolution_function_is_affine_multivariate_p
21498 before calling gcd_of_steps_may_divide_p.
21500 2018-01-31 Julia Koval <julia.koval@intel.com>
21503 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
21504 * config/i386/i386.md (rdpid_rex64) New.
21505 (rdpid): Make 32bit only.
21507 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
21510 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
21511 an IDENTIFIER_NODE for FUNCTION_TYPE's.
21513 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21516 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21518 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21520 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21522 PR rtl-optimization/84071
21523 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
21524 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
21526 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
21528 * config/arc/arc.c (arc_handle_aux_attribute): New function.
21529 (arc_attribute_table): Add 'aux' attribute.
21530 (arc_in_small_data_p): Consider aux like variables.
21531 (arc_is_aux_reg_p): New function.
21532 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
21533 (arc_get_aux_arg): New function.
21534 (prepare_move_operands): Handle aux-register access.
21535 (arc_handle_aux_attribute): New function.
21536 * doc/extend.texi (ARC Variable attributes): Add subsection.
21538 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
21540 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
21541 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
21542 (arc_attribute_table): Add 'uncached' attribute.
21543 (arc_print_operand): Print '.di' flag for uncached memory
21545 (arc_in_small_data_p): Do not consider for small data the uncached
21547 (arc_is_uncached_mem_p): New function.
21548 * config/arc/predicates.md (compact_store_memory_operand): Check
21549 for uncached memory accesses.
21550 (nonvol_nonimm_operand): Likewise.
21551 * doc/extend.texi (ARC Type Attribute): New subsection.
21553 2018-01-31 Jakub Jelinek <jakub@redhat.com>
21556 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
21557 falign-loops=): Add Optimization flag.
21559 2018-01-30 Jeff Law <law@redhat.com>
21562 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
21563 INT_REGISTERS_SAVED. Check it prior to calling
21564 get_scratch_register_on_entry.
21565 (ix86_adjust_stack_and_probe): Similarly.
21566 (ix86_emit_probe_stack_range): Similarly.
21567 (ix86_expand_prologue): Corresponding changes.
21569 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21572 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
21573 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
21575 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
21578 * lra-constraints.c (curr_insn_transform): Process AND in the
21581 2018-01-30 Jakub Jelinek <jakub@redhat.com>
21583 PR rtl-optimization/83986
21584 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
21585 dependence against last_pending_memory_flush in addition to
21586 pending_jump_insns.
21588 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
21590 PR tree-optimization/81611
21591 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
21594 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21597 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
21600 2018-01-30 Richard Biener <rguenther@suse.de>
21601 Jakub Jelinek <jakub@redhat.com>
21603 PR tree-optimization/84111
21604 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
21605 inner loops added during recursion, as they don't have up-to-date
21608 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
21611 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
21612 (can_inline_edge_by_limits_p): ... here.
21613 (can_early_inline_edge_p, check_callers,
21614 update_caller_keys, update_callee_keys, recursive_inlining,
21615 add_new_edges_to_heap, speculation_useful_p,
21616 inline_small_functions,
21617 inline_small_functions, flatten_function,
21618 inline_to_all_callers_1): Update.
21620 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
21622 * profile-count.c (profile_count::combine_with_ipa_count): Handle
21625 2018-01-30 Richard Biener <rguenther@suse.de>
21627 PR tree-optimization/83008
21628 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
21629 invariant and constant vector uses in stmts when they need
21630 more than one stmt.
21632 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21635 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
21636 * configure: Regenerate.
21638 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21640 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
21642 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
21643 Use gen_rtx_REG rather than gen_lowpart.
21645 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21647 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
21648 rather than 0 when creating partial subregs.
21650 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21652 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
21655 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
21658 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
21659 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
21660 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
21661 flags. This restores the settings used before the 2017-07-24.
21662 Turning off pre increment/decrement/modify allows IVOPTS to
21663 optimize DF/SF loops where the index is an int.
21665 2018-01-29 Richard Biener <rguenther@suse.de>
21666 Kelvin Nilsen <kelvin@gcc.gnu.org>
21669 * tree-vect-stmts.c (vectorizable_call): Don't call
21670 targetm.vectorize_builtin_md_vectorized_function if callee is
21673 2018-01-22 Carl Love <cel@us.ibm.com>
21675 * doc/extend.tex: Fix typo in second arg in
21676 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
21678 2018-01-29 Richard Biener <rguenther@suse.de>
21680 PR tree-optimization/84086
21681 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
21682 (flush_ssaname_freelist): When SSA names were released reset
21683 the SCEV hash table.
21685 2018-01-29 Richard Biener <rguenther@suse.de>
21687 PR tree-optimization/84057
21688 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
21689 removed paths when removing edges.
21691 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
21693 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
21694 -mfunction-return=@var{choice}.
21696 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
21698 PR diagnostic/84034
21699 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
21700 Handle CR like TAB.
21701 (layout::print_source_line): Likewise.
21702 (test_get_line_width_without_trailing_whitespace): Add test cases.
21704 2018-01-27 Jakub Jelinek <jakub@redhat.com>
21706 PR middle-end/84040
21707 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
21710 2018-01-26 Jim Wilson <jimw@sifive.com>
21712 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
21714 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
21717 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21719 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
21720 and CMP + SUB-immediate -> SUBS.
21722 2018-01-26 Martin Sebor <msebor@redhat.com>
21724 PR tree-optimization/83896
21725 * tree-ssa-strlen.c (get_string_len): Rename...
21726 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
21727 Avoid assuming length is constant.
21728 (handle_char_store): Use HOST_WIDE_INT for string length.
21730 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
21733 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
21734 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
21736 2018-01-26 Richard Biener <rguenther@suse.de>
21738 PR rtl-optimization/84003
21739 * dse.c (record_store): Only record redundant stores when
21740 the earlier store aliases at least all accesses the later one does.
21742 2018-01-26 Jakub Jelinek <jakub@redhat.com>
21744 PR rtl-optimization/83985
21745 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
21746 REG_CFA_RESTORE insns.
21747 (delete_unmarked_insns): Don't ignore separate shrink wrapping
21748 REG_CFA_RESTORE insns here.
21751 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
21752 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
21754 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21756 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
21757 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
21758 (arc_init): Likewise.
21759 (arc_override_options): Likewise.
21760 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
21762 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
21764 * config/arc/arc.h (TARGET_DBNZ): Define.
21765 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
21766 properly set the tune attribute.
21767 (dbnz): Use TARGET_DBNZ guard.
21768 * config/arc/arc.opt (mtune): Add core3 option.
21770 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21772 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
21773 recognize new pic like addresses.
21774 (arc_delegitimize_address): Clean up.
21776 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21778 * config/arc/arc-arches.def: Option mrf16 valid for all
21780 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
21781 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
21782 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
21783 * config/arc/arc-tables.opt: Regenerate.
21784 * config/arc/arc.c (arc_conditional_register_usage): Handle
21785 reduced register file case.
21786 (arc_file_start): Set must have build attributes.
21787 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
21788 mrf16 option value.
21789 * config/arc/arc.opt (mrf16): Add new option.
21790 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
21791 * config/arc/genmultilib.awk: Handle new mrf16 option.
21792 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
21793 * config/arc/t-multilib: Regenerate.
21794 * doc/invoke.texi (ARC Options): Document mrf16 option.
21796 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21798 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
21799 * config/arc/arc.c (arc_handle_secure_attribute): New function.
21800 (arc_attribute_table): Add 'secure_call' attribute.
21801 (arc_print_operand): Print secure call operand.
21802 (arc_function_ok_for_sibcall): Don't optimize tail calls when
21804 (arc_is_secure_call_p): New function. * config/arc/arc.md
21805 (call_i): Add support for sjli instruction.
21806 (call_value_i): Likewise.
21807 * config/arc/constraints.md (Csc): New constraint.
21809 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21810 John Eric Martin <John.Martin@emmicro-us.com>
21812 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
21813 * config/arc/arc.c (_arc_jli_section): New struct.
21814 (arc_jli_section): New type.
21815 (rc_jli_sections): New static variable.
21816 (arc_handle_jli_attribute): New function.
21817 (arc_attribute_table): Add jli_always and jli_fixed attribute.
21818 (arc_file_end): New function.
21819 (TARGET_ASM_FILE_END): Define.
21820 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
21821 (arc_add_jli_section): New function.
21822 (jli_call_scan): Likewise.
21823 (arc_reorg): Call jli_call_scan.
21824 (arc_output_addsi): Remove 'S' from printing asm operand.
21825 (arc_is_jli_call_p): New function.
21826 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
21828 (movhi_insn): Likewise.
21829 (movsi_insn): Likewise.
21830 (movsi_set_cc_insn): Likewise.
21831 (loadqi_update): Likewise.
21832 (load_zeroextendqisi_update): Likewise.
21833 (load_signextendqisi_update): Likewise.
21834 (loadhi_update): Likewise.
21835 (load_zeroextendhisi_update): Likewise.
21836 (load_signextendhisi_update): Likewise.
21837 (loadsi_update): Likewise.
21838 (loadsf_update): Likewise.
21839 (movsicc_insn): Likewise.
21840 (bset_insn): Likewise.
21841 (bxor_insn): Likewise.
21842 (bclr_insn): Likewise.
21843 (bmsk_insn): Likewise.
21844 (bicsi3_insn): Likewise.
21845 (cmpsi_cc_c_insn): Likewise.
21846 (movsi_ne): Likewise.
21847 (movsi_cond_exec): Likewise.
21848 (clrsbsi2): Likewise.
21849 (norm_f): Likewise.
21859 (call_i): Remove 'S' asm letter, add jli instruction.
21860 (call_value_i): Likewise.
21861 * config/arc/arc.op (mjli-always): New option.
21862 * config/arc/constraints.md (Cji): New constraint.
21863 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
21865 (subsf3_fpx): Likewise.
21866 (mulsf3_fpx): Likewise.
21867 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
21869 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
21870 function attrbutes.
21871 * doc/invoke.texi (ARC): Document mjli-always option.
21873 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
21875 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
21876 avoid addition with 0 and use incw and decw where possible.
21878 2018-01-26 Richard Biener <rguenther@suse.de>
21880 PR tree-optimization/81082
21881 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
21882 association if it requires casting to unsigned.
21883 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
21884 from fold_plusminus_mult_expr to catch important cases late when
21885 range info is available.
21887 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21889 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
21890 * configure.ac (hidden_linkonce): New test.
21891 * configure: Regenerate.
21892 * config.in: Regenerate.
21894 2018-01-26 Julia Koval <julia.koval@intel.com>
21896 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
21897 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
21898 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
21899 _mm_mask_bitshuffle_epi64_mask): Fix type.
21900 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
21901 USI_FTYPE_V4DI_V4DI_USI): Remove.
21902 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
21903 __builtin_ia32_vpshufbitqmb256_mask,
21904 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
21905 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
21906 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
21908 2018-01-26 Alan Modra <amodra@gmail.com>
21911 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
21912 UNSPEC_VBPERMQ. Sort other unspecs.
21914 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
21916 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
21918 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
21920 PR middle-end/83055
21921 * predict.c (drop_profile): Do not push/pop cfun; update also
21923 (handle_missing_profiles): Fix logic looking for zero profiles.
21925 2018-01-25 Jakub Jelinek <jakub@redhat.com>
21927 PR middle-end/83977
21928 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
21929 on functions with #pragma omp declare simd or functions with simd
21931 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
21932 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
21933 Remove trailing \n from warning_at calls.
21935 2018-01-25 Tom de Vries <tom@codesourcery.com>
21938 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21939 for neutered workers.
21941 2018-01-24 Joseph Myers <joseph@codesourcery.com>
21944 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
21945 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
21947 2018-01-24 Jeff Law <law@redhat.com>
21950 * i386.c (get_probe_interval): Move to earlier point.
21951 (ix86_compute_frame_layout): If -fstack-clash-protection and
21952 the frame is larger than the probe interval, then use pushes
21953 to save registers rather than reg->mem moves.
21954 (ix86_expand_prologue): Remove conditional for int_registers_saved
21957 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
21960 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
21961 min/max for never referenced object.
21963 2018-01-24 Jakub Jelinek <jakub@redhat.com>
21965 PR middle-end/83977
21966 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
21968 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
21969 attributes from DECL_ATTRIBUTES (decl) without affecting
21970 DECL_ATTRIBUTES (current_function_decl).
21971 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
21972 functions with non-NULL DECL_ABSTRACT_ORIGIN.
21974 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
21976 PR tree-optimization/83979
21977 * fold-const.c (fold_comparison): Use constant_boolean_node
21978 instead of boolean_{true,false}_node.
21980 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
21982 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
21985 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21987 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21988 Simplify the clause that sets the length attribute.
21989 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21990 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
21991 clause that sets the length attribute.
21992 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21994 2018-01-24 Tom de Vries <tom@codesourcery.com>
21997 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
21998 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
21999 Add strict parameter.
22000 (prevent_branch_around_nothing): Insert dummy insn between branch to
22001 label and label with no ptx insn inbetween.
22002 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22004 2018-01-24 Tom de Vries <tom@codesourcery.com>
22007 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22008 for neutered threads in warp.
22009 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22011 2018-01-24 Richard Biener <rguenther@suse.de>
22013 PR tree-optimization/83176
22014 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22017 2018-01-24 Richard Biener <rguenther@suse.de>
22019 PR tree-optimization/82819
22020 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22021 code generating pluses that are no-ops in the target precision.
22023 2018-01-24 Richard Biener <rguenther@suse.de>
22025 PR middle-end/84000
22026 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22028 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22030 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22031 to merge probabilities.
22032 * predict.c (probably_never_executed): Also mark as cold functions
22033 with global 0 profile and guessed local profile.
22034 * profile-count.c (profile_probability::combine_with_count): New
22036 * profile-count.h (profile_probability::operator*,
22037 profile_probability::operator*=, profile_probability::operator/,
22038 profile_probability::operator/=): Reduce precision to adjusted
22039 and set value to guessed on contradictory divisions.
22040 (profile_probability::combine_with_freq): Remove.
22041 (profile_probability::combine_wiht_count): Declare.
22042 (profile_count::force_nonzero):: Set to adjusted.
22043 (profile_count::probability_in):: Set quality to adjusted.
22044 * tree-ssa-tail-merge.c (replace_block_by): Use
22045 combine_with_count.
22047 2018-01-23 Andrew Waterman <andrew@sifive.com>
22048 Jim Wilson <jimw@sifive.com>
22050 * config/riscv/riscv.c (riscv_stack_boundary): New.
22051 (riscv_option_override): Set riscv_stack_boundary. Handle
22052 riscv_preferred_stack_boundary_arg.
22053 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22054 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22055 (STACK_BOUNDARY): Set to riscv_stack_boundary.
22056 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22057 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22058 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22060 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
22063 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22064 of struct ix86_frame.
22065 (ix86_expand_epilogue): Likewise. Add a local variable for
22066 the reg_save_offset field in struct ix86_frame.
22068 2018-01-23 Bin Cheng <bin.cheng@arm.com>
22070 PR tree-optimization/82604
22071 * tree-loop-distribution.c (enum partition_kind): New enum item
22072 PKIND_PARTIAL_MEMSET.
22073 (partition_builtin_p): Support above new enum item.
22074 (generate_code_for_partition): Ditto.
22075 (compute_access_range): Differentiate cases that equality can be
22076 proven at all loops, the innermost loops or no loops.
22077 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22078 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
22079 (finalize_partitions, distribute_loop): Don't fuse partition of
22080 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22081 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22082 parloop is enabled.
22084 2018-01-23 Martin Liska <mliska@suse.cz>
22086 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22087 order to ignore the predictor.
22088 (PRED_POLYMORPHIC_CALL): Likewise.
22089 (PRED_RECURSIVE_CALL): Likewise.
22091 2018-01-23 Martin Liska <mliska@suse.cz>
22093 * tree-profile.c (tree_profiling): Print function header to
22094 aware reader which function we are working on.
22095 * value-prof.c (gimple_find_values_to_profile): Do not print
22096 not interesting value histograms.
22098 2018-01-23 Martin Liska <mliska@suse.cz>
22100 * profile-count.h (enum profile_quality): Add
22101 profile_uninitialized as the first value. Do not number values
22102 as they are zero based.
22103 (profile_count::verify): Update sanity check.
22104 (profile_probability::verify): Likewise.
22106 2018-01-23 Nathan Sidwell <nathan@acm.org>
22108 * doc/invoke.texi (ffor-scope): Deprecate.
22110 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22112 PR tree-optimization/83510
22113 * domwalk.c (set_all_edges_as_executable): New function.
22114 (dom_walker::dom_walker): Convert bool param
22115 "skip_unreachable_blocks" to enum reachability. Move setup of
22116 edge flags to set_all_edges_as_executable and only do it when
22117 reachability is REACHABLE_BLOCKS.
22118 * domwalk.h (enum dom_walker::reachability): New enum.
22119 (dom_walker::dom_walker): Convert bool param
22120 "skip_unreachable_blocks" to enum reachability.
22121 (set_all_edges_as_executable): New decl.
22122 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
22123 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22125 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22126 but converting true to REACHABLE_BLOCKS.
22127 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22129 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22130 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22131 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22133 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22134 if check_all_array_refs will be called.
22136 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22138 * tree.c (selftest::test_location_wrappers): Add more test
22141 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22143 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22144 (selftest::test_bit_in_range): Likewise.
22146 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22149 * doc/sourcebuild.texi (vect_float): Say that the selector
22150 only describes the situation when -funsafe-math-optimizations is on.
22151 (vect_float_strict): Document.
22153 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22155 PR tree-optimization/83965
22156 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22157 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22158 instead of checking only for a reduction.
22159 (vect_recog_widen_sum_pattern): Likewise.
22161 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22163 * predict.c (probably_never_executed): Only use precise profile info.
22164 (compute_function_frequency): Skip after inlining hack since we now
22165 have quality checking.
22167 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22169 * profile-count.h (profile_probability::very_unlikely,
22170 profile_probability::unlikely, profile_probability::even): Set
22171 precision to guessed.
22173 2018-01-23 Richard Biener <rguenther@suse.de>
22175 PR tree-optimization/83963
22176 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22177 Properly terminate dominator walk when crossing the exit edge not
22178 when visiting its source block.
22180 2018-01-23 Jakub Jelinek <jakub@redhat.com>
22183 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22184 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22186 2018-01-22 Jakub Jelinek <jakub@redhat.com>
22188 PR tree-optimization/83957
22189 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
22190 semicolon after for body surrounded by braces.
22192 PR tree-optimization/83081
22193 * profile-count.h (profile_probability::split): New method.
22194 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22195 Use profile_probability::split.
22196 (do_compare_rtx_and_jump): Fix adjustment of probabilities
22197 when splitting a single conditional jump into 2.
22199 2018-01-22 David Malcolm <dmalcolm@redhat.com>
22201 PR tree-optimization/69452
22202 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22205 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22207 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22208 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22209 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22211 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22213 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22215 * config/rl78/rl78.md (movdi): New define_expand.
22216 * config/rl78/rl78.c (rl78_split_movdi): New function.
22218 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
22221 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22223 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22224 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22225 128-bit to produce an UNSPEC move to get the double word with the
22226 signbit and then a shift directly to do signbit.
22227 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22228 implementation with a new version that just does either a direct
22229 move or a regular move. Move memory interface to separate insns.
22230 Move insns so they are next to the expander.
22231 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22232 with signbit move. Split big and little endian case.
22233 (signbit<mode>2_dm_mem_le): Likewise.
22234 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22235 (signbit<mode>2_dm2): Likewise.
22237 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22239 * config/rl78/rl78.md (anddi3): New define_expand.
22241 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22243 * config/rl78/rl78.md (umindi3): New define_expand.
22245 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22247 * config/rl78/rl78.md (smindi3): New define_expand.
22249 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22251 * config/rl78/rl78.md (smaxdi3): New define_expand.
22253 2018-01-22 Carl Love <cel@us.ibm.com>
22255 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22256 LVX_V1TI): Add macro expansion.
22257 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22258 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22259 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22260 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22261 Change check to determine if the instruction is a byte reversing
22262 entry. Fix typo in comment.
22263 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22264 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22265 Add def_builtin calls for new builtins.
22266 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22267 Add define_insn expansion.
22269 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22271 * config/rl78/rl78.md (umaxdi3): New define_expand.
22273 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22275 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22276 for non-QImode registers.
22278 2018-01-22 Richard Biener <rguenther@suse.de>
22280 PR tree-optimization/83963
22281 * graphite-scop-detection.c (scop_detection::get_sese): Delay
22282 including the loop exit block.
22283 (scop_detection::merge_sese): Likewise.
22284 (scop_detection::add_scop): Do it here instead.
22286 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22288 * doc/sourcebuild.texi (arm_softfloat): Document.
22290 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
22293 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22294 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22295 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22297 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22298 David Edelsohn <dje.gcc@gmail.com>
22301 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22302 Change "crset eq" to "crset 2".
22303 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22304 (*call_indirect_aix<mode>_nospec): Likewise.
22305 (*call_value_indirect_aix<mode>_nospec): Likewise.
22306 (*call_indirect_elfv2<mode>_nospec): Likewise.
22307 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22308 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22309 change assembly output from . to $.
22310 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22311 (indirect_jump<mode>_nospec): Change assembly output from . to $.
22312 (*tablejump<mode>_internal1_nospec): Likewise.
22314 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
22317 * config/sh/sh_optimize_sett_clrt.cc:
22318 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22320 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
22322 PR tree-optimization/83940
22323 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22324 offset_dt to vect_constant_def rather than vect_unknown_def_type.
22325 (vect_check_load_store_mask): Add a mask_dt_out parameter and
22326 use it to pass back the definition type.
22327 (vect_check_store_rhs): Likewise rhs_dt_out.
22328 (vect_build_gather_load_calls): Add a mask_dt argument and use
22329 it instead of a call to vect_is_simple_use.
22330 (vectorizable_store): Update calls to vect_check_load_store_mask
22331 and vect_check_store_rhs. Use the dt returned by the latter instead
22332 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
22333 instead of calls to vect_is_simple_use. Pass the scalar rather
22334 than the vector operand to vect_is_simple_use when handling
22335 second and subsequent copies of an rhs value.
22336 (vectorizable_load): Update calls to vect_check_load_store_mask
22337 and vect_build_gather_load_calls. Use the cached mask_dt and
22338 gs_info.offset_dt instead of calls to vect_is_simple_use.
22340 2018-01-20 Jakub Jelinek <jakub@redhat.com>
22342 PR middle-end/83945
22343 * tree-emutls.c: Include gimplify.h.
22344 (lower_emutls_2): New function.
22345 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
22346 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
22347 it before further processing.
22350 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
22351 UINTVAL (trueop1) instead of INTVAL (op1).
22353 2018-01-19 Jakub Jelinek <jakub@redhat.com>
22357 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
22358 INCOMING_FRAME_SP_OFFSET if not defined.
22359 (scan_trace): Add ENTRY argument. If true and
22360 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
22361 emit a note to adjust the CFA offset.
22362 (create_cfi_notes): Adjust scan_trace callers.
22363 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
22364 INCOMING_FRAME_SP_OFFSET in the CIE.
22365 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
22366 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
22368 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
22369 * doc/tm.texi: Regenerated.
22371 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22373 PR rtl-optimization/83147
22374 * lra-constraints.c (remove_inheritance_pseudos): Use
22375 lra_substitute_pseudo_within_insn.
22377 2018-01-19 Tom de Vries <tom@codesourcery.com>
22378 Cesar Philippidis <cesar@codesourcery.com>
22381 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
22383 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
22386 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
22387 spaces for function labels.
22389 2018-01-19 Martin Liska <mliska@suse.cz>
22391 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
22392 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
22393 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
22394 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
22395 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
22396 (PRED_CONST_RETURN): Change from 69 to 65.
22397 (PRED_NULL_RETURN): Change from 91 to 71.
22398 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
22399 (PRED_LOOP_GUARD): Change from 66 to 73.
22401 2018-01-19 Martin Liska <mliska@suse.cz>
22403 * predict.c (predict_insn_def): Add new assert.
22404 (struct branch_predictor): Change type to signed integer.
22405 (test_prediction_value_range): Amend test to cover
22406 PROB_UNINITIALIZED.
22407 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
22408 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
22409 (PRED_LOOP_ITERATIONS_MAX): Likewise.
22410 (PRED_LOOP_IV_COMPARE): Likewise.
22411 * predict.h (PROB_UNINITIALIZED): Define new constant.
22413 2018-01-19 Martin Liska <mliska@suse.cz>
22415 * predict.c (dump_prediction): Add new format for
22416 analyze_brprob.py script which is enabled with -details
22418 * profile-count.h (precise_p): New function.
22420 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
22422 PR tree-optimization/83922
22423 * tree-vect-loop.c (vect_verify_full_masking): Return false if
22424 there are no statements that need masking.
22425 (vect_active_double_reduction_p): New function.
22426 (vect_analyze_loop_operations): Use it when handling phis that
22427 are not in the loop header.
22429 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
22431 PR tree-optimization/83914
22432 * tree-vect-loop.c (vectorizable_induction): Don't convert
22433 init_expr or apply the peeling adjustment for inductions
22434 that are nested within the vectorized loop.
22436 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22438 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
22441 2018-01-18 Jakub Jelinek <jakub@redhat.com>
22445 * function.h (gimplify_parameters): Add gimple_seq * argument.
22446 * function.c: Include gimple.h and options.h.
22447 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
22448 for the added local temporaries if needed.
22449 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
22450 if there are any parameter cleanups, wrap whole body into a
22451 try/finally with the cleanups.
22453 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
22456 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
22457 Use GET_MODE_CLASS for scalar floating point.
22459 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
22463 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
22464 Fix call of call_cgraph_insertion_hooks.
22466 2018-01-18 Martin Sebor <msebor@redhat.com>
22468 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
22470 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
22473 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
22476 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
22479 * common.opt: (-ffile-prefix-map): New option.
22480 * opts.c (common_handle_option): Defer it.
22481 * opts-global.c (handle_common_deferred_options): Handle it.
22482 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
22483 * file-prefix-map.h: New file.
22484 (remap_debug_filename, add_debug_prefix_map): ...here.
22485 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
22486 * final.c (debug_prefix_map, add_debug_prefix_map
22487 remap_debug_filename): Move to...
22488 * file-prefix-map.c: New file.
22489 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
22490 generalize, get rid of alloca(), use strrchr() instead of strchr().
22491 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
22492 Implement in terms of add_prefix_map().
22493 (remap_macro_filename, remap_debug_filename): Implement in term of
22495 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
22496 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
22497 * dbxout.c: Include file-prefix-map.h.
22498 * varasm.c: Likewise.
22499 * vmsdbgout.c: Likewise.
22500 * xcoffout.c: Likewise.
22501 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
22502 * doc/cppopts.texi (-fmacro-prefix-map): Document.
22503 * doc/invoke.texi (-ffile-prefix-map): Document.
22504 (-fdebug-prefix-map): Update description.
22506 2018-01-18 Martin Liska <mliska@suse.cz>
22508 * config/i386/i386.c (indirect_thunk_name): Document that also
22510 (output_indirect_thunk): Document why both instructions
22511 (pause and lfence) are generated.
22513 2018-01-18 Richard Biener <rguenther@suse.de>
22515 PR tree-optimization/83887
22516 * graphite-scop-detection.c
22517 (scop_detection::get_nearest_dom_with_single_entry): Remove.
22518 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
22519 (scop_detection::merge_sese): Re-implement with a flood-fill
22520 algorithm that properly finds a SESE region if it exists.
22522 2018-01-18 Jakub Jelinek <jakub@redhat.com>
22525 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
22526 pointer_diff optimizations use view_convert instead of convert.
22528 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22530 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22531 Generate different code for -mno-speculate-indirect-jumps.
22532 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22533 (*call_indirect_aix<mode>): Disable for
22534 -mno-speculate-indirect-jumps.
22535 (*call_indirect_aix<mode>_nospec): New define_insn.
22536 (*call_value_indirect_aix<mode>): Disable for
22537 -mno-speculate-indirect-jumps.
22538 (*call_value_indirect_aix<mode>_nospec): New define_insn.
22539 (*sibcall_nonlocal_sysv<mode>): Generate different code for
22540 -mno-speculate-indirect-jumps.
22541 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22543 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
22545 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
22546 long double type, set the flags for noting the default long double
22547 type, even if we don't pass or return a long double type.
22549 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
22552 * ipa-inline.c (flatten_function): Do not overwrite final inlining
22555 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
22557 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
22558 support for merge[hl].
22559 (fold_mergehl_helper): New helper function.
22560 (tree-vector-builder.h): New #include for tree_vector_builder usage.
22561 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
22562 (altivec_vmrglw_direct): Add xxmrglw insn.
22564 2018-01-17 Andrew Waterman <andrew@sifive.com>
22566 * config/riscv/riscv.c (riscv_conditional_register_usage): If
22567 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
22569 2018-01-17 David Malcolm <dmalcolm@redhat.com>
22572 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
22573 call the lto_location_cache before reading the
22574 DECL_SOURCE_LOCATION of the types.
22576 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
22577 Richard Sandiford <richard.sandiford@linaro.org>
22579 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
22580 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
22581 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
22582 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
22583 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
22585 * config/aarch64/constraints.md (aarch64_movti_operand):
22587 * config/aarch64/predicates.md (Uti): Add new constraint.
22589 2018-01-17 Carl Love <cel@us.ibm.com>
22591 * config/rs6000/vsx.md (define_expand xl_len_r,
22592 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
22593 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
22595 (define_expand, define_insn): Move the shift left from the
22596 define_insn to the define_expand for lxvl and stxvl instructions.
22597 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
22598 and XL_LEN_R definitions to PURE.
22600 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
22602 * config/i386/i386.c (indirect_thunk_name): Declare regno
22603 as unsigned int. Compare regno with INVALID_REGNUM.
22604 (output_indirect_thunk): Ditto.
22605 (output_indirect_thunk_function): Ditto.
22606 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
22607 in the call to output_indirect_thunk_function.
22609 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
22611 PR middle-end/83884
22612 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
22613 rather than the size of inner_type to determine the stack slot size
22614 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
22616 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
22619 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
22622 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
22624 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
22625 endian Linux systems to optionally enable multilibs for selecting
22626 the long double type if the user configured an explicit type.
22627 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
22628 have no long double multilibs if not defined.
22629 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
22630 warn if the user used -mabi={ieee,ibm}longdouble and we built
22631 multilibs for long double.
22632 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
22633 appropriate multilib option.
22634 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
22636 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
22637 for building long double multilibs.
22638 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
22640 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
22642 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
22645 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
22647 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
22650 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
22653 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
22656 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
22658 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
22659 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
22661 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
22663 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
22664 different rtl trees depending on TARGET_64BIT.
22665 (rs6000_gen_lvx): Likewise.
22667 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
22669 * config/visium/visium.md (nop): Tweak comment.
22670 (hazard_nop): Likewise.
22672 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22674 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
22675 -mspeculate-indirect-jumps.
22676 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
22677 for -mno-speculate-indirect-jumps.
22678 (*call_indirect_elfv2<mode>_nospec): New define_insn.
22679 (*call_value_indirect_elfv2<mode>): Disable for
22680 -mno-speculate-indirect-jumps.
22681 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
22682 (indirect_jump): Emit different RTL for
22683 -mno-speculate-indirect-jumps.
22684 (*indirect_jump<mode>): Disable for
22685 -mno-speculate-indirect-jumps.
22686 (*indirect_jump<mode>_nospec): New define_insn.
22687 (tablejump): Emit different RTL for
22688 -mno-speculate-indirect-jumps.
22689 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
22690 (tablejumpsi_nospec): New define_expand.
22691 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
22692 (tablejumpdi_nospec): New define_expand.
22693 (*tablejump<mode>_internal1): Disable for
22694 -mno-speculate-indirect-jumps.
22695 (*tablejump<mode>_internal1_nospec): New define_insn.
22696 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
22699 2018-01-16 Artyom Skrobov tyomitch@gmail.com
22701 * caller-save.c (insert_save): Drop unnecessary parameter. All
22704 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22705 Richard Biener <rguenth@suse.de>
22708 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
22709 return early, inline manually is_gimple_sizepos. Make sure if we
22710 call gimplify_expr we don't end up with a gimple constant.
22711 * tree.c (variably_modified_type_p): Don't return true for
22712 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
22713 * gimplify.h (is_gimple_sizepos): Remove.
22715 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22717 PR tree-optimization/83857
22718 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
22719 vectorizable_live_operation for pure SLP statements.
22720 (vectorizable_live_operation): Handle PHIs.
22722 2018-01-16 Richard Biener <rguenther@suse.de>
22724 PR tree-optimization/83867
22725 * tree-vect-stmts.c (vect_transform_stmt): Precompute
22726 nested_in_vect_loop_p since the scalar stmt may get invalidated.
22728 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22731 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
22732 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
22733 If off is not INTEGER_CST, issue a may not be aligned warning
22734 rather than isn't aligned. Use isn%'t rather than isn't.
22735 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
22737 <case MULT_EXPR>: Improve the case when bottom and one of the
22738 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
22739 operand, in that case check if the other operand is multiple of
22740 bottom divided by the INTEGER_CST operand.
22742 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22745 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
22746 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
22747 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
22748 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
22749 * config/pa/pa.c (pa_function_arg_advance): Likewise.
22750 (pa_function_arg, pa_arg_partial_bytes): Likewise.
22751 (pa_function_arg_size): New function.
22753 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22755 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
22756 in a separate statement.
22758 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22760 PR tree-optimization/83847
22761 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
22762 group gathers and scatters.
22764 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22766 PR rtl-optimization/86620
22767 * params.def (max-sched-ready-insns): Bump minimum value to 1.
22769 PR rtl-optimization/83213
22770 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
22771 to last if both are JUMP_INSNs.
22773 PR tree-optimization/83843
22774 * gimple-ssa-store-merging.c
22775 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
22776 store_immediate_info for bswap/nop orig_stores.
22778 2018-01-15 Andrew Waterman <andrew@sifive.com>
22780 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
22782 <UDIV>: Increase cost if !TARGET_DIV.
22784 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
22786 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
22787 (define_attr "cr_logical_3op"): New.
22788 (cceq_ior_compare): Adjust.
22789 (cceq_ior_compare_complement): Adjust.
22790 (*cceq_rev_compare): Adjust.
22791 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
22792 (is_cracked_insn): Adjust.
22793 (insn_must_be_first_in_group): Adjust.
22794 * config/rs6000/40x.md: Adjust.
22795 * config/rs6000/440.md: Adjust.
22796 * config/rs6000/476.md: Adjust.
22797 * config/rs6000/601.md: Adjust.
22798 * config/rs6000/603.md: Adjust.
22799 * config/rs6000/6xx.md: Adjust.
22800 * config/rs6000/7450.md: Adjust.
22801 * config/rs6000/7xx.md: Adjust.
22802 * config/rs6000/8540.md: Adjust.
22803 * config/rs6000/cell.md: Adjust.
22804 * config/rs6000/e300c2c3.md: Adjust.
22805 * config/rs6000/e500mc.md: Adjust.
22806 * config/rs6000/e500mc64.md: Adjust.
22807 * config/rs6000/e5500.md: Adjust.
22808 * config/rs6000/e6500.md: Adjust.
22809 * config/rs6000/mpc.md: Adjust.
22810 * config/rs6000/power4.md: Adjust.
22811 * config/rs6000/power5.md: Adjust.
22812 * config/rs6000/power6.md: Adjust.
22813 * config/rs6000/power7.md: Adjust.
22814 * config/rs6000/power8.md: Adjust.
22815 * config/rs6000/power9.md: Adjust.
22816 * config/rs6000/rs64.md: Adjust.
22817 * config/rs6000/titan.md: Adjust.
22819 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22821 * config/i386/predicates.md (indirect_branch_operand): Rewrite
22822 ix86_indirect_branch_register logic.
22824 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22826 * config/i386/constraints.md (Bs): Update
22827 ix86_indirect_branch_register check. Don't check
22828 ix86_indirect_branch_register with GOT_memory_operand.
22830 * config/i386/predicates.md (GOT_memory_operand): Don't check
22831 ix86_indirect_branch_register here.
22832 (GOT32_symbol_operand): Likewise.
22834 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22836 * config/i386/predicates.md (constant_call_address_operand):
22837 Rewrite ix86_indirect_branch_register logic.
22838 (sibcall_insn_operand): Likewise.
22840 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22842 * config/i386/constraints.md (Bs): Replace
22843 ix86_indirect_branch_thunk_register with
22844 ix86_indirect_branch_register.
22846 * config/i386/i386.md (indirect_jump): Likewise.
22847 (tablejump): Likewise.
22848 (*sibcall_memory): Likewise.
22849 (*sibcall_value_memory): Likewise.
22850 Peepholes of indirect call and jump via memory: Likewise.
22851 * config/i386/i386.opt: Likewise.
22852 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22853 (GOT_memory_operand): Likewise.
22854 (call_insn_operand): Likewise.
22855 (sibcall_insn_operand): Likewise.
22856 (GOT32_symbol_operand): Likewise.
22858 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22860 PR middle-end/83837
22861 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
22862 type rather than type addr's type points to.
22863 (expand_omp_atomic_mutex): Likewise.
22864 (expand_omp_atomic): Likewise.
22866 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22869 * config/i386/i386.c (output_indirect_thunk_function): Use
22870 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
22871 for __x86_return_thunk.
22873 2018-01-15 Richard Biener <rguenther@suse.de>
22875 PR middle-end/83850
22876 * expmed.c (extract_bit_field_1): Fix typo.
22878 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22881 * config/arm/iterators.md (VF): New mode iterator.
22882 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
22883 Remove integer-related logic from pattern.
22884 (neon_vabd<mode>_3): Likewise.
22886 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22888 PR middle-end/82694
22889 * common.opt (fstrict-overflow): No longer an alias.
22890 (fwrapv-pointer): New option.
22891 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
22892 also for pointer types based on flag_wrapv_pointer.
22893 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
22894 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
22895 opts->x_flag_wrapv got set.
22896 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
22897 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
22898 POINTER_TYPE_OVERFLOW_UNDEFINED.
22899 * match.pd: Likewise in address comparison pattern.
22900 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
22902 2018-01-15 Richard Biener <rguenther@suse.de>
22905 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
22906 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
22907 Reset type names to their identifier if their TYPE_DECL doesn't
22908 have linkage (and thus is used for ODR and devirt).
22909 (save_debug_info_for_decl): Remove.
22910 (save_debug_info_for_type): Likewise.
22911 (add_tree_to_fld_list): Adjust.
22912 * tree-pretty-print.c (dump_generic_node): Make dumping of
22913 type names more robust.
22915 2018-01-15 Richard Biener <rguenther@suse.de>
22917 * BASE-VER: Bump to 8.0.1.
22919 2018-01-14 Martin Sebor <msebor@redhat.com>
22922 * builtins.c (check_access): Avoid warning when the no-warning bit
22925 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
22927 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
22928 * ira-color (allocno_hard_regs_compare): Likewise.
22930 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
22933 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
22934 Use .pushsection/.popsection.
22936 2018-01-14 Martin Sebor <msebor@redhat.com>
22939 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
22941 2018-01-14 Jakub Jelinek <jakub@redhat.com>
22943 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
22944 entry from extra_headers.
22945 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
22946 extra_headers, make the list bitwise identical to the i?86-*-* one.
22948 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22950 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
22951 -mcmodel=large with -mindirect-branch=thunk,
22952 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
22953 -mfunction-return=thunk-extern.
22954 * doc/invoke.texi: Document -mcmodel=large is incompatible with
22955 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
22956 -mfunction-return=thunk and -mfunction-return=thunk-extern.
22958 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22960 * config/i386/i386.c (print_reg): Print the name of the full
22961 integer register without '%'.
22962 (ix86_print_operand): Handle 'V'.
22963 * doc/extend.texi: Document 'V' modifier.
22965 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22967 * config/i386/constraints.md (Bs): Disallow memory operand for
22968 -mindirect-branch-register.
22970 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22971 (GOT_memory_operand): Likewise.
22972 (call_insn_operand): Likewise.
22973 (sibcall_insn_operand): Likewise.
22974 (GOT32_symbol_operand): Likewise.
22975 * config/i386/i386.md (indirect_jump): Call convert_memory_address
22976 for -mindirect-branch-register.
22977 (tablejump): Likewise.
22978 (*sibcall_memory): Likewise.
22979 (*sibcall_value_memory): Likewise.
22980 Disallow peepholes of indirect call and jump via memory for
22981 -mindirect-branch-register.
22982 (*call_pop): Replace m with Bw.
22983 (*call_value_pop): Likewise.
22984 (*sibcall_pop_memory): Replace m with Bs.
22985 * config/i386/i386.opt (mindirect-branch-register): New option.
22986 * doc/invoke.texi: Document -mindirect-branch-register option.
22988 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22990 * config/i386/i386-protos.h (ix86_output_function_return): New.
22991 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
22992 set function_return_type.
22993 (indirect_thunk_name): Add ret_p to indicate thunk for function
22995 (output_indirect_thunk_function): Pass false to
22996 indirect_thunk_name.
22997 (ix86_output_indirect_branch_via_reg): Likewise.
22998 (ix86_output_indirect_branch_via_push): Likewise.
22999 (output_indirect_thunk_function): Create alias for function
23000 return thunk if regno < 0.
23001 (ix86_output_function_return): New function.
23002 (ix86_handle_fndecl_attribute): Handle function_return.
23003 (ix86_attribute_table): Add function_return.
23004 * config/i386/i386.h (machine_function): Add
23005 function_return_type.
23006 * config/i386/i386.md (simple_return_internal): Use
23007 ix86_output_function_return.
23008 (simple_return_internal_long): Likewise.
23009 * config/i386/i386.opt (mfunction-return=): New option.
23010 (indirect_branch): Mention -mfunction-return=.
23011 * doc/extend.texi: Document function_return function attribute.
23012 * doc/invoke.texi: Document -mfunction-return= option.
23014 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23016 * config/i386/i386-opts.h (indirect_branch): New.
23017 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23018 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23019 with local indirect jump when converting indirect call and jump.
23020 (ix86_set_indirect_branch_type): New.
23021 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23022 (indirectlabelno): New.
23023 (indirect_thunk_needed): Likewise.
23024 (indirect_thunk_bnd_needed): Likewise.
23025 (indirect_thunks_used): Likewise.
23026 (indirect_thunks_bnd_used): Likewise.
23027 (INDIRECT_LABEL): Likewise.
23028 (indirect_thunk_name): Likewise.
23029 (output_indirect_thunk): Likewise.
23030 (output_indirect_thunk_function): Likewise.
23031 (ix86_output_indirect_branch_via_reg): Likewise.
23032 (ix86_output_indirect_branch_via_push): Likewise.
23033 (ix86_output_indirect_branch): Likewise.
23034 (ix86_output_indirect_jmp): Likewise.
23035 (ix86_code_end): Call output_indirect_thunk_function if needed.
23036 (ix86_output_call_insn): Call ix86_output_indirect_branch if
23038 (ix86_handle_fndecl_attribute): Handle indirect_branch.
23039 (ix86_attribute_table): Add indirect_branch.
23040 * config/i386/i386.h (machine_function): Add indirect_branch_type
23041 and has_local_indirect_jump.
23042 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23044 (tablejump): Likewise.
23045 (*indirect_jump): Use ix86_output_indirect_jmp.
23046 (*tablejump_1): Likewise.
23047 (simple_return_indirect_internal): Likewise.
23048 * config/i386/i386.opt (mindirect-branch=): New option.
23049 (indirect_branch): New.
23052 (thunk-inline): Likewise.
23053 (thunk-extern): Likewise.
23054 * doc/extend.texi: Document indirect_branch function attribute.
23055 * doc/invoke.texi: Document -mindirect-branch= option.
23057 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
23060 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23062 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
23064 * ipa-inline.c (want_inline_small_function_p): Return false if
23065 inlining has already failed with CIF_FINAL_ERROR.
23066 (update_caller_keys): Call want_inline_small_function_p before
23068 (update_callee_keys): Likewise.
23070 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
23072 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23074 (rs6000_quadword_masked_address_p): Likewise.
23075 (quad_aligned_load_p): Likewise.
23076 (quad_aligned_store_p): Likewise.
23077 (const_load_sequence_p): Add comment to describe the outer-most loop.
23078 (mimic_memory_attributes_and_flags): New function.
23079 (rs6000_gen_stvx): Likewise.
23080 (replace_swapped_aligned_store): Likewise.
23081 (rs6000_gen_lvx): Likewise.
23082 (replace_swapped_aligned_load): Likewise.
23083 (replace_swapped_load_constant): Capitalize argument name in
23084 comment describing this function.
23085 (rs6000_analyze_swaps): Add a third pass to search for vector loads
23086 and stores that access quad-word aligned addresses and replace
23087 with stvx or lvx instructions when appropriate.
23088 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23089 New function prototype.
23090 (rs6000_quadword_masked_address_p): Likewise.
23091 (rs6000_gen_lvx): Likewise.
23092 (rs6000_gen_stvx): Likewise.
23093 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23094 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23095 when memory address is aligned.
23096 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23097 this split to select lvx instruction when memory address is aligned.
23098 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23099 instruction when memory address is aligned.
23100 (*vsx_le_perm_load_v16qi): Likewise.
23101 (four unnamed splitters): Modify to select the stvx instruction
23102 when memory is aligned.
23104 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
23106 * predict.c (determine_unlikely_bbs): Handle correctly BBs
23107 which appears in the queue multiple times.
23109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23110 Alan Hayward <alan.hayward@arm.com>
23111 David Sherwood <david.sherwood@arm.com>
23113 * tree-vectorizer.h (vec_lower_bound): New structure.
23114 (_loop_vec_info): Add check_nonzero and lower_bounds.
23115 (LOOP_VINFO_CHECK_NONZERO): New macro.
23116 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23117 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23118 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23119 fields. Make seg_len the distance travelled, not including the
23121 (dr_direction_indicator): Declare.
23122 (dr_zero_step_indicator): Likewise.
23123 (dr_known_forward_stride_p): Likewise.
23124 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23126 (runtime_alias_check_p): Allow runtime alias checks with
23128 (operator ==): Compare access_size and align.
23129 (prune_runtime_alias_test_list): Rework for new distinction between
23130 the access_size and seg_len.
23131 (create_intersect_range_checks_index): Likewise. Cope with polynomial
23133 (get_segment_min_max): New function.
23134 (create_intersect_range_checks): Use it.
23135 (dr_step_indicator): New function.
23136 (dr_direction_indicator): Likewise.
23137 (dr_zero_step_indicator): Likewise.
23138 (dr_known_forward_stride_p): Likewise.
23139 * tree-loop-distribution.c (data_ref_segment_size): Return
23140 DR_STEP * (niters - 1).
23141 (compute_alias_check_pairs): Update call to the dr_with_seg_len
23143 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23144 (vect_preserves_scalar_order_p): New function, split out from...
23145 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
23146 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23147 (vect_vfa_access_size): New function.
23148 (vect_vfa_align): Likewise.
23149 (vect_compile_time_alias): Take access_size_a and access_b arguments.
23150 (dump_lower_bound): New function.
23151 (vect_check_lower_bound): Likewise.
23152 (vect_small_gap_p): Likewise.
23153 (vectorizable_with_step_bound_p): Likewise.
23154 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23155 depencies if the vectorization factor is 1. Convert the checks
23156 for nonzero steps into checks on the bounds of DR_STEP. Try using
23157 a bunds check for variable steps if the minimum required step is
23158 relatively small. Update calls to the dr_with_seg_len
23159 constructor and to vect_compile_time_alias.
23160 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23162 (vect_loop_versioning): Call it.
23163 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23165 (vect_estimate_min_profitable_iters): Account for any bounds checks.
23167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23168 Alan Hayward <alan.hayward@arm.com>
23169 David Sherwood <david.sherwood@arm.com>
23171 * doc/sourcebuild.texi (vect_scatter_store): Document.
23172 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23174 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23176 * genopinit.c (main): Add supports_vec_scatter_store and
23177 supports_vec_scatter_store_cached to target_optabs.
23178 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23179 IFN_MASK_SCATTER_STORE.
23180 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23182 * internal-fn.h (internal_store_fn_p): Declare.
23183 (internal_fn_stored_value_index): Likewise.
23184 * internal-fn.c (scatter_store_direct): New macro.
23185 (expand_scatter_store_optab_fn): New function.
23186 (direct_scatter_store_optab_supported_p): New macro.
23187 (internal_store_fn_p): New function.
23188 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23189 IFN_MASK_SCATTER_STORE.
23190 (internal_fn_mask_index): Likewise.
23191 (internal_fn_stored_value_index): New function.
23192 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23193 for scatter stores.
23194 * optabs-query.h (supports_vec_scatter_store_p): Declare.
23195 * optabs-query.c (supports_vec_scatter_store_p): New function.
23196 * tree-vectorizer.h (vect_get_store_rhs): Declare.
23197 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23198 true for scatter stores.
23199 (vect_gather_scatter_fn_p): Handle scatter stores too.
23200 (vect_check_gather_scatter): Consider using scatter stores if
23201 supports_vec_scatter_store_p.
23202 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23203 scatter stores too.
23204 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23205 internal_fn_stored_value_index.
23206 (check_load_store_masking): Handle scatter stores too.
23207 (vect_get_store_rhs): Make public.
23208 (vectorizable_call): Use internal_store_fn_p.
23209 (vectorizable_store): Handle scatter store internal functions.
23210 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23211 when deciding whether the end of the group has been reached.
23212 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23213 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23214 (mask_scatter_store<mode>): New insns.
23216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23217 Alan Hayward <alan.hayward@arm.com>
23218 David Sherwood <david.sherwood@arm.com>
23220 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23221 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23222 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23224 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23225 Use vect_truncate_gather_scatter_offset if we can't treat the
23226 operation as a normal gather load or scatter store.
23227 (get_group_load_store_type): Take the gather_scatter_info
23228 as argument. Try using a gather load or scatter store for
23229 single-element groups.
23230 (get_load_store_type): Update calls to get_group_load_store_type
23231 and vect_use_strided_gather_scatters_p.
23233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23234 Alan Hayward <alan.hayward@arm.com>
23235 David Sherwood <david.sherwood@arm.com>
23237 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23238 optional tree argument.
23239 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23241 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23242 but continue to use the current value as a fallback.
23243 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23244 to compare the updates.
23245 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23246 (get_load_store_type): Use it when handling a strided access.
23247 (vect_get_strided_load_store_ops): New function.
23248 (vect_get_data_ptr_increment): Likewise.
23249 (vectorizable_load): Handle strided gather loads. Always pass
23250 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
23257 (mask_gather_load@var{m}): Likewise.
23258 * genopinit.c (main): Add supports_vec_gather_load and
23259 supports_vec_gather_load_cached to target_optabs.
23260 * optabs-tree.c (init_tree_optimization_optabs): Use
23261 ggc_cleared_alloc to allocate target_optabs.
23262 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23263 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23265 * internal-fn.h (internal_load_fn_p): Declare.
23266 (internal_gather_scatter_fn_p): Likewise.
23267 (internal_fn_mask_index): Likewise.
23268 (internal_gather_scatter_fn_supported_p): Likewise.
23269 * internal-fn.c (gather_load_direct): New macro.
23270 (expand_gather_load_optab_fn): New function.
23271 (direct_gather_load_optab_supported_p): New macro.
23272 (direct_internal_fn_optab): New function.
23273 (internal_load_fn_p): Likewise.
23274 (internal_gather_scatter_fn_p): Likewise.
23275 (internal_fn_mask_index): Likewise.
23276 (internal_gather_scatter_fn_supported_p): Likewise.
23277 * optabs-query.c (supports_at_least_one_mode_p): New function.
23278 (supports_vec_gather_load_p): Likewise.
23279 * optabs-query.h (supports_vec_gather_load_p): Declare.
23280 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23281 and memory_type field.
23282 (NUM_PATTERNS): Bump to 15.
23283 * tree-vect-data-refs.c: Include internal-fn.h.
23284 (vect_gather_scatter_fn_p): New function.
23285 (vect_describe_gather_scatter_call): Likewise.
23286 (vect_check_gather_scatter): Try using internal functions for
23287 gather loads. Recognize existing calls to a gather load function.
23288 (vect_analyze_data_refs): Consider using gather loads if
23289 supports_vec_gather_load_p.
23290 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23291 (vect_get_gather_scatter_offset_type): Likewise.
23292 (vect_convert_mask_for_vectype): Likewise.
23293 (vect_add_conversion_to_patterm): Likewise.
23294 (vect_try_gather_scatter_pattern): Likewise.
23295 (vect_recog_gather_scatter_pattern): New pattern recognizer.
23296 (vect_vect_recog_func_ptrs): Add it.
23297 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23298 internal_fn_mask_index and internal_gather_scatter_fn_p.
23299 (check_load_store_masking): Take the gather_scatter_info as an
23300 argument and handle gather loads.
23301 (vect_get_gather_scatter_ops): New function.
23302 (vectorizable_call): Check internal_load_fn_p.
23303 (vectorizable_load): Likewise. Handle gather load internal
23305 (vectorizable_store): Update call to check_load_store_masking.
23306 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23307 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23308 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23309 (aarch64_gather_scale_operand_d): New predicates.
23310 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23311 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
23318 * doc/md.texi (fold_left_plus_@var{m}): Document.
23319 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23320 * internal-fn.c (fold_left_direct): Define.
23321 (expand_fold_left_optab_fn): Likewise.
23322 (direct_fold_left_optab_supported_p): Likewise.
23323 * fold-const-call.c (fold_const_fold_left): New function.
23324 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23325 * tree-parloops.c (valid_reduction_p): New function.
23326 (gather_scalar_reductions): Use it.
23327 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23328 (vect_finish_replace_stmt): Declare.
23329 * tree-vect-loop.c (fold_left_reduction_fn): New function.
23330 (needs_fold_left_reduction_p): New function, split out from...
23331 (vect_is_simple_reduction): ...here. Accept reductions that
23332 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23333 (vect_force_simple_reduction): Also store the reduction type in
23334 the assignment's STMT_VINFO_REDUC_TYPE.
23335 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
23336 (merge_with_identity): New function.
23337 (vect_expand_fold_left): Likewise.
23338 (vectorize_fold_left_reduction): Likewise.
23339 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
23340 scalar phi in place for it. Check for target support and reject
23341 cases that would reassociate the operation. Defer the transform
23342 phase to vectorize_fold_left_reduction.
23343 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
23344 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
23345 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
23347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23349 * tree-if-conv.c (predicate_mem_writes): Remove redundant
23350 call to ifc_temp_var.
23352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23353 Alan Hayward <alan.hayward@arm.com>
23354 David Sherwood <david.sherwood@arm.com>
23356 * target.def (legitimize_address_displacement): Take the original
23357 offset as a poly_int.
23358 * targhooks.h (default_legitimize_address_displacement): Update
23360 * targhooks.c (default_legitimize_address_displacement): Likewise.
23361 * doc/tm.texi: Regenerate.
23362 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
23363 as an argument, moving assert of ad->disp == ad->disp_term to...
23364 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
23365 Try calling targetm.legitimize_address_displacement before expanding
23366 the address rather than afterwards, and adjust for the new interface.
23367 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
23368 Match the new hook interface. Handle SVE addresses.
23369 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
23370 new hook interface.
23372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23374 * Makefile.in (OBJS): Add early-remat.o.
23375 * target.def (select_early_remat_modes): New hook.
23376 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
23377 * doc/tm.texi: Regenerate.
23378 * targhooks.h (default_select_early_remat_modes): Declare.
23379 * targhooks.c (default_select_early_remat_modes): New function.
23380 * timevar.def (TV_EARLY_REMAT): New timevar.
23381 * passes.def (pass_early_remat): New pass.
23382 * tree-pass.h (make_pass_early_remat): Declare.
23383 * early-remat.c: New file.
23384 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
23386 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
23388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23389 Alan Hayward <alan.hayward@arm.com>
23390 David Sherwood <david.sherwood@arm.com>
23392 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
23393 vfm1 with a bound_epilog parameter.
23394 (vect_do_peeling): Update calls accordingly, and move the prologue
23395 call earlier in the function. Treat the base bound_epilog as 0 for
23396 fully-masked loops and retain vf - 1 for other loops. Add 1 to
23397 this base when peeling for gaps.
23398 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
23399 with fully-masked loops.
23400 (vect_estimate_min_profitable_iters): Handle the single peeled
23401 iteration in that case.
23403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23404 Alan Hayward <alan.hayward@arm.com>
23405 David Sherwood <david.sherwood@arm.com>
23407 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
23408 single-element interleaving even if the size is not a power of 2.
23409 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
23410 accesses for single-element interleaving if the group size is
23413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23414 Alan Hayward <alan.hayward@arm.com>
23415 David Sherwood <david.sherwood@arm.com>
23417 * doc/md.texi (fold_extract_last_@var{m}): Document.
23418 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
23419 * optabs.def (fold_extract_last_optab): New optab.
23420 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
23421 * internal-fn.c (fold_extract_direct): New macro.
23422 (expand_fold_extract_optab_fn): Likewise.
23423 (direct_fold_extract_optab_supported_p): Likewise.
23424 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
23425 * tree-vect-loop.c (vect_model_reduction_cost): Handle
23426 EXTRACT_LAST_REDUCTION.
23427 (get_initial_def_for_reduction): Do not create an initial vector
23428 for EXTRACT_LAST_REDUCTION reductions.
23429 (vectorizable_reduction): Leave the scalar phi in place for
23430 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
23431 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
23432 epilogue code for EXTRACT_LAST_REDUCTION and defer the
23433 transform phase to vectorizable_condition.
23434 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
23436 (vect_finish_stmt_generation): ...here.
23437 (vect_finish_replace_stmt): New function.
23438 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
23439 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
23441 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
23443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23444 Alan Hayward <alan.hayward@arm.com>
23445 David Sherwood <david.sherwood@arm.com>
23447 * doc/md.texi (extract_last_@var{m}): Document.
23448 * optabs.def (extract_last_optab): New optab.
23449 * internal-fn.def (EXTRACT_LAST): New internal function.
23450 * internal-fn.c (cond_unary_direct): New macro.
23451 (expand_cond_unary_optab_fn): Likewise.
23452 (direct_cond_unary_optab_supported_p): Likewise.
23453 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
23454 loops using EXTRACT_LAST.
23455 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
23456 (extract_last_<mode>): ...this optab.
23457 (vec_extract<mode><Vel>): Update accordingly.
23459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23460 Alan Hayward <alan.hayward@arm.com>
23461 David Sherwood <david.sherwood@arm.com>
23463 * target.def (empty_mask_is_expensive): New hook.
23464 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
23465 * doc/tm.texi: Regenerate.
23466 * targhooks.h (default_empty_mask_is_expensive): Declare.
23467 * targhooks.c (default_empty_mask_is_expensive): New function.
23468 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
23469 if the target says that empty masks are expensive.
23470 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
23472 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
23474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23475 Alan Hayward <alan.hayward@arm.com>
23476 David Sherwood <david.sherwood@arm.com>
23478 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
23479 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
23480 (vect_use_loop_mask_for_alignment_p): New function.
23481 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
23482 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
23483 niters_skip argument. Make sure that the first niters_skip elements
23484 of the first iteration are inactive.
23485 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
23486 Update call to vect_set_loop_masks_directly.
23487 (get_misalign_in_elems): New function, split out from...
23488 (vect_gen_prolog_loop_niters): ...here.
23489 (vect_update_init_of_dr): Take a code argument that specifies whether
23490 the adjustment should be added or subtracted.
23491 (vect_update_init_of_drs): Likewise.
23492 (vect_prepare_for_masked_peels): New function.
23493 (vect_do_peeling): Skip prologue peeling if we're using a mask
23494 instead. Update call to vect_update_inits_of_drs.
23495 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23497 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
23498 alignment. Do not include the number of peeled iterations in
23499 the minimum threshold in that case.
23500 (vectorizable_induction): Adjust the start value down by
23501 LOOP_VINFO_MASK_SKIP_NITERS iterations.
23502 (vect_transform_loop): Call vect_prepare_for_masked_peels.
23503 Take the number of skipped iterations into account when calculating
23505 * tree-vect-stmts.c (vect_gen_while_not): New function.
23507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23508 Alan Hayward <alan.hayward@arm.com>
23509 David Sherwood <david.sherwood@arm.com>
23511 * doc/sourcebuild.texi (vect_fully_masked): Document.
23512 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
23513 default value to 0.
23514 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
23516 (vect_analyze_loop_2): ...here. Don't check the vectorization
23517 factor against the number of loop iterations if the loop is
23520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23521 Alan Hayward <alan.hayward@arm.com>
23522 David Sherwood <david.sherwood@arm.com>
23524 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
23525 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
23526 (dump_groups): Update accordingly.
23527 (iv_use::mem_type): New member variable.
23528 (address_p): New function.
23529 (record_use): Add a mem_type argument and initialize the new
23531 (record_group_use): Add a mem_type argument. Use address_p.
23532 Remove obsolete null checks of base_object. Update call to record_use.
23533 (find_interesting_uses_op): Update call to record_group_use.
23534 (find_interesting_uses_cond): Likewise.
23535 (find_interesting_uses_address): Likewise.
23536 (get_mem_type_for_internal_fn): New function.
23537 (find_address_like_use): Likewise.
23538 (find_interesting_uses_stmt): Try find_address_like_use before
23539 calling find_interesting_uses_op.
23540 (addr_offset_valid_p): Use the iv mem_type field as the type
23541 of the addressed memory.
23542 (add_autoinc_candidates): Likewise.
23543 (get_address_cost): Likewise.
23544 (split_small_address_groups_p): Use address_p.
23545 (split_address_groups): Likewise.
23546 (add_iv_candidate_for_use): Likewise.
23547 (autoinc_possible_for_pair): Likewise.
23548 (rewrite_groups): Likewise.
23549 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
23550 (determine_group_iv_cost): Update after split of USE_ADDRESS.
23551 (get_alias_ptr_type_for_ptr_address): New function.
23552 (rewrite_use_address): Rewrite address uses in calls that were
23553 identified by find_address_like_use.
23555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23556 Alan Hayward <alan.hayward@arm.com>
23557 David Sherwood <david.sherwood@arm.com>
23559 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
23561 * gimple-expr.h (is_gimple_addressable: Likewise.
23562 * gimple-expr.c (is_gimple_address): Likewise.
23563 * internal-fn.c (expand_call_mem_ref): New function.
23564 (expand_mask_load_optab_fn): Use it.
23565 (expand_mask_store_optab_fn): Likewise.
23567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23568 Alan Hayward <alan.hayward@arm.com>
23569 David Sherwood <david.sherwood@arm.com>
23571 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
23572 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
23573 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
23574 (cond_umax@var{mode}): Document.
23575 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
23576 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
23577 (cond_umin_optab, cond_umax_optab): New optabs.
23578 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
23579 (COND_IOR, COND_XOR): New internal functions.
23580 * internal-fn.h (get_conditional_internal_fn): Declare.
23581 * internal-fn.c (cond_binary_direct): New macro.
23582 (expand_cond_binary_optab_fn): Likewise.
23583 (direct_cond_binary_optab_supported_p): Likewise.
23584 (get_conditional_internal_fn): New function.
23585 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
23586 Cope with reduction statements that are vectorized as calls rather
23588 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
23589 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
23590 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
23591 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
23592 (UNSPEC_COND_EOR): New unspecs.
23593 (optab): Add mappings for them.
23594 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
23595 (sve_int_op, sve_fp_op): New int attributes.
23597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23598 Alan Hayward <alan.hayward@arm.com>
23599 David Sherwood <david.sherwood@arm.com>
23601 * optabs.def (while_ult_optab): New optab.
23602 * doc/md.texi (while_ult@var{m}@var{n}): Document.
23603 * internal-fn.def (WHILE_ULT): New internal function.
23604 * internal-fn.h (direct_internal_fn_supported_p): New override
23605 that takes two types as argument.
23606 * internal-fn.c (while_direct): New macro.
23607 (expand_while_optab_fn): New function.
23608 (convert_optab_supported_p): Likewise.
23609 (direct_while_optab_supported_p): New macro.
23610 * wide-int.h (wi::udiv_ceil): New function.
23611 * tree-vectorizer.h (rgroup_masks): New structure.
23612 (vec_loop_masks): New typedef.
23613 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
23614 and fully_masked_p.
23615 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
23616 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
23617 (vect_max_vf): New function.
23618 (slpeel_make_loop_iterate_ntimes): Delete.
23619 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
23620 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
23621 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
23622 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
23623 internal-fn.h, stor-layout.h and optabs-query.h.
23624 (vect_set_loop_mask): New function.
23625 (add_preheader_seq): Likewise.
23626 (add_header_seq): Likewise.
23627 (interleave_supported_p): Likewise.
23628 (vect_maybe_permute_loop_masks): Likewise.
23629 (vect_set_loop_masks_directly): Likewise.
23630 (vect_set_loop_condition_masked): Likewise.
23631 (vect_set_loop_condition_unmasked): New function, split out from
23632 slpeel_make_loop_iterate_ntimes.
23633 (slpeel_make_loop_iterate_ntimes): Rename to..
23634 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
23635 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
23636 (vect_do_peeling): Update call accordingly.
23637 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
23639 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23640 mask_compare_type, can_fully_mask_p and fully_masked_p.
23641 (release_vec_loop_masks): New function.
23642 (_loop_vec_info): Use it to free the loop masks.
23643 (can_produce_all_loop_masks_p): New function.
23644 (vect_get_max_nscalars_per_iter): Likewise.
23645 (vect_verify_full_masking): Likewise.
23646 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
23647 retries, and free the mask rgroups before retrying. Check loop-wide
23648 reasons for disallowing fully-masked loops. Make the final decision
23649 about whether use a fully-masked loop or not.
23650 (vect_estimate_min_profitable_iters): Do not assume that peeling
23651 for the number of iterations will be needed for fully-masked loops.
23652 (vectorizable_reduction): Disable fully-masked loops.
23653 (vectorizable_live_operation): Likewise.
23654 (vect_halve_mask_nunits): New function.
23655 (vect_double_mask_nunits): Likewise.
23656 (vect_record_loop_mask): Likewise.
23657 (vect_get_loop_mask): Likewise.
23658 (vect_transform_loop): Handle the case in which the final loop
23659 iteration might handle a partial vector. Call vect_set_loop_condition
23660 instead of slpeel_make_loop_iterate_ntimes.
23661 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
23662 (check_load_store_masking): New function.
23663 (prepare_load_store_mask): Likewise.
23664 (vectorizable_store): Handle fully-masked loops.
23665 (vectorizable_load): Likewise.
23666 (supportable_widening_operation): Use vect_halve_mask_nunits for
23668 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
23669 (vect_gen_while): New function.
23670 * config/aarch64/aarch64.md (umax<mode>3): New expander.
23671 (aarch64_uqdec<mode>): New insn.
23673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23674 Alan Hayward <alan.hayward@arm.com>
23675 David Sherwood <david.sherwood@arm.com>
23677 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
23678 (reduc_xor_scal_optab): New optabs.
23679 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
23680 (reduc_xor_scal_@var{m}): Document.
23681 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
23682 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
23683 internal functions.
23684 * fold-const-call.c (fold_const_call): Handle them.
23685 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
23686 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
23687 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
23688 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
23689 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
23690 (UNSPEC_XORV): New unspecs.
23691 (optab): Add entries for them.
23692 (BITWISEV): New int iterator.
23693 (bit_reduc_op): New int attributes.
23695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23696 Alan Hayward <alan.hayward@arm.com>
23697 David Sherwood <david.sherwood@arm.com>
23699 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
23700 * internal-fn.def (VEC_SHL_INSERT): New internal function.
23701 * optabs.def (vec_shl_insert_optab): New optab.
23702 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
23703 (duplicate_and_interleave): Likewise.
23704 * tree-vect-loop.c: Include internal-fn.h.
23705 (neutral_op_for_slp_reduction): New function, split out from
23706 get_initial_defs_for_reduction.
23707 (get_initial_def_for_reduction): Handle option 2 for variable-length
23708 vectors by loading the neutral value into a vector and then shifting
23709 the initial value into element 0.
23710 (get_initial_defs_for_reduction): Replace the code argument with
23711 the neutral value calculated by neutral_op_for_slp_reduction.
23712 Use gimple_build_vector for constant-length vectors.
23713 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
23714 but the first group_size elements have a neutral value.
23715 Use duplicate_and_interleave otherwise.
23716 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
23717 Update call to get_initial_defs_for_reduction. Handle SLP
23718 reductions for variable-length vectors by creating one vector
23719 result for each scalar result, with the elements associated
23720 with other scalar results stubbed out with the neutral value.
23721 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
23722 Require IFN_VEC_SHL_INSERT for double reductions on
23723 variable-length vectors, or SLP reductions that have
23724 a neutral value. Require can_duplicate_and_interleave_p
23725 support for variable-length unchained SLP reductions if there
23726 is no neutral value, such as for MIN/MAX reductions. Also require
23727 the number of vector elements to be a multiple of the number of
23728 SLP statements when doing variable-length unchained SLP reductions.
23729 Update call to vect_create_epilog_for_reduction.
23730 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
23731 and remove initial values.
23732 (duplicate_and_interleave): Make public.
23733 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
23734 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
23736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23737 Alan Hayward <alan.hayward@arm.com>
23738 David Sherwood <david.sherwood@arm.com>
23740 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
23741 (can_duplicate_and_interleave_p): New function.
23742 (vect_get_and_check_slp_defs): Take the vector of statements
23743 rather than just the current one. Remove excess parentheses.
23744 Restriction rejectinon of vect_constant_def and vect_external_def
23745 for variable-length vectors to boolean types, or types for which
23746 can_duplicate_and_interleave_p is false.
23747 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
23748 (duplicate_and_interleave): New function.
23749 (vect_get_constant_vectors): Use gimple_build_vector for
23750 constant-length vectors and suitable variable-length constant
23751 vectors. Use duplicate_and_interleave for other variable-length
23752 vectors. Don't defer the update when inserting new statements.
23754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23755 Alan Hayward <alan.hayward@arm.com>
23756 David Sherwood <david.sherwood@arm.com>
23758 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
23759 min_profitable_iters doesn't go negative.
23761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23762 Alan Hayward <alan.hayward@arm.com>
23763 David Sherwood <david.sherwood@arm.com>
23765 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
23766 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
23767 * optabs.def (vec_mask_load_lanes_optab): New optab.
23768 (vec_mask_store_lanes_optab): Likewise.
23769 * internal-fn.def (MASK_LOAD_LANES): New internal function.
23770 (MASK_STORE_LANES): Likewise.
23771 * internal-fn.c (mask_load_lanes_direct): New macro.
23772 (mask_store_lanes_direct): Likewise.
23773 (expand_mask_load_optab_fn): Handle masked operations.
23774 (expand_mask_load_lanes_optab_fn): New macro.
23775 (expand_mask_store_optab_fn): Handle masked operations.
23776 (expand_mask_store_lanes_optab_fn): New macro.
23777 (direct_mask_load_lanes_optab_supported_p): Likewise.
23778 (direct_mask_store_lanes_optab_supported_p): Likewise.
23779 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
23781 (vect_load_lanes_supported): Likewise.
23782 * tree-vect-data-refs.c (strip_conversion): New function.
23783 (can_group_stmts_p): Likewise.
23784 (vect_analyze_data_ref_accesses): Use it instead of checking
23785 for a pair of assignments.
23786 (vect_store_lanes_supported): Take a masked_p parameter.
23787 (vect_load_lanes_supported): Likewise.
23788 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
23789 vect_store_lanes_supported and vect_load_lanes_supported.
23790 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
23791 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
23792 parameter. Don't allow gaps for masked accesses.
23793 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
23794 and vect_load_lanes_supported.
23795 (get_load_store_type): Take a masked_p parameter and update
23796 call to get_group_load_store_type.
23797 (vectorizable_store): Update call to get_load_store_type.
23798 Handle IFN_MASK_STORE_LANES.
23799 (vectorizable_load): Update call to get_load_store_type.
23800 Handle IFN_MASK_LOAD_LANES.
23802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23803 Alan Hayward <alan.hayward@arm.com>
23804 David Sherwood <david.sherwood@arm.com>
23806 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
23808 * config/aarch64/aarch64-protos.h
23809 (aarch64_sve_struct_memory_operand_p): Declare.
23810 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
23811 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
23812 (VPRED, vpred): Handle SVE structure modes.
23813 * config/aarch64/constraints.md (Utx): New constraint.
23814 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
23815 (aarch64_sve_struct_nonimmediate_operand): New predicates.
23816 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
23817 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
23818 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
23819 structure modes. Split into pieces after RA.
23820 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
23821 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
23823 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
23824 SVE structure modes.
23825 (aarch64_classify_address): Likewise.
23826 (sizetochar): Move earlier in file.
23827 (aarch64_print_operand): Handle SVE register lists.
23828 (aarch64_array_mode): New function.
23829 (aarch64_sve_struct_memory_operand_p): Likewise.
23830 (TARGET_ARRAY_MODE): Redefine.
23832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23833 Alan Hayward <alan.hayward@arm.com>
23834 David Sherwood <david.sherwood@arm.com>
23836 * target.def (array_mode): New target hook.
23837 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
23838 * doc/tm.texi: Regenerate.
23839 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
23840 * hooks.c (hook_optmode_mode_uhwi_none): New function.
23841 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
23842 targetm.array_mode.
23843 * stor-layout.c (mode_for_array): Likewise. Support polynomial
23846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23847 Alan Hayward <alan.hayward@arm.com>
23848 David Sherwood <david.sherwood@arm.com>
23850 * fold-const.c (fold_binary_loc): Check the argument types
23851 rather than the result type when testing for a vector operation.
23853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23855 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
23856 * doc/tm.texi: Regenerate.
23858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23859 Alan Hayward <alan.hayward@arm.com>
23860 David Sherwood <david.sherwood@arm.com>
23862 * doc/invoke.texi (-msve-vector-bits=): Document new option.
23863 (sve): Document new AArch64 extension.
23864 * doc/md.texi (w): Extend the description of the AArch64
23865 constraint to include SVE vectors.
23866 (Upl, Upa): Document new AArch64 predicate constraints.
23867 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
23869 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
23870 (msve-vector-bits=): New option.
23871 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
23872 SVE when these are disabled.
23873 (sve): New extension.
23874 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
23875 modes. Adjust their number of units based on aarch64_sve_vg.
23876 (MAX_BITSIZE_MODE_ANY_MODE): Define.
23877 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
23878 aarch64_addr_query_type.
23879 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
23880 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
23881 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
23882 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
23883 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
23884 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
23885 (aarch64_simd_imm_zero_p): Delete.
23886 (aarch64_check_zero_based_sve_index_immediate): Declare.
23887 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23888 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23889 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23890 (aarch64_sve_float_mul_immediate_p): Likewise.
23891 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23892 rather than an rtx.
23893 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
23894 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
23895 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
23896 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
23897 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
23898 (aarch64_regmode_natural_size): Likewise.
23899 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
23900 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
23902 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
23903 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
23904 for VG and the SVE predicate registers.
23905 (V_ALIASES): Add a "z"-prefixed alias.
23906 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
23907 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
23908 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
23909 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
23910 (REG_CLASS_NAMES): Add entries for them.
23911 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
23912 and the predicate registers.
23913 (aarch64_sve_vg): Declare.
23914 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
23915 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
23916 (REGMODE_NATURAL_SIZE): Define.
23917 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
23919 * config/aarch64/aarch64.c: Include cfgrtl.h.
23920 (simd_immediate_info): Add a constructor for series vectors,
23921 and an associated step field.
23922 (aarch64_sve_vg): New variable.
23923 (aarch64_dbx_register_number): Handle VG and the predicate registers.
23924 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
23925 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
23926 (VEC_ANY_DATA, VEC_STRUCT): New constants.
23927 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
23928 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
23929 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
23930 (aarch64_get_mask_mode): New functions.
23931 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
23932 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
23933 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
23934 predicate modes and predicate registers. Explicitly restrict
23935 GPRs to modes of 16 bytes or smaller. Only allow FP registers
23936 to store a vector mode if it is recognized by
23937 aarch64_classify_vector_mode.
23938 (aarch64_regmode_natural_size): New function.
23939 (aarch64_hard_regno_caller_save_mode): Return the original mode
23941 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
23942 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
23943 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
23944 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
23946 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
23947 does not overlap dest if the function is frame-related. Handle
23949 (aarch64_split_add_offset): New function.
23950 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
23951 them aarch64_add_offset.
23952 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
23953 and update call to aarch64_sub_sp.
23954 (aarch64_add_cfa_expression): New function.
23955 (aarch64_expand_prologue): Pass extra temporary registers to the
23956 functions above. Handle the case in which we need to emit new
23957 DW_CFA_expressions for registers that were originally saved
23958 relative to the stack pointer, but now have to be expressed
23959 relative to the frame pointer.
23960 (aarch64_output_mi_thunk): Pass extra temporary registers to the
23962 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
23963 IP0 and IP1 values for SVE frames.
23964 (aarch64_expand_vec_series): New function.
23965 (aarch64_expand_sve_widened_duplicate): Likewise.
23966 (aarch64_expand_sve_const_vector): Likewise.
23967 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
23968 Handle SVE constants. Use emit_move_insn to move a force_const_mem
23969 into the register, rather than emitting a SET directly.
23970 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
23971 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
23972 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
23973 (offset_9bit_signed_scaled_p): New functions.
23974 (aarch64_replicate_bitmask_imm): New function.
23975 (aarch64_bitmask_imm): Use it.
23976 (aarch64_cannot_force_const_mem): Reject expressions involving
23977 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
23978 (aarch64_classify_index): Handle SVE indices, by requiring
23979 a plain register index with a scale that matches the element size.
23980 (aarch64_classify_address): Handle SVE addresses. Assert that
23981 the mode of the address is VOIDmode or an integer mode.
23982 Update call to aarch64_classify_symbol.
23983 (aarch64_classify_symbolic_expression): Update call to
23984 aarch64_classify_symbol.
23985 (aarch64_const_vec_all_in_range_p): New function.
23986 (aarch64_print_vector_float_operand): Likewise.
23987 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
23988 "vN" for FP registers with SVE modes. Handle (const ...) vectors
23989 and the FP immediates 1.0 and 0.5.
23990 (aarch64_print_address_internal): Handle SVE addresses.
23991 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
23992 (aarch64_regno_regclass): Handle predicate registers.
23993 (aarch64_secondary_reload): Handle big-endian reloads of SVE
23995 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
23996 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
23997 (aarch64_convert_sve_vector_bits): New function.
23998 (aarch64_override_options): Use it to handle -msve-vector-bits=.
23999 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24000 rather than an rtx.
24001 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24002 Handle SVE vector and predicate modes. Accept VL-based constants
24003 that need only one temporary register, and VL offsets that require
24004 no temporary registers.
24005 (aarch64_conditional_register_usage): Mark the predicate registers
24006 as fixed if SVE isn't available.
24007 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24008 Return true for SVE vector and predicate modes.
24009 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24010 rather than an unsigned int. Handle SVE modes.
24011 (aarch64_preferred_simd_mode): Update call accordingly. Handle
24013 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24015 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24016 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24017 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24018 (aarch64_sve_float_mul_immediate_p): New functions.
24019 (aarch64_sve_valid_immediate): New function.
24020 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24021 Explicitly reject structure modes. Check for INDEX constants.
24022 Handle PTRUE and PFALSE constants.
24023 (aarch64_check_zero_based_sve_index_immediate): New function.
24024 (aarch64_simd_imm_zero_p): Delete.
24025 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24026 vector modes. Accept constants in the range of CNT[BHWD].
24027 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24028 ask for an Advanced SIMD mode.
24029 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24030 (aarch64_simd_vector_alignment): Handle SVE predicates.
24031 (aarch64_vectorize_preferred_vector_alignment): New function.
24032 (aarch64_simd_vector_alignment_reachable): Use it instead of
24034 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24035 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24037 (MAX_VECT_LEN): Delete.
24038 (expand_vec_perm_d): Add a vec_flags field.
24039 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24040 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24041 (aarch64_evpc_ext): Don't apply a big-endian lane correction
24043 (aarch64_evpc_rev): Rename to...
24044 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
24045 (aarch64_evpc_rev_global): New function.
24046 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24047 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24049 (aarch64_evpc_sve_tbl): New function.
24050 (aarch64_expand_vec_perm_const_1): Update after rename of
24051 aarch64_evpc_rev. Handle SVE permutes too, trying
24052 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24053 than aarch64_evpc_tbl.
24054 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24055 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24056 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24057 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24058 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24059 (aarch64_expand_sve_vcond): New functions.
24060 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24061 of aarch64_vector_mode_p.
24062 (aarch64_dwarf_poly_indeterminate_value): New function.
24063 (aarch64_compute_pressure_classes): Likewise.
24064 (aarch64_can_change_mode_class): Likewise.
24065 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24066 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24067 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24068 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24069 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24070 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24071 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24072 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24074 (Dn, Dl, Dr): Accept const as well as const_vector.
24075 (Dz): Likewise. Compare against CONST0_RTX.
24076 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24077 of "vector" where appropriate.
24078 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24079 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24080 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24081 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24082 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24083 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24084 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24085 (v_int_equiv): Extend to SVE modes.
24086 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24088 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24089 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24090 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24091 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24092 (SVE_COND_FP_CMP): New int iterators.
24093 (perm_hilo): Handle the new unpack unspecs.
24094 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24096 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24097 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24098 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24099 (aarch64_equality_operator, aarch64_constant_vector_operand)
24100 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24101 (aarch64_sve_nonimmediate_operand): Likewise.
24102 (aarch64_sve_general_operand): Likewise.
24103 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24104 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24105 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24106 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24107 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24108 (aarch64_sve_float_arith_immediate): Likewise.
24109 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24110 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24111 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24112 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24113 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24114 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24115 (aarch64_sve_float_arith_operand): Likewise.
24116 (aarch64_sve_float_arith_with_sub_operand): Likewise.
24117 (aarch64_sve_float_mul_operand): Likewise.
24118 (aarch64_sve_vec_perm_operand): Likewise.
24119 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24120 (aarch64_mov_operand): Accept const_poly_int and const_vector.
24121 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24122 as well as const_vector.
24123 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24124 in file. Use CONST0_RTX and CONSTM1_RTX.
24125 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
24126 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24127 Use aarch64_simd_imm_zero.
24128 * config/aarch64/aarch64-sve.md: New file.
24129 * config/aarch64/aarch64.md: Include it.
24130 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24131 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24132 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24133 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24134 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24135 (sve): New attribute.
24136 (enabled): Disable instructions with the sve attribute unless
24138 (movqi, movhi): Pass CONST_POLY_INT operaneds through
24139 aarch64_expand_mov_immediate.
24140 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24141 CNT[BHSD] immediates.
24142 (movti): Split CONST_POLY_INT moves into two halves.
24143 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24144 Split additions that need a temporary here if the destination
24145 is the stack pointer.
24146 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24147 (*add<mode>3_poly_1): New instruction.
24148 (set_clobber_cc): New expander.
24150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24152 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24153 parameter and use it instead of GET_MODE_SIZE (innermode). Use
24154 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24155 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24156 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
24157 Change innermode from fixed_mode_size to machine_mode.
24158 (simplify_subreg): Update call accordingly. Handle a constant-sized
24159 subreg of a variable-length CONST_VECTOR.
24161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24162 Alan Hayward <alan.hayward@arm.com>
24163 David Sherwood <david.sherwood@arm.com>
24165 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24166 (add_offset_to_base): New function, split out from...
24167 (create_mem_ref): ...here. When handling a scale other than 1,
24168 check first whether the address is valid without the offset.
24169 Add it into the base if so, leaving the index and scale as-is.
24171 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24174 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24175 fold_for_warn before checking if arg2 is INTEGER_CST.
24177 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
24179 * config/rs6000/predicates.md (load_multiple_operation): Delete.
24180 (store_multiple_operation): Delete.
24181 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24182 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24183 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24184 guarded by TARGET_STRING.
24185 (rs6000_output_load_multiple): Delete.
24186 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24187 OPTION_MASK_STRING / TARGET_STRING handling.
24188 (print_operand) <'N', 'O'>: Add comment that these are unused now.
24189 (const rs6000_opt_masks) <"string">: Change mask to 0.
24190 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24191 (MASK_STRING): Delete.
24192 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24194 (load_multiple): Delete.
24201 (store_multiple): Delete.
24208 (movmemsi_8reg): Delete.
24209 (corresponding unnamed define_insn): Delete.
24210 (movmemsi_6reg): Delete.
24211 (corresponding unnamed define_insn): Delete.
24212 (movmemsi_4reg): Delete.
24213 (corresponding unnamed define_insn): Delete.
24214 (movmemsi_2reg): Delete.
24215 (corresponding unnamed define_insn): Delete.
24216 (movmemsi_1reg): Delete.
24217 (corresponding unnamed define_insn): Delete.
24218 * config/rs6000/rs6000.opt (mno-string): New.
24219 (mstring): Replace by deprecation warning stub.
24220 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24222 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24224 * regrename.c (regrename_do_replace): If replacing the same
24225 reg multiple times, try to reuse last created gen_raw_REG.
24228 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24229 main to workaround a bug in GDB.
24231 2018-01-12 Tom de Vries <tom@codesourcery.com>
24234 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24236 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
24238 PR rtl-optimization/80481
24239 * ira-color.c (get_cap_member): New function.
24240 (allocnos_conflict_by_live_ranges_p): Use it.
24241 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24242 (setup_slot_coalesced_allocno_live_ranges): Ditto.
24244 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
24247 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24248 (*saddl_se_1): Ditto.
24249 (*ssubsi_1): Ditto.
24250 (*ssubl_se_1): Ditto.
24252 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24254 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24255 rather than wi::to_widest for DR_INITs.
24256 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24257 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24258 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24260 (vect_analyze_group_access_1): Note that here.
24262 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24264 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24265 polynomial type sizes.
24267 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24269 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24270 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24271 (gimple_add_tmp_var): Likewise.
24273 2018-01-12 Martin Liska <mliska@suse.cz>
24275 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24276 (gimple_alloc_sizes): Likewise.
24277 (dump_gimple_statistics): Use PRIu64 in printf format.
24278 * gimple.h: Change uint64_t to int.
24280 2018-01-12 Martin Liska <mliska@suse.cz>
24282 * tree-core.h: Use uint64_t instead of int.
24283 * tree.c (tree_node_counts): Likewise.
24284 (tree_node_sizes): Likewise.
24285 (dump_tree_statistics): Use PRIu64 in printf format.
24287 2018-01-12 Martin Liska <mliska@suse.cz>
24289 * Makefile.in: As qsort_chk is implemented in vec.c, add
24290 vec.o to linkage of gencfn-macros.
24291 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24292 passing the info to record_node_allocation_statistics.
24293 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24295 * ggc-common.c (struct ggc_usage): Add operator== and use
24296 it in operator< and compare function.
24297 * mem-stats.h (struct mem_usage): Likewise.
24298 * vec.c (struct vec_usage): Remove operator< and compare
24299 function. Can be simply inherited.
24301 2018-01-12 Martin Jambor <mjambor@suse.cz>
24304 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24305 * tree-ssa-math-opts.c: Include domwalk.h.
24306 (convert_mult_to_fma_1): New function.
24307 (fma_transformation_info): New type.
24308 (fma_deferring_state): Likewise.
24309 (cancel_fma_deferring): New function.
24310 (result_of_phi): Likewise.
24311 (last_fma_candidate_feeds_initial_phi): Likewise.
24312 (convert_mult_to_fma): Added deferring logic, split actual
24313 transformation to convert_mult_to_fma_1.
24314 (math_opts_dom_walker): New type.
24315 (math_opts_dom_walker::after_dom_children): New method, body moved
24316 here from pass_optimize_widening_mul::execute, added deferring logic
24318 (pass_optimize_widening_mul::execute): Moved most of code to
24319 math_opts_dom_walker::after_dom_children.
24320 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24321 * config/i386/i386.c (ix86_option_override_internal): Added
24322 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24324 2018-01-12 Richard Biener <rguenther@suse.de>
24327 * dwarf2out.c (gen_variable_die): Do not reset old_die for
24328 inline instance vars.
24330 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
24333 * config/rx/rx.c (rx_is_restricted_memory_address):
24334 Handle SUBREG case.
24336 2018-01-12 Richard Biener <rguenther@suse.de>
24338 PR tree-optimization/80846
24339 * target.def (split_reduction): New target hook.
24340 * targhooks.c (default_split_reduction): New function.
24341 * targhooks.h (default_split_reduction): Declare.
24342 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
24343 target requests first reduce vectors by combining low and high
24345 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
24346 (get_vectype_for_scalar_type_and_size): Export.
24347 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
24348 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
24349 * doc/tm.texi: Regenerate.
24350 * config/i386/i386.c (ix86_split_reduction): Implement
24351 TARGET_VECTORIZE_SPLIT_REDUCTION.
24353 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24356 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
24357 in PIC mode except for TARGET_VXWORKS_RTP.
24358 * config/sparc/sparc.c: Include cfgrtl.h.
24359 (TARGET_INIT_PIC_REG): Define.
24360 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
24361 (sparc_pic_register_p): New predicate.
24362 (sparc_legitimate_address_p): Use it.
24363 (sparc_legitimize_pic_address): Likewise.
24364 (sparc_delegitimize_address): Likewise.
24365 (sparc_mode_dependent_address_p): Likewise.
24366 (gen_load_pcrel_sym): Remove 4th parameter.
24367 (load_got_register): Adjust call to above. Remove obsolete stuff.
24368 (sparc_expand_prologue): Do not call load_got_register here.
24369 (sparc_flat_expand_prologue): Likewise.
24370 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
24371 (sparc_use_pseudo_pic_reg): New function.
24372 (sparc_init_pic_reg): Likewise.
24373 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24374 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
24376 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24378 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
24379 Add item for branch_cost.
24381 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24383 PR rtl-optimization/83565
24384 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
24385 not extend the result to a larger mode for rotate operations.
24386 (num_sign_bit_copies1): Likewise.
24388 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24391 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
24393 Use values-Xc.o for -pedantic.
24394 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
24396 2018-01-12 Martin Liska <mliska@suse.cz>
24399 * ipa-devirt.c (final_warning_record::grow_type_warnings):
24401 (possible_polymorphic_call_targets): Use it.
24402 (ipa_devirt): Likewise.
24404 2018-01-12 Martin Liska <mliska@suse.cz>
24406 * profile-count.h (enum profile_quality): Use 0 as invalid
24407 enum value of profile_quality.
24409 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
24411 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
24412 -mext-string options.
24414 2018-01-12 Richard Biener <rguenther@suse.de>
24416 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
24417 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
24418 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
24420 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
24422 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
24424 * configure.ac (--with-long-double-format): Add support for the
24425 configuration option to change the default long double format on
24427 * config.gcc (powerpc*-linux*-*): Likewise.
24428 * configure: Regenerate.
24429 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
24430 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
24431 used without modification.
24433 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24435 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
24436 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
24437 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
24438 MISC_BUILTIN_SPEC_BARRIER.
24439 (rs6000_init_builtins): Likewise.
24440 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
24442 (speculation_barrier): New define_insn.
24443 * doc/extend.texi: Document __builtin_speculation_barrier.
24445 2018-01-11 Jakub Jelinek <jakub@redhat.com>
24448 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
24449 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
24450 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
24452 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
24453 integral modes instead of "ss" and "sd".
24454 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
24455 vectors with 32-bit and 64-bit elements.
24456 (vecdupssescalarmodesuffix): New mode attribute.
24457 (vec_dup<mode>): Use it.
24459 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
24462 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
24463 frame if argument is passed on stack.
24465 2018-01-11 Jakub Jelinek <jakub@redhat.com>
24468 * ree.c (combine_reaching_defs): Optimize also
24469 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
24470 reg2=any_extend(exp); reg1=reg2;, formatting fix.
24472 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
24474 PR middle-end/83189
24475 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
24477 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
24479 PR middle-end/83718
24480 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
24481 after they are computed.
24483 2018-01-11 Bin Cheng <bin.cheng@arm.com>
24485 PR tree-optimization/83695
24486 * gimple-loop-linterchange.cc
24487 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
24488 reset cached scev information after interchange.
24489 (pass_linterchange::execute): Remove call to scev_reset_htab.
24491 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24493 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
24494 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
24495 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
24496 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
24497 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
24498 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
24499 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
24500 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
24501 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
24502 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
24503 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
24504 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
24505 (V_lane_reg): Likewise.
24506 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
24508 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
24509 (vfmal_lane_low<mode>_intrinsic,
24510 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
24511 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
24512 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
24513 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
24514 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
24515 vfmsl_lane_high<mode>_intrinsic): New define_insns.
24517 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24519 * config/arm/arm-cpus.in (fp16fml): New feature.
24520 (ALL_SIMD): Add fp16fml.
24521 (armv8.2-a): Add fp16fml as an option.
24522 (armv8.3-a): Likewise.
24523 (armv8.4-a): Add fp16fml as part of fp16.
24524 * config/arm/arm.h (TARGET_FP16FML): Define.
24525 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
24527 * config/arm/arm-modes.def (V2HF): Define.
24528 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
24529 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
24530 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
24531 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
24532 vfmsl_low, vfmsl_high): New set of builtins.
24533 * config/arm/iterators.md (PLUSMINUS): New code iterator.
24534 (vfml_op): New code attribute.
24535 (VFMLHALVES): New int iterator.
24536 (VFML, VFMLSEL): New mode attributes.
24537 (V_reg): Define mapping for V2HF.
24538 (V_hi, V_lo): New mode attributes.
24539 (VF_constraint): Likewise.
24540 (vfml_half, vfml_half_selector): New int attributes.
24541 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
24543 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
24544 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
24546 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
24547 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
24548 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
24549 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
24551 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
24552 Document new effective target and option set.
24554 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24556 * config/arm/arm-cpus.in (armv8_4): New feature.
24557 (ARMv8_4a): New fgroup.
24558 (armv8.4-a): New arch.
24559 * config/arm/arm-tables.opt: Regenerate.
24560 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
24561 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
24562 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
24563 Add matching rules for -march=armv8.4-a and extensions.
24564 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
24566 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
24569 * config/rx/rx.md (BW): New mode attribute.
24570 (sync_lock_test_and_setsi): Add mode suffix to insn output.
24572 2018-01-11 Richard Biener <rguenther@suse.de>
24574 PR tree-optimization/83435
24575 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
24576 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
24577 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
24579 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24580 Alan Hayward <alan.hayward@arm.com>
24581 David Sherwood <david.sherwood@arm.com>
24583 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
24585 (aarch64_classify_address): Initialize it. Track polynomial offsets.
24586 (aarch64_print_address_internal): Use it to check for a zero offset.
24588 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24589 Alan Hayward <alan.hayward@arm.com>
24590 David Sherwood <david.sherwood@arm.com>
24592 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
24593 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
24594 Return a poly_int64 rather than a HOST_WIDE_INT.
24595 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
24596 rather than a HOST_WIDE_INT.
24597 * config/aarch64/aarch64.h (aarch64_frame): Protect with
24598 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
24599 hard_fp_offset, frame_size, initial_adjust, callee_offset and
24600 final_offset from HOST_WIDE_INT to poly_int64.
24601 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
24602 to_constant when getting the number of units in an Advanced SIMD
24604 (aarch64_builtin_vectorized_function): Check for a constant number
24606 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
24608 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
24609 attribute instead of GET_MODE_NUNITS.
24610 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
24611 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
24612 GET_MODE_SIZE for fixed-size registers.
24613 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
24614 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
24615 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
24616 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
24617 (aarch64_print_operand, aarch64_print_address_internal)
24618 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
24619 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
24620 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
24621 Handle polynomial GET_MODE_SIZE.
24622 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
24623 wider than SImode without modification.
24624 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
24625 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
24626 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
24627 passing and returning SVE modes.
24628 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
24629 rather than GEN_INT.
24630 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
24631 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
24632 (aarch64_allocate_and_probe_stack_space): Likewise.
24633 (aarch64_layout_frame): Cope with polynomial offsets.
24634 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
24635 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
24636 polynomial offsets.
24637 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
24638 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
24639 poly_int64 rather than a HOST_WIDE_INT.
24640 (aarch64_get_separate_components, aarch64_process_components)
24641 (aarch64_expand_prologue, aarch64_expand_epilogue)
24642 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
24643 (aarch64_anchor_offset): New function, split out from...
24644 (aarch64_legitimize_address): ...here.
24645 (aarch64_builtin_vectorization_cost): Handle polynomial
24646 TYPE_VECTOR_SUBPARTS.
24647 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
24649 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
24650 number of elements from the PARALLEL rather than the mode.
24651 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
24652 rather than GET_MODE_BITSIZE.
24653 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
24654 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
24655 (aarch64_expand_vec_perm_const_1): Handle polynomial
24656 d->perm.length () and d->perm elements.
24657 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
24658 Apply to_constant to d->perm elements.
24659 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
24660 polynomial CONST_VECTOR_NUNITS.
24661 (aarch64_move_pointer): Take amount as a poly_int64 rather
24663 (aarch64_progress_pointer): Avoid temporary variable.
24664 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
24665 the mode attribute instead of GET_MODE.
24667 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24668 Alan Hayward <alan.hayward@arm.com>
24669 David Sherwood <david.sherwood@arm.com>
24671 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
24672 x exists before using it.
24673 (aarch64_add_constant_internal): Rename to...
24674 (aarch64_add_offset_1): ...this. Replace regnum with separate
24675 src and dest rtxes. Handle the case in which they're different,
24676 including when the offset is zero. Replace scratchreg with an rtx.
24677 Use 2 additions if there is no spare register into which we can
24678 move a 16-bit constant.
24679 (aarch64_add_constant): Delete.
24680 (aarch64_add_offset): Replace reg with separate src and dest
24681 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
24682 Use aarch64_add_offset_1.
24683 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
24684 an rtx rather than an int. Take the delta as a poly_int64
24685 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
24686 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
24687 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
24688 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
24689 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
24690 and aarch64_add_sp.
24691 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
24692 aarch64_add_constant.
24694 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24696 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
24697 Use scalar_float_mode.
24699 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24701 * config/aarch64/aarch64-simd.md
24702 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
24703 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
24704 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
24705 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
24706 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
24707 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
24708 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
24709 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
24710 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
24711 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
24713 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24716 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
24717 targ_options->x_arm_arch_string is non NULL.
24719 2018-01-11 Tamar Christina <tamar.christina@arm.com>
24721 * config/aarch64/aarch64.h
24722 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
24724 2018-01-11 Sudakshina Das <sudi.das@arm.com>
24727 * expmed.c (emit_store_flag_force): Swap if const op0
24728 and change VOIDmode to mode of op0.
24730 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24732 PR rtl-optimization/83761
24733 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
24734 than bytes to mode_for_size.
24736 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24738 PR middle-end/83189
24739 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
24740 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
24743 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24745 PR middle-end/83575
24746 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
24747 when in layout mode.
24748 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
24749 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
24752 2018-01-10 Michael Collison <michael.collison@arm.com>
24754 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
24755 * config/aarch64/aarch64-option-extension.def: Add
24756 AARCH64_OPT_EXTENSION of 'fp16fml'.
24757 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24758 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
24759 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
24760 * config/aarch64/constraints.md (Ui7): New constraint.
24761 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
24762 (VFMLA_SEL_W): Ditto.
24765 (VFMLA16_LOW): New int iterator.
24766 (VFMLA16_HIGH): Ditto.
24767 (UNSPEC_FMLAL): New unspec.
24768 (UNSPEC_FMLSL): Ditto.
24769 (UNSPEC_FMLAL2): Ditto.
24770 (UNSPEC_FMLSL2): Ditto.
24771 (f16mac): New code attribute.
24772 * config/aarch64/aarch64-simd-builtins.def
24773 (aarch64_fmlal_lowv2sf): Ditto.
24774 (aarch64_fmlsl_lowv2sf): Ditto.
24775 (aarch64_fmlalq_lowv4sf): Ditto.
24776 (aarch64_fmlslq_lowv4sf): Ditto.
24777 (aarch64_fmlal_highv2sf): Ditto.
24778 (aarch64_fmlsl_highv2sf): Ditto.
24779 (aarch64_fmlalq_highv4sf): Ditto.
24780 (aarch64_fmlslq_highv4sf): Ditto.
24781 (aarch64_fmlal_lane_lowv2sf): Ditto.
24782 (aarch64_fmlsl_lane_lowv2sf): Ditto.
24783 (aarch64_fmlal_laneq_lowv2sf): Ditto.
24784 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
24785 (aarch64_fmlalq_lane_lowv4sf): Ditto.
24786 (aarch64_fmlsl_lane_lowv4sf): Ditto.
24787 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
24788 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
24789 (aarch64_fmlal_lane_highv2sf): Ditto.
24790 (aarch64_fmlsl_lane_highv2sf): Ditto.
24791 (aarch64_fmlal_laneq_highv2sf): Ditto.
24792 (aarch64_fmlsl_laneq_highv2sf): Ditto.
24793 (aarch64_fmlalq_lane_highv4sf): Ditto.
24794 (aarch64_fmlsl_lane_highv4sf): Ditto.
24795 (aarch64_fmlalq_laneq_highv4sf): Ditto.
24796 (aarch64_fmlsl_laneq_highv4sf): Ditto.
24797 * config/aarch64/aarch64-simd.md:
24798 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
24799 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24800 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
24801 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24802 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
24803 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
24804 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
24805 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
24806 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
24807 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
24808 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
24809 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
24810 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
24811 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
24812 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
24813 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
24814 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
24815 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
24816 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
24817 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
24818 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
24819 (vfmlsl_low_u32): Ditto.
24820 (vfmlalq_low_u32): Ditto.
24821 (vfmlslq_low_u32): Ditto.
24822 (vfmlal_high_u32): Ditto.
24823 (vfmlsl_high_u32): Ditto.
24824 (vfmlalq_high_u32): Ditto.
24825 (vfmlslq_high_u32): Ditto.
24826 (vfmlal_lane_low_u32): Ditto.
24827 (vfmlsl_lane_low_u32): Ditto.
24828 (vfmlal_laneq_low_u32): Ditto.
24829 (vfmlsl_laneq_low_u32): Ditto.
24830 (vfmlalq_lane_low_u32): Ditto.
24831 (vfmlslq_lane_low_u32): Ditto.
24832 (vfmlalq_laneq_low_u32): Ditto.
24833 (vfmlslq_laneq_low_u32): Ditto.
24834 (vfmlal_lane_high_u32): Ditto.
24835 (vfmlsl_lane_high_u32): Ditto.
24836 (vfmlal_laneq_high_u32): Ditto.
24837 (vfmlsl_laneq_high_u32): Ditto.
24838 (vfmlalq_lane_high_u32): Ditto.
24839 (vfmlslq_lane_high_u32): Ditto.
24840 (vfmlalq_laneq_high_u32): Ditto.
24841 (vfmlslq_laneq_high_u32): Ditto.
24842 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
24843 (AARCH64_FL_FOR_ARCH8_4): New.
24844 (AARCH64_ISA_F16FML): New ISA flag.
24845 (TARGET_F16FML): New feature flag for fp16fml.
24846 (doc/invoke.texi): Document new fp16fml option.
24848 2018-01-10 Michael Collison <michael.collison@arm.com>
24850 * config/aarch64/aarch64-builtins.c:
24851 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
24852 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24853 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
24854 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
24855 (AARCH64_ISA_SHA3): New ISA flag.
24856 (TARGET_SHA3): New feature flag for sha3.
24857 * config/aarch64/iterators.md (sha512_op): New int attribute.
24858 (CRYPTO_SHA512): New int iterator.
24859 (UNSPEC_SHA512H): New unspec.
24860 (UNSPEC_SHA512H2): Ditto.
24861 (UNSPEC_SHA512SU0): Ditto.
24862 (UNSPEC_SHA512SU1): Ditto.
24863 * config/aarch64/aarch64-simd-builtins.def
24864 (aarch64_crypto_sha512hqv2di): New builtin.
24865 (aarch64_crypto_sha512h2qv2di): Ditto.
24866 (aarch64_crypto_sha512su0qv2di): Ditto.
24867 (aarch64_crypto_sha512su1qv2di): Ditto.
24868 (aarch64_eor3qv8hi): Ditto.
24869 (aarch64_rax1qv2di): Ditto.
24870 (aarch64_xarqv2di): Ditto.
24871 (aarch64_bcaxqv8hi): Ditto.
24872 * config/aarch64/aarch64-simd.md:
24873 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
24874 (aarch64_crypto_sha512su0qv2di): Ditto.
24875 (aarch64_crypto_sha512su1qv2di): Ditto.
24876 (aarch64_eor3qv8hi): Ditto.
24877 (aarch64_rax1qv2di): Ditto.
24878 (aarch64_xarqv2di): Ditto.
24879 (aarch64_bcaxqv8hi): Ditto.
24880 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
24881 (vsha512h2q_u64): Ditto.
24882 (vsha512su0q_u64): Ditto.
24883 (vsha512su1q_u64): Ditto.
24884 (veor3q_u16): Ditto.
24885 (vrax1q_u64): Ditto.
24886 (vxarq_u64): Ditto.
24887 (vbcaxq_u16): Ditto.
24888 * config/arm/types.md (crypto_sha512): New type attribute.
24889 (crypto_sha3): Ditto.
24890 (doc/invoke.texi): Document new sha3 option.
24892 2018-01-10 Michael Collison <michael.collison@arm.com>
24894 * config/aarch64/aarch64-builtins.c:
24895 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
24896 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24897 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
24898 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
24899 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
24900 (AARCH64_ISA_SM4): New ISA flag.
24901 (TARGET_SM4): New feature flag for sm4.
24902 * config/aarch64/aarch64-simd-builtins.def
24903 (aarch64_sm3ss1qv4si): Ditto.
24904 (aarch64_sm3tt1aq4si): Ditto.
24905 (aarch64_sm3tt1bq4si): Ditto.
24906 (aarch64_sm3tt2aq4si): Ditto.
24907 (aarch64_sm3tt2bq4si): Ditto.
24908 (aarch64_sm3partw1qv4si): Ditto.
24909 (aarch64_sm3partw2qv4si): Ditto.
24910 (aarch64_sm4eqv4si): Ditto.
24911 (aarch64_sm4ekeyqv4si): Ditto.
24912 * config/aarch64/aarch64-simd.md:
24913 (aarch64_sm3ss1qv4si): Ditto.
24914 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
24915 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
24916 (aarch64_sm4eqv4si): Ditto.
24917 (aarch64_sm4ekeyqv4si): Ditto.
24918 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
24919 (sm3part_op): Ditto.
24920 (CRYPTO_SM3TT): Ditto.
24921 (CRYPTO_SM3PART): Ditto.
24922 (UNSPEC_SM3SS1): New unspec.
24923 (UNSPEC_SM3TT1A): Ditto.
24924 (UNSPEC_SM3TT1B): Ditto.
24925 (UNSPEC_SM3TT2A): Ditto.
24926 (UNSPEC_SM3TT2B): Ditto.
24927 (UNSPEC_SM3PARTW1): Ditto.
24928 (UNSPEC_SM3PARTW2): Ditto.
24929 (UNSPEC_SM4E): Ditto.
24930 (UNSPEC_SM4EKEY): Ditto.
24931 * config/aarch64/constraints.md (Ui2): New constraint.
24932 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
24933 * config/arm/types.md (crypto_sm3): New type attribute.
24934 (crypto_sm4): Ditto.
24935 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
24936 (vsm3tt1aq_u32): Ditto.
24937 (vsm3tt1bq_u32): Ditto.
24938 (vsm3tt2aq_u32): Ditto.
24939 (vsm3tt2bq_u32): Ditto.
24940 (vsm3partw1q_u32): Ditto.
24941 (vsm3partw2q_u32): Ditto.
24942 (vsm4eq_u32): Ditto.
24943 (vsm4ekeyq_u32): Ditto.
24944 (doc/invoke.texi): Document new sm4 option.
24946 2018-01-10 Michael Collison <michael.collison@arm.com>
24948 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
24949 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
24950 (AARCH64_FL_FOR_ARCH8_4): New.
24951 (AARCH64_FL_V8_4): New flag.
24952 (doc/invoke.texi): Document new armv8.4-a option.
24954 2018-01-10 Michael Collison <michael.collison@arm.com>
24956 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24957 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
24958 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
24959 * config/aarch64/aarch64-option-extension.def: Add
24960 AARCH64_OPT_EXTENSION of 'sha2'.
24961 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
24962 (crypto): Disable sha2 and aes if crypto disabled.
24963 (crypto): Enable aes and sha2 if enabled.
24964 (simd): Disable sha2 and aes if simd disabled.
24965 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
24967 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
24968 (TARGET_SHA2): New feature flag for sha2.
24969 (TARGET_AES): New feature flag for aes.
24970 * config/aarch64/aarch64-simd.md:
24971 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
24972 conditional on TARGET_AES.
24973 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
24974 (aarch64_crypto_sha1hsi): Make pattern conditional
24976 (aarch64_crypto_sha1hv4si): Ditto.
24977 (aarch64_be_crypto_sha1hv4si): Ditto.
24978 (aarch64_crypto_sha1su1v4si): Ditto.
24979 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
24980 (aarch64_crypto_sha1su0v4si): Ditto.
24981 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
24982 (aarch64_crypto_sha256su0v4si): Ditto.
24983 (aarch64_crypto_sha256su1v4si): Ditto.
24984 (doc/invoke.texi): Document new aes and sha2 options.
24986 2018-01-10 Martin Sebor <msebor@redhat.com>
24988 PR tree-optimization/83781
24989 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
24992 2018-01-11 Martin Sebor <msebor@gmail.com>
24993 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24995 PR tree-optimization/83501
24996 PR tree-optimization/81703
24998 * tree-ssa-strlen.c (get_string_cst): Rename...
24999 (get_string_len): ...to this. Handle global constants.
25000 (handle_char_store): Adjust.
25002 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
25003 Jim Wilson <jimw@sifive.com>
25005 * config/riscv/riscv-protos.h (riscv_output_return): New.
25006 * config/riscv/riscv.c (struct machine_function): New naked_p field.
25007 (riscv_attribute_table, riscv_output_return),
25008 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25009 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25010 (riscv_compute_frame_info): Only compute frame->mask if not a naked
25012 (riscv_expand_prologue): Add early return for naked function.
25013 (riscv_expand_epilogue): Likewise.
25014 (riscv_function_ok_for_sibcall): Return false for naked function.
25015 (riscv_set_current_function): New.
25016 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25017 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25018 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25019 * doc/extend.texi (RISC-V Function Attributes): New.
25021 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
25023 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25024 check for 128-bit long double before checking TCmode.
25025 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25026 128-bit long doubles before checking TFmode or TCmode.
25027 (FLOAT128_IBM_P): Likewise.
25029 2018-01-10 Martin Sebor <msebor@redhat.com>
25031 PR tree-optimization/83671
25032 * builtins.c (c_strlen): Unconditionally return zero for the empty
25034 Use -Warray-bounds for warnings.
25035 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25036 for non-constant array indices with COMPONENT_REF, arrays of
25037 arrays, and pointers to arrays.
25038 (gimple_fold_builtin_strlen): Determine and set length range for
25039 non-constant character arrays.
25041 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
25043 PR middle-end/81897
25044 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25047 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
25049 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25051 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25054 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25055 VECTOR_MEM_ALTIVEC_OR_VSX_P.
25056 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25057 indexed_or_indirect_operand predicate.
25058 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25059 (*vsx_le_perm_load_v8hi): Likewise.
25060 (*vsx_le_perm_load_v16qi): Likewise.
25061 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25062 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25063 (*vsx_le_perm_store_v8hi): Likewise.
25064 (*vsx_le_perm_store_v16qi): Likewise.
25065 (eight unnamed splitters): Likewise.
25067 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25069 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25070 * config/rs6000/emmintrin.h: Likewise.
25071 * config/rs6000/mmintrin.h: Likewise.
25072 * config/rs6000/xmmintrin.h: Likewise.
25074 2018-01-10 David Malcolm <dmalcolm@redhat.com>
25077 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25079 * tree.c (tree_nop_conversion): Return true for location wrapper
25081 (maybe_wrap_with_location): New function.
25082 (selftest::check_strip_nops): New function.
25083 (selftest::test_location_wrappers): New function.
25084 (selftest::tree_c_tests): Call it.
25085 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25086 (maybe_wrap_with_location): New decl.
25087 (EXPR_LOCATION_WRAPPER_P): New macro.
25088 (location_wrapper_p): New inline function.
25089 (tree_strip_any_location_wrapper): New inline function.
25091 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
25094 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25095 stack_realign_offset for the largest alignment of stack slot
25097 (ix86_find_max_used_stack_alignment): New function.
25098 (ix86_finalize_stack_frame_flags): Use it. Set
25099 max_used_stack_alignment if we don't realign stack.
25100 * config/i386/i386.h (machine_function): Add
25101 max_used_stack_alignment.
25103 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
25105 * config/arm/arm.opt (-mbranch-cost): New option.
25106 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25109 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
25112 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25113 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25115 2018-01-10 Richard Biener <rguenther@suse.de>
25118 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25119 early out so it also covers the case where we have a non-NULL
25122 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25124 PR tree-optimization/83753
25125 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25126 for non-strided grouped accesses if the number of elements is 1.
25128 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25131 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25132 * i386.h (TARGET_USE_GATHER): Define.
25133 * x86-tune.def (X86_TUNE_USE_GATHER): New.
25135 2018-01-10 Martin Liska <mliska@suse.cz>
25138 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25139 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25141 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25142 CLEANUP_NO_PARTITIONING is not set.
25144 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25146 * doc/rtl.texi: Remove documentation of (const ...) wrappers
25147 for vectors, as a partial revert of r254296.
25148 * rtl.h (const_vec_p): Delete.
25149 (const_vec_duplicate_p): Don't test for vector CONSTs.
25150 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25151 * expmed.c (make_tree): Likewise.
25154 * common.md (E, F): Use CONSTANT_P instead of checking for
25156 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25157 checking for CONST_VECTOR.
25159 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25161 PR middle-end/83575
25162 * predict.c (force_edge_cold): Handle in more sane way edges
25163 with no prediction.
25165 2018-01-09 Carl Love <cel@us.ibm.com>
25167 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25169 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25170 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25171 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25172 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
25173 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25174 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
25175 * config/rs6000/rs6000-protos.h: Add extern defition for
25176 rs6000_generate_float2_double_code.
25177 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25179 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25180 (float2_v2df): Add define_expand.
25182 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
25185 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25186 op_mode in the force_to_mode call.
25188 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25190 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25191 instead of checking each element individually.
25192 (aarch64_evpc_uzp): Likewise.
25193 (aarch64_evpc_zip): Likewise.
25194 (aarch64_evpc_ext): Likewise.
25195 (aarch64_evpc_rev): Likewise.
25196 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25197 instead of checking each element individually. Return true without
25199 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25200 whether all selected elements come from the same input, instead of
25201 checking each element individually. Remove calls to gen_rtx_REG,
25202 start_sequence and end_sequence and instead assert that no rtl is
25205 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25207 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25208 order of HIGH and CONST checks.
25210 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25212 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25213 if the destination isn't an SSA_NAME.
25215 2018-01-09 Richard Biener <rguenther@suse.de>
25217 PR tree-optimization/83668
25218 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25220 (canonicalize_loop_form): ... here, renamed from ...
25221 (canonicalize_loop_closed_ssa_form): ... this and amended to
25222 swap successor edges for loop exit blocks to make us use
25223 the RPO order we need for initial schedule generation.
25225 2018-01-09 Joseph Myers <joseph@codesourcery.com>
25227 PR tree-optimization/64811
25228 * match.pd: When optimizing comparisons with Inf, avoid
25229 introducing or losing exceptions from comparisons with NaN.
25231 2018-01-09 Martin Liska <mliska@suse.cz>
25234 * asan.c (shadow_mem_size): Add gcc_assert.
25236 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
25238 Don't save registers in main().
25241 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25242 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25243 * config/avr/avr.c (avr_set_current_function): Don't error if
25244 naked, OS_task or OS_main are specified at the same time.
25245 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25247 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25249 * common/config/avr/avr-common.c (avr_option_optimization_table):
25250 Switch on -mmain-is-OS_task for optimizing compilations.
25252 2018-01-09 Richard Biener <rguenther@suse.de>
25254 PR tree-optimization/83572
25255 * graphite.c: Include cfganal.h.
25256 (graphite_transform_loops): Connect infinite loops to exit
25257 and remove fake edges at the end.
25259 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25261 * ipa-inline.c (edge_badness): Revert accidental checkin.
25263 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25266 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25267 symbols; not inline clones.
25269 2018-01-09 Jakub Jelinek <jakub@redhat.com>
25272 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25273 hard registers. Formatting fixes.
25275 PR preprocessor/83722
25276 * gcc.c (try_generate_repro): Pass
25277 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25278 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25281 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
25282 Kito Cheng <kito.cheng@gmail.com>
25284 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25285 (riscv_leaf_function_p): Delete.
25286 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25288 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25290 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25292 (do_ifelse): New function.
25293 (do_isel): New function.
25294 (do_sub3): New function.
25295 (do_add3): New function.
25296 (do_load_mask_compare): New function.
25297 (do_overlap_load_compare): New function.
25298 (expand_compare_loop): New function.
25299 (expand_block_compare): Call expand_compare_loop() when appropriate.
25300 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25301 option description.
25302 (-mblock-compare-inline-loop-limit): New option.
25304 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25307 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25308 Reverse order of second and third operands in first alternative.
25309 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25310 of first and second elements in UNSPEC_VPERMR vector.
25311 (altivec_expand_vec_perm_le): Likewise.
25313 2018-01-08 Jeff Law <law@redhat.com>
25315 PR rtl-optimizatin/81308
25316 * tree-switch-conversion.c (cfg_altered): New file scoped static.
25317 (process_switch): If group_case_labels makes a change, then set
25319 (pass_convert_switch::execute): If a switch is converted, then
25320 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
25322 PR rtl-optimization/81308
25323 * recog.c (split_all_insns): Conditionally cleanup the CFG after
25326 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
25328 PR target/83663 - Revert r255946
25329 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25330 generation for cases where splatting a value is not useful.
25331 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25332 across a vec_duplicate and a paradoxical subreg forming a vector
25333 mode to a vec_concat.
25335 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25337 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
25338 -march=armv8.3-a variants.
25339 * config/arm/t-multilib: Likewise.
25340 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
25342 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25344 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
25346 (cceq_ior_compare_complement): Give it a name so I can use it, and
25347 change boolean_or_operator predicate to boolean_operator so it can
25348 be used to generate a crand.
25349 (eqne): New code iterator.
25350 (bd/bd_neg): New code_attrs.
25351 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
25352 a single define_insn.
25353 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
25354 decrement (bdnzt/bdnzf/bdzt/bdzf).
25355 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
25356 with the new names of the branch decrement patterns, and added the
25357 names of the branch decrement conditional patterns.
25359 2018-01-08 Richard Biener <rguenther@suse.de>
25361 PR tree-optimization/83563
25362 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
25365 2018-01-08 Richard Biener <rguenther@suse.de>
25367 PR middle-end/83713
25368 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
25370 2018-01-08 Richard Biener <rguenther@suse.de>
25372 PR tree-optimization/83685
25373 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
25374 references to abnormals.
25376 2018-01-08 Richard Biener <rguenther@suse.de>
25379 * dwarf2out.c (output_indirect_strings): Handle empty
25380 skeleton_debug_str_hash.
25381 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
25383 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
25385 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
25386 (emit_store_direct): Likewise.
25387 (arc_trampoline_adjust_address): Likewise.
25388 (arc_asm_trampoline_template): New function.
25389 (arc_initialize_trampoline): Use asm_trampoline_template.
25390 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
25391 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
25392 * config/arc/arc.md (flush_icache): Delete pattern.
25394 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
25396 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
25397 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
25400 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
25403 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
25404 by not USED_FOR_TARGET.
25405 (make_pass_resolve_sw_modes): Likewise.
25407 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
25409 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
25412 2018-01-08 Richard Biener <rguenther@suse.de>
25414 PR middle-end/83580
25415 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
25417 2018-01-08 Richard Biener <rguenther@suse.de>
25419 PR middle-end/83517
25420 * match.pd ((t * 2) / 2) -> t): Add missing :c.
25422 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
25424 PR middle-end/81897
25425 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
25426 basic blocks with a small number of successors.
25427 (convert_control_dep_chain_into_preds): Improve handling of
25429 (dump_predicates): Split apart into...
25430 (dump_pred_chain): ...here...
25431 (dump_pred_info): ...and here.
25432 (can_one_predicate_be_invalidated_p): Add debugging printfs.
25433 (can_chain_union_be_invalidated_p): Improve check for invalidation
25435 (uninit_uses_cannot_happen): Avoid unnecessary if
25436 convert_control_dep_chain_into_preds yielded nothing.
25438 2018-01-06 Martin Sebor <msebor@redhat.com>
25440 PR tree-optimization/83640
25441 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
25442 subtracting negative offset from size.
25443 (builtin_access::overlap): Adjust offset bounds of the access to fall
25444 within the size of the object if possible.
25446 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
25448 PR rtl-optimization/83699
25449 * expmed.c (extract_bit_field_1): Restrict the vector usage of
25450 extract_bit_field_as_subreg to cases in which the extracted
25451 value is also a vector.
25453 * lra-constraints.c (process_alt_operands): Test for the equivalence
25454 substitutions when detecting a possible reload cycle.
25456 2018-01-06 Jakub Jelinek <jakub@redhat.com>
25459 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
25460 by default if flag_selective_schedling{,2}. Formatting fixes.
25462 PR rtl-optimization/83682
25463 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
25464 if it has non-VECTOR_MODE element mode.
25465 (vec_duplicate_p): Likewise.
25467 PR middle-end/83694
25468 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
25469 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
25471 2018-01-05 Jakub Jelinek <jakub@redhat.com>
25474 * config/i386/i386-builtin.def
25475 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
25476 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
25477 Require also OPTION_MASK_ISA_AVX512F in addition to
25478 OPTION_MASK_ISA_GFNI.
25479 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
25480 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
25481 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
25482 to OPTION_MASK_ISA_GFNI.
25483 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
25484 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
25485 OPTION_MASK_ISA_AVX512BW.
25486 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
25487 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
25488 addition to OPTION_MASK_ISA_GFNI.
25489 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
25490 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
25491 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
25492 to OPTION_MASK_ISA_GFNI.
25493 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
25494 a requirement for all ISAs rather than any of them with a few
25496 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
25498 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
25499 bitmasks to be enabled with 3 exceptions, instead of requiring any
25500 enabled ISA with lots of exceptions.
25501 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
25502 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
25503 Change avx512bw in isa attribute to avx512f.
25504 * config/i386/sgxintrin.h: Add license boilerplate.
25505 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
25506 to __AVX512F__ and __AVX512VL to __AVX512VL__.
25507 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
25508 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
25510 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
25511 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
25512 temporarily sse2 rather than sse if not enabled already.
25515 * config/i386/sse.md (VI248_VLBW): Rename to ...
25516 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
25517 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
25518 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
25519 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
25520 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
25521 mode iterator instead of VI248_VLBW.
25523 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
25525 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
25526 (record_modified): Skip clobbers; add debug output.
25527 (param_change_prob): Use sreal frequencies.
25529 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
25531 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25532 punt for user-aligned variables.
25534 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
25536 * tree-chrec.c (chrec_contains_symbols): Return true for
25539 2018-01-05 Sudakshina Das <sudi.das@arm.com>
25542 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
25543 of (x|y) == x for BICS pattern.
25545 2018-01-05 Jakub Jelinek <jakub@redhat.com>
25547 PR tree-optimization/83605
25548 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
25549 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
25552 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
25554 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
25555 * config/epiphany/rtems.h: New file.
25557 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25558 Uros Bizjak <ubizjak@gmail.com>
25561 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
25562 QIreg_operand instead of register_operand predicate.
25563 * config/i386/i386.c (ix86_rop_should_change_byte_p,
25564 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
25565 comments instead of -fmitigate[-_]rop.
25567 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25570 * cgraphunit.c (symbol_table::compile): Switch to text_section
25571 before calling assembly_start debug hook.
25572 * run-rtl-passes.c (run_rtl_passes): Likewise.
25575 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25577 * tree-vrp.c (extract_range_from_binary_expr_1): Check
25578 range_int_cst_p rather than !symbolic_range_p before calling
25579 extract_range_from_multiplicative_op_1.
25581 2018-01-04 Jeff Law <law@redhat.com>
25583 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
25584 redundant test in assertion.
25586 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25588 * doc/rtl.texi: Document machine_mode wrapper classes.
25590 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25592 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
25593 using tree_to_uhwi.
25595 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25597 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
25598 the VEC_PERM_EXPR fold to fail.
25600 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25603 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
25604 to switched_sections.
25606 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25609 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
25610 test for d.testing.
25612 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
25615 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
25616 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
25618 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25621 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
25622 is BLKmode and bitpos not zero or mode change is needed.
25624 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25627 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
25630 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
25633 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
25634 instead of MULT rtx. Update all corresponding splitters.
25635 (*saddl_se): Ditto.
25636 (*ssub<modesuffix>): Ditto.
25637 (*ssubl_se): Ditto.
25638 (*cmp_sadd_di): Update split patterns.
25639 (*cmp_sadd_si): Ditto.
25640 (*cmp_sadd_sidi): Ditto.
25641 (*cmp_ssub_di): Ditto.
25642 (*cmp_ssub_si): Ditto.
25643 (*cmp_ssub_sidi): Ditto.
25644 * config/alpha/predicates.md (const23_operand): New predicate.
25645 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
25646 Look for ASHIFT, not MULT inner operand.
25647 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
25649 2018-01-04 Martin Liska <mliska@suse.cz>
25651 PR gcov-profile/83669
25652 * gcov.c (output_intermediate_file): Add version to intermediate
25654 * doc/gcov.texi: Document new field 'version' in intermediate
25655 file format. Fix location of '-k' option of gcov command.
25657 2018-01-04 Martin Liska <mliska@suse.cz>
25660 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
25662 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25664 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
25666 2018-01-03 Martin Sebor <msebor@redhat.com>
25668 PR tree-optimization/83655
25669 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
25670 checking calls with invalid arguments.
25672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25674 * tree-vect-stmts.c (vect_get_store_rhs): New function.
25675 (vectorizable_mask_load_store): Delete.
25676 (vectorizable_call): Return false for masked loads and stores.
25677 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
25678 instead of gimple_assign_rhs1.
25679 (vectorizable_load): Handle IFN_MASK_LOAD.
25680 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
25682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25684 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
25686 (vectorizable_mask_load_store): ...here.
25687 (vectorizable_load): ...and here.
25689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25691 * tree-vect-stmts.c (vect_build_all_ones_mask)
25692 (vect_build_zero_merge_argument): New functions, split out from...
25693 (vectorizable_load): ...here.
25695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25697 * tree-vect-stmts.c (vect_check_store_rhs): New function,
25699 (vectorizable_mask_load_store): ...here.
25700 (vectorizable_store): ...and here.
25702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25704 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
25706 (vectorizable_mask_load_store): ...here.
25708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25710 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
25711 (vect_model_store_cost): Take a vec_load_store_type instead of a
25713 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
25714 (vect_model_store_cost): Take a vec_load_store_type instead of a
25716 (vectorizable_mask_load_store): Update accordingly.
25717 (vectorizable_store): Likewise.
25718 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
25720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25722 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
25723 IFN_MASK_LOAD calls here rather than...
25724 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
25726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25727 Alan Hayward <alan.hayward@arm.com>
25728 David Sherwood <david.sherwood@arm.com>
25730 * expmed.c (extract_bit_field_1): For vector extracts,
25731 fall back to extract_bit_field_as_subreg if vec_extract
25734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25735 Alan Hayward <alan.hayward@arm.com>
25736 David Sherwood <david.sherwood@arm.com>
25738 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
25739 they are variable or constant sized.
25740 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
25741 slots for constant-sized data.
25743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25744 Alan Hayward <alan.hayward@arm.com>
25745 David Sherwood <david.sherwood@arm.com>
25747 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
25748 handling COND_EXPRs with boolean comparisons, try to find a better
25749 basis for the mask type than the boolean itself.
25751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25753 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
25754 is calculated and how it can be overridden.
25755 * genmodes.c (max_bitsize_mode_any_mode): New variable.
25756 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
25758 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
25761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25762 Alan Hayward <alan.hayward@arm.com>
25763 David Sherwood <david.sherwood@arm.com>
25765 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
25766 Remove the mode argument.
25767 (aarch64_simd_valid_immediate): Remove the mode and inverse
25769 * config/aarch64/iterators.md (bitsize): New iterator.
25770 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
25771 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
25772 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
25773 aarch64_simd_valid_immediate.
25774 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
25775 (aarch64_reg_or_bic_imm): Likewise.
25776 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
25777 with an insn_type enum and msl with a modifier_type enum.
25778 Replace element_width with a scalar_mode. Change the shift
25779 to unsigned int. Add constructors for scalar_float_mode and
25780 scalar_int_mode elements.
25781 (aarch64_vect_float_const_representable_p): Delete.
25782 (aarch64_can_const_movi_rtx_p)
25783 (aarch64_simd_scalar_immediate_valid_for_move)
25784 (aarch64_simd_make_constant): Update call to
25785 aarch64_simd_valid_immediate.
25786 (aarch64_advsimd_valid_immediate_hs): New function.
25787 (aarch64_advsimd_valid_immediate): Likewise.
25788 (aarch64_simd_valid_immediate): Remove mode and inverse
25789 arguments. Rewrite to use the above. Use const_vec_duplicate_p
25790 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
25791 and aarch64_float_const_representable_p on the result.
25792 (aarch64_output_simd_mov_immediate): Remove mode argument.
25793 Update call to aarch64_simd_valid_immediate and use of
25794 simd_immediate_info.
25795 (aarch64_output_scalar_simd_mov_immediate): Update call
25798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25799 Alan Hayward <alan.hayward@arm.com>
25800 David Sherwood <david.sherwood@arm.com>
25802 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
25803 (mode_nunits): Likewise CONST_MODE_NUNITS.
25804 * machmode.def (ADJUST_NUNITS): Document.
25805 * genmodes.c (mode_data::need_nunits_adj): New field.
25806 (blank_mode): Update accordingly.
25807 (adj_nunits): New variable.
25808 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
25810 (emit_mode_size_inline): Set need_bytesize_adj for all modes
25811 listed in adj_nunits.
25812 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
25813 listed in adj_nunits. Don't emit case statements for such modes.
25814 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
25815 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
25816 nothing if adj_nunits is nonnull.
25817 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
25818 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
25819 (emit_mode_fbit): Update use of print_maybe_const_decl.
25820 (emit_move_size): Likewise. Treat the array as non-const
25822 (emit_mode_adjustments): Handle adj_nunits.
25824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25826 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
25827 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
25828 (VECTOR_MODES): Use it.
25829 (make_vector_modes): Take the prefix as an argument.
25831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25832 Alan Hayward <alan.hayward@arm.com>
25833 David Sherwood <david.sherwood@arm.com>
25835 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
25836 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
25837 for MODE_VECTOR_BOOL.
25838 * machmode.def (VECTOR_BOOL_MODE): Document.
25839 * genmodes.c (VECTOR_BOOL_MODE): New macro.
25840 (make_vector_bool_mode): New function.
25841 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
25843 * lto-streamer-in.c (lto_input_mode_table): Likewise.
25844 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
25846 * stor-layout.c (int_mode_for_mode): Likewise.
25847 * tree.c (build_vector_type_for_mode): Likewise.
25848 * varasm.c (output_constant_pool_2): Likewise.
25849 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
25850 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
25851 for MODE_VECTOR_BOOL.
25852 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
25853 of mode class checks.
25854 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
25855 instead of a list of mode class checks.
25856 (expand_vector_scalar_condition): Likewise.
25857 (type_for_widest_vector_mode): Handle BImode as an inner mode.
25859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25860 Alan Hayward <alan.hayward@arm.com>
25861 David Sherwood <david.sherwood@arm.com>
25863 * machmode.h (mode_size): Change from unsigned short to
25865 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
25866 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25867 or if measurement_type is not polynomial.
25868 (fixed_size_mode::includes_p): Check for constant-sized modes.
25869 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
25870 return a poly_uint16 rather than an unsigned short.
25871 (emit_mode_size): Change the type of mode_size from unsigned short
25872 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
25873 (emit_mode_adjustments): Cope with polynomial vector sizes.
25874 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25876 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25878 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
25879 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
25880 * caller-save.c (setup_save_areas): Likewise.
25881 (replace_reg_with_saved_mem): Likewise.
25882 * calls.c (emit_library_call_value_1): Likewise.
25883 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
25884 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
25885 (gen_lowpart_for_combine): Likewise.
25886 * convert.c (convert_to_integer_1): Likewise.
25887 * cse.c (equiv_constant, cse_insn): Likewise.
25888 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
25889 (cselib_subst_to_values): Likewise.
25890 * dce.c (word_dce_process_block): Likewise.
25891 * df-problems.c (df_word_lr_mark_ref): Likewise.
25892 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
25893 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
25894 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
25895 (rtl_for_decl_location): Likewise.
25896 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
25897 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
25898 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
25899 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
25900 (expand_expr_real_1): Likewise.
25901 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
25902 (pad_below): Likewise.
25903 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25904 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
25905 * ira.c (get_subreg_tracking_sizes): Likewise.
25906 * ira-build.c (ira_create_allocno_objects): Likewise.
25907 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
25908 (ira_sort_regnos_for_alter_reg): Likewise.
25909 * ira-costs.c (record_operand_costs): Likewise.
25910 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
25911 (resolve_simple_move): Likewise.
25912 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
25913 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
25914 (lra_constraints): Likewise.
25915 (CONST_POOL_OK_P): Reject variable-sized modes.
25916 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
25917 (add_pseudo_to_slot, lra_spill): Likewise.
25918 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25919 * optabs-query.c (get_best_extraction_insn): Likewise.
25920 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25921 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
25922 (expand_mult_highpart, valid_multiword_target_p): Likewise.
25923 * recog.c (offsettable_address_addr_space_p): Likewise.
25924 * regcprop.c (maybe_mode_change): Likewise.
25925 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
25926 * regrename.c (build_def_use): Likewise.
25927 * regstat.c (dump_reg_info): Likewise.
25928 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
25929 (find_reloads, find_reloads_subreg_address): Likewise.
25930 * reload1.c (eliminate_regs_1): Likewise.
25931 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
25932 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
25933 (simplify_binary_operation_1, simplify_subreg): Likewise.
25934 * targhooks.c (default_function_arg_padding): Likewise.
25935 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
25936 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
25937 (verify_gimple_assign_ternary): Likewise.
25938 * tree-inline.c (estimate_move_cost): Likewise.
25939 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25940 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
25941 (get_address_cost_ainc): Likewise.
25942 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
25943 (vect_supportable_dr_alignment): Likewise.
25944 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25945 (vectorizable_reduction): Likewise.
25946 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
25947 (vectorizable_operation, vectorizable_load): Likewise.
25948 * tree.c (build_same_sized_truth_vector_type): Likewise.
25949 * valtrack.c (cleanup_auto_inc_dec): Likewise.
25950 * var-tracking.c (emit_note_insn_var_location): Likewise.
25951 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
25952 (ADDR_VEC_ALIGN): Likewise.
25954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25955 Alan Hayward <alan.hayward@arm.com>
25956 David Sherwood <david.sherwood@arm.com>
25958 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
25960 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25961 or if measurement_type is polynomial.
25962 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
25963 * combine.c (make_extraction): Likewise.
25964 * dse.c (find_shift_sequence): Likewise.
25965 * dwarf2out.c (mem_loc_descriptor): Likewise.
25966 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
25967 (extract_bit_field, extract_low_bits): Likewise.
25968 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
25969 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
25970 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
25971 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
25972 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25973 * reload.c (find_reloads): Likewise.
25974 * reload1.c (alter_reg): Likewise.
25975 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
25976 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
25977 * tree-if-conv.c (predicate_mem_writes): Likewise.
25978 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
25979 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
25980 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
25981 * valtrack.c (dead_debug_insert_temp): Likewise.
25982 * varasm.c (mergeable_constant_section): Likewise.
25983 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
25985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25986 Alan Hayward <alan.hayward@arm.com>
25987 David Sherwood <david.sherwood@arm.com>
25989 * expr.c (expand_assignment): Cope with polynomial mode sizes
25990 when assigning to a CONCAT.
25992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25993 Alan Hayward <alan.hayward@arm.com>
25994 David Sherwood <david.sherwood@arm.com>
25996 * machmode.h (mode_precision): Change from unsigned short to
25998 (mode_to_precision): Return a poly_uint16 rather than an unsigned
26000 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26001 or if measurement_type is not polynomial.
26002 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
26003 in which the mode is already known to be a scalar_int_mode.
26004 * genmodes.c (emit_mode_precision): Change the type of mode_precision
26005 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
26007 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26008 for GET_MODE_PRECISION.
26009 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26010 for GET_MODE_PRECISION.
26011 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26013 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26014 (expand_field_assignment, make_extraction): Likewise.
26015 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26016 (get_last_value): Likewise.
26017 * convert.c (convert_to_integer_1): Likewise.
26018 * cse.c (cse_insn): Likewise.
26019 * expr.c (expand_expr_real_1): Likewise.
26020 * lra-constraints.c (simplify_operand_subreg): Likewise.
26021 * optabs-query.c (can_atomic_load_p): Likewise.
26022 * optabs.c (expand_atomic_load): Likewise.
26023 (expand_atomic_store): Likewise.
26024 * ree.c (combine_reaching_defs): Likewise.
26025 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26026 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26027 * tree.h (type_has_mode_precision_p): Likewise.
26028 * ubsan.c (instrument_si_overflow): Likewise.
26030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26031 Alan Hayward <alan.hayward@arm.com>
26032 David Sherwood <david.sherwood@arm.com>
26034 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26035 polynomial numbers of units.
26036 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26037 (valid_vector_subparts_p): New function.
26038 (build_vector_type): Remove temporary shim and take the number
26039 of units as a poly_uint64 rather than an int.
26040 (build_opaque_vector_type): Take the number of units as a
26041 poly_uint64 rather than an int.
26042 * tree.c (build_vector_from_ctor): Handle polynomial
26043 TYPE_VECTOR_SUBPARTS.
26044 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26045 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26046 (build_vector_from_val): If the number of units is variable,
26047 use build_vec_duplicate_cst for constant operands and
26048 VEC_DUPLICATE_EXPR otherwise.
26049 (make_vector_type): Remove temporary is_constant ().
26050 (build_vector_type, build_opaque_vector_type): Take the number of
26051 units as a poly_uint64 rather than an int.
26052 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26054 * cfgexpand.c (expand_debug_expr): Likewise.
26055 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26056 (store_constructor, expand_expr_real_1): Likewise.
26057 (const_scalar_mask_from_tree): Likewise.
26058 * fold-const-call.c (fold_const_reduction): Likewise.
26059 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26060 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26061 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26062 (fold_relational_const): Likewise.
26063 (native_interpret_vector): Likewise. Change the size from an
26064 int to an unsigned int.
26065 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26066 TYPE_VECTOR_SUBPARTS.
26067 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26068 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26069 duplicating a non-constant operand into a variable-length vector.
26070 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26071 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26072 * ipa-icf.c (sem_variable::equals): Likewise.
26073 * match.pd: Likewise.
26074 * omp-simd-clone.c (simd_clone_subparts): Likewise.
26075 * print-tree.c (print_node): Likewise.
26076 * stor-layout.c (layout_type): Likewise.
26077 * targhooks.c (default_builtin_vectorization_cost): Likewise.
26078 * tree-cfg.c (verify_gimple_comparison): Likewise.
26079 (verify_gimple_assign_binary): Likewise.
26080 (verify_gimple_assign_ternary): Likewise.
26081 (verify_gimple_assign_single): Likewise.
26082 * tree-pretty-print.c (dump_generic_node): Likewise.
26083 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26084 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26085 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26086 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26087 (vect_shift_permute_load_chain): Likewise.
26088 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26089 (expand_vector_condition, optimize_vector_constructor): Likewise.
26090 (lower_vec_perm, get_compute_type): Likewise.
26091 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26092 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26093 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26094 (vect_recog_mask_conversion_pattern): Likewise.
26095 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26096 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26097 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26098 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26099 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26100 (vectorizable_shift, vectorizable_operation, vectorizable_store)
26101 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26102 (supportable_widening_operation): Likewise.
26103 (supportable_narrowing_operation): Likewise.
26104 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26106 * varasm.c (output_constant): Likewise.
26108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26109 Alan Hayward <alan.hayward@arm.com>
26110 David Sherwood <david.sherwood@arm.com>
26112 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26113 so that both the length == 3 and length != 3 cases set up their
26114 own permute vectors. Add comments explaining why we know the
26115 number of elements is constant.
26116 (vect_permute_load_chain): Likewise.
26118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26119 Alan Hayward <alan.hayward@arm.com>
26120 David Sherwood <david.sherwood@arm.com>
26122 * machmode.h (mode_nunits): Change from unsigned char to
26124 (ONLY_FIXED_SIZE_MODES): New macro.
26125 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26126 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26127 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26129 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26130 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26131 or if measurement_type is not polynomial.
26132 * genmodes.c (ZERO_COEFFS): New macro.
26133 (emit_mode_nunits_inline): Make mode_nunits_inline return a
26135 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26136 Use ZERO_COEFFS when emitting initializers.
26137 * data-streamer.h (bp_pack_poly_value): New function.
26138 (bp_unpack_poly_value): Likewise.
26139 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26140 for GET_MODE_NUNITS.
26141 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26142 for GET_MODE_NUNITS.
26143 * tree.c (make_vector_type): Remove temporary shim and make
26144 the real function take the number of units as a poly_uint64
26145 rather than an int.
26146 (build_vector_type_for_mode): Handle polynomial nunits.
26147 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26148 * emit-rtl.c (const_vec_series_p_1): Likewise.
26149 (gen_rtx_CONST_VECTOR): Likewise.
26150 * fold-const.c (test_vec_duplicate_folding): Likewise.
26151 * genrecog.c (validate_pattern): Likewise.
26152 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26153 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26154 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26155 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26156 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26157 * rtlanal.c (subreg_get_info): Likewise.
26158 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26159 (vect_grouped_load_supported): Likewise.
26160 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26161 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26162 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26163 (simplify_const_unary_operation, simplify_binary_operation_1)
26164 (simplify_const_binary_operation, simplify_ternary_operation)
26165 (test_vector_ops_duplicate, test_vector_ops): Likewise.
26166 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26167 instead of CONST_VECTOR_NUNITS.
26168 * varasm.c (output_constant_pool_2): Likewise.
26169 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26170 explicit-encoded elements in the XVEC for variable-length vectors.
26172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26174 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26177 Alan Hayward <alan.hayward@arm.com>
26178 David Sherwood <david.sherwood@arm.com>
26180 * coretypes.h (fixed_size_mode): Declare.
26181 (fixed_size_mode_pod): New typedef.
26182 * builtins.h (target_builtins::x_apply_args_mode)
26183 (target_builtins::x_apply_result_mode): Change type to
26184 fixed_size_mode_pod.
26185 * builtins.c (apply_args_size, apply_result_size, result_vector)
26186 (expand_builtin_apply_args_1, expand_builtin_apply)
26187 (expand_builtin_return): Update accordingly.
26189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26191 * cse.c (hash_rtx_cb): Hash only the encoded elements.
26192 * cselib.c (cselib_hash_rtx): Likewise.
26193 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26194 CONST_VECTOR encoding.
26196 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26197 Jeff Law <law@redhat.com>
26200 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26201 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26202 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26203 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26206 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26207 explicitly probe *sp in a noreturn function if there were any callee
26208 register saves or frame pointer is needed.
26210 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26213 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26214 BLKmode for ternary, binary or unary expressions.
26217 * var-tracking.c (delete_vta_debug_insn): New inline function.
26218 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26219 insns from get_insns () to NULL instead of each bb separately.
26220 Use delete_vta_debug_insn. No longer static.
26221 (vt_debug_insns_local, variable_tracking_main_1): Adjust
26222 delete_vta_debug_insns callers.
26223 * rtl.h (delete_vta_debug_insns): Declare.
26224 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26225 instead of variable_tracking_main.
26227 2018-01-03 Martin Sebor <msebor@redhat.com>
26229 PR tree-optimization/83603
26230 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26231 arguments past the endof the argument list in functions declared
26232 without a prototype.
26233 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26234 Avoid checking when arguments are null.
26236 2018-01-03 Martin Sebor <msebor@redhat.com>
26239 * doc/extend.texi (attribute const): Fix a typo.
26240 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26241 issuing -Wsuggest-attribute for void functions.
26243 2018-01-03 Martin Sebor <msebor@redhat.com>
26245 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26246 offset_int::from instead of wide_int::to_shwi.
26247 (maybe_diag_overlap): Remove assertion.
26248 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26249 * gimple-ssa-sprintf.c (format_directive): Same.
26250 (parse_directive): Same.
26251 (sprintf_dom_walker::compute_format_length): Same.
26252 (try_substitute_return_value): Same.
26254 2018-01-03 Jeff Law <law@redhat.com>
26256 PR middle-end/83654
26257 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26258 non-constant residual for zero at runtime and avoid probing in
26259 that case. Reorganize code for trailing problem to mirror handling
26262 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26264 PR tree-optimization/83501
26265 * tree-ssa-strlen.c (get_string_cst): New.
26266 (handle_char_store): Call get_string_cst.
26268 2018-01-03 Martin Liska <mliska@suse.cz>
26270 PR tree-optimization/83593
26271 * tree-ssa-strlen.c: Include tree-cfg.h.
26272 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26273 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26274 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26276 (strlen_dom_walker::before_dom_children): Call
26277 gimple_purge_dead_eh_edges. Dump tranformation with details
26279 (strlen_dom_walker::before_dom_children): Update call by adding
26280 new argument cleanup_eh.
26281 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26283 2018-01-03 Martin Liska <mliska@suse.cz>
26286 * cif-code.def (VARIADIC_THUNK): New enum value.
26287 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26290 2018-01-03 Jan Beulich <jbeulich@suse.com>
26292 * sse.md (mov<mode>_internal): Tighten condition for when to use
26293 vmovdqu<ssescalarsize> for TI and OI modes.
26295 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26297 Update copyright years.
26299 2018-01-03 Martin Liska <mliska@suse.cz>
26302 * ipa-visibility.c (function_and_variable_visibility): Skip
26303 functions with noipa attribure.
26305 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26307 * gcc.c (process_command): Update copyright notice dates.
26308 * gcov-dump.c (print_version): Ditto.
26309 * gcov.c (print_version): Ditto.
26310 * gcov-tool.c (print_version): Ditto.
26311 * gengtype.c (create_file): Ditto.
26312 * doc/cpp.texi: Bump @copying's copyright year.
26313 * doc/cppinternals.texi: Ditto.
26314 * doc/gcc.texi: Ditto.
26315 * doc/gccint.texi: Ditto.
26316 * doc/gcov.texi: Ditto.
26317 * doc/install.texi: Ditto.
26318 * doc/invoke.texi: Ditto.
26320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26322 * vector-builder.h (vector_builder::m_full_nelts): Change from
26323 unsigned int to poly_uint64.
26324 (vector_builder::full_nelts): Update prototype accordingly.
26325 (vector_builder::new_vector): Likewise.
26326 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26327 (vector_builder::operator ==): Likewise.
26328 (vector_builder::finalize): Likewise.
26329 * int-vector-builder.h (int_vector_builder::int_vector_builder):
26330 Take the number of elements as a poly_uint64 rather than an
26332 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26333 from unsigned int to poly_uint64.
26334 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
26335 (vec_perm_indices::new_vector): Likewise.
26336 (vec_perm_indices::length): Likewise.
26337 (vec_perm_indices::nelts_per_input): Likewise.
26338 (vec_perm_indices::input_nelts): Likewise.
26339 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
26340 number of elements per input as a poly_uint64 rather than an
26341 unsigned int. Use the original encoding for variable-length
26342 vectors, rather than clamping each individual element.
26343 For the second and subsequent elements in each pattern,
26344 clamp the step and base before clamping their sum.
26345 (vec_perm_indices::series_p): Handle polynomial element counts.
26346 (vec_perm_indices::all_in_range_p): Likewise.
26347 (vec_perm_indices_to_tree): Likewise.
26348 (vec_perm_indices_to_rtx): Likewise.
26349 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
26350 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
26351 (tree_vector_builder::new_binary_operation): Handle polynomial
26352 element counts. Return false if we need to know the number
26353 of elements at compile time.
26354 * fold-const.c (fold_vec_perm): Punt if the number of elements
26355 isn't known at compile time.
26357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26359 * vec-perm-indices.h (vec_perm_builder): Change element type
26360 from HOST_WIDE_INT to poly_int64.
26361 (vec_perm_indices::element_type): Update accordingly.
26362 (vec_perm_indices::clamp): Handle polynomial element_types.
26363 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26364 (vec_perm_indices::all_in_range_p): Likewise.
26365 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
26367 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
26368 polynomial vec_perm_indices element types.
26369 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
26370 * fold-const.c (fold_vec_perm): Likewise.
26371 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
26372 * tree-vect-generic.c (lower_vec_perm): Likewise.
26373 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26374 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
26375 element type to HOST_WIDE_INT.
26377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26378 Alan Hayward <alan.hayward@arm.com>
26379 David Sherwood <david.sherwood@arm.com>
26381 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
26382 rather than an int. Use plus_constant.
26383 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
26384 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
26386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26387 Alan Hayward <alan.hayward@arm.com>
26388 David Sherwood <david.sherwood@arm.com>
26390 * calls.c (emit_call_1, expand_call): Change struct_value_size from
26391 a HOST_WIDE_INT to a poly_int64.
26393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26394 Alan Hayward <alan.hayward@arm.com>
26395 David Sherwood <david.sherwood@arm.com>
26397 * calls.c (load_register_parameters): Cope with polynomial
26398 mode sizes. Require a constant size for BLKmode parameters
26399 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
26400 forces a parameter to be padded at the lsb end in order to
26401 fill a complete number of words, require the parameter size
26402 to be ordered wrt UNITS_PER_WORD.
26404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26405 Alan Hayward <alan.hayward@arm.com>
26406 David Sherwood <david.sherwood@arm.com>
26408 * reload1.c (spill_stack_slot_width): Change element type
26409 from unsigned int to poly_uint64_pod.
26410 (alter_reg): Treat mode sizes as polynomial.
26412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26413 Alan Hayward <alan.hayward@arm.com>
26414 David Sherwood <david.sherwood@arm.com>
26416 * reload.c (complex_word_subreg_p): New function.
26417 (reload_inner_reg_of_subreg, push_reload): Use it.
26419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26420 Alan Hayward <alan.hayward@arm.com>
26421 David Sherwood <david.sherwood@arm.com>
26423 * lra-constraints.c (process_alt_operands): Reject matched
26424 operands whose sizes aren't ordered.
26425 (match_reload): Refer to this check here.
26427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26428 Alan Hayward <alan.hayward@arm.com>
26429 David Sherwood <david.sherwood@arm.com>
26431 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
26432 that the mode size is in the set {1, 2, 4, 8, 16}.
26434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26435 Alan Hayward <alan.hayward@arm.com>
26436 David Sherwood <david.sherwood@arm.com>
26438 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
26439 Use plus_constant instead of gen_rtx_PLUS.
26441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26442 Alan Hayward <alan.hayward@arm.com>
26443 David Sherwood <david.sherwood@arm.com>
26445 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
26446 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
26447 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
26448 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
26449 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
26450 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
26451 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
26452 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
26453 * config/i386/i386.c (ix86_push_rounding): ...this new function.
26454 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
26456 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
26457 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
26458 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
26459 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
26460 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
26461 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
26462 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
26463 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
26464 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
26465 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
26467 * expr.c (emit_move_resolve_push): Treat the input and result
26468 of PUSH_ROUNDING as a poly_int64.
26469 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
26470 (emit_push_insn): Likewise.
26471 * lra-eliminations.c (mark_not_eliminable): Likewise.
26472 * recog.c (push_operand): Likewise.
26473 * reload1.c (elimination_effects): Likewise.
26474 * rtlanal.c (nonzero_bits1): Likewise.
26475 * calls.c (store_one_arg): Likewise. Require the padding to be
26476 known at compile time.
26478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26479 Alan Hayward <alan.hayward@arm.com>
26480 David Sherwood <david.sherwood@arm.com>
26482 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
26483 Use plus_constant instead of gen_rtx_PLUS.
26485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26486 Alan Hayward <alan.hayward@arm.com>
26487 David Sherwood <david.sherwood@arm.com>
26489 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
26490 rather than an int.
26492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26493 Alan Hayward <alan.hayward@arm.com>
26494 David Sherwood <david.sherwood@arm.com>
26496 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
26497 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
26498 via stack temporaries. Treat the mode size as polynomial too.
26500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26501 Alan Hayward <alan.hayward@arm.com>
26502 David Sherwood <david.sherwood@arm.com>
26504 * expr.c (expand_expr_real_2): When handling conversions involving
26505 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
26506 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
26507 as a poly_uint64 too.
26509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26510 Alan Hayward <alan.hayward@arm.com>
26511 David Sherwood <david.sherwood@arm.com>
26513 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
26515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26516 Alan Hayward <alan.hayward@arm.com>
26517 David Sherwood <david.sherwood@arm.com>
26519 * combine.c (can_change_dest_mode): Handle polynomial
26520 REGMODE_NATURAL_SIZE.
26521 * expmed.c (store_bit_field_1): Likewise.
26522 * expr.c (store_constructor): Likewise.
26523 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
26524 and polynomial REGMODE_NATURAL_SIZE.
26525 (gen_lowpart_common): Likewise.
26526 * reginfo.c (record_subregs_of_mode): Likewise.
26527 * rtlanal.c (read_modify_subreg_p): Likewise.
26529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26530 Alan Hayward <alan.hayward@arm.com>
26531 David Sherwood <david.sherwood@arm.com>
26533 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
26534 numbers of elements.
26536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26537 Alan Hayward <alan.hayward@arm.com>
26538 David Sherwood <david.sherwood@arm.com>
26540 * match.pd: Cope with polynomial numbers of vector elements.
26542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26543 Alan Hayward <alan.hayward@arm.com>
26544 David Sherwood <david.sherwood@arm.com>
26546 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
26547 in a POINTER_PLUS_EXPR.
26549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26550 Alan Hayward <alan.hayward@arm.com>
26551 David Sherwood <david.sherwood@arm.com>
26553 * omp-simd-clone.c (simd_clone_subparts): New function.
26554 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
26555 (ipa_simd_modify_function_body): Likewise.
26557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26558 Alan Hayward <alan.hayward@arm.com>
26559 David Sherwood <david.sherwood@arm.com>
26561 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
26562 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
26563 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
26564 (expand_vector_condition, vector_element): Likewise.
26565 (subparts_gt): New function.
26566 (get_compute_type): Use subparts_gt.
26567 (count_type_subparts): Delete.
26568 (expand_vector_operations_1): Use subparts_gt instead of
26569 count_type_subparts.
26571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26572 Alan Hayward <alan.hayward@arm.com>
26573 David Sherwood <david.sherwood@arm.com>
26575 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
26576 (vect_compile_time_alias): ...this new function. Do the calculation
26577 on poly_ints rather than trees.
26578 (vect_prune_runtime_alias_test_list): Update call accordingly.
26580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26581 Alan Hayward <alan.hayward@arm.com>
26582 David Sherwood <david.sherwood@arm.com>
26584 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
26586 (vect_schedule_slp_instance): Likewise.
26588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26589 Alan Hayward <alan.hayward@arm.com>
26590 David Sherwood <david.sherwood@arm.com>
26592 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
26593 constant and extern definitions for variable-length vectors.
26594 (vect_get_constant_vectors): Note that the number of units
26595 is known to be constant.
26597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26598 Alan Hayward <alan.hayward@arm.com>
26599 David Sherwood <david.sherwood@arm.com>
26601 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
26602 of units as polynomial. Choose between WIDE and NARROW based
26605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26606 Alan Hayward <alan.hayward@arm.com>
26607 David Sherwood <david.sherwood@arm.com>
26609 * tree-vect-stmts.c (simd_clone_subparts): New function.
26610 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
26612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26613 Alan Hayward <alan.hayward@arm.com>
26614 David Sherwood <david.sherwood@arm.com>
26616 * tree-vect-stmts.c (vectorizable_call): Treat the number of
26617 vectors as polynomial. Use build_index_vector for
26618 IFN_GOMP_SIMD_LANE.
26620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26621 Alan Hayward <alan.hayward@arm.com>
26622 David Sherwood <david.sherwood@arm.com>
26624 * tree-vect-stmts.c (get_load_store_type): Treat the number of
26625 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
26626 for variable-length vectors.
26627 (vectorizable_mask_load_store): Treat the number of units as
26628 polynomial, asserting that it is constant if the condition has
26629 already been enforced.
26630 (vectorizable_store, vectorizable_load): 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-vect-loop.c (vectorizable_live_operation): Treat the number
26637 of units as polynomial. Punt if we can't tell at compile time
26638 which vector contains the final result.
26640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26641 Alan Hayward <alan.hayward@arm.com>
26642 David Sherwood <david.sherwood@arm.com>
26644 * tree-vect-loop.c (vectorizable_induction): Treat the number
26645 of units as polynomial. Punt on SLP inductions. Use an integer
26646 VEC_SERIES_EXPR for variable-length integer reductions. Use a
26647 cast of such a series for variable-length floating-point
26650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26651 Alan Hayward <alan.hayward@arm.com>
26652 David Sherwood <david.sherwood@arm.com>
26654 * tree.h (build_index_vector): Declare.
26655 * tree.c (build_index_vector): New function.
26656 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
26657 of units as polynomial, forcibly converting it to a constant if
26658 vectorizable_reduction has already enforced the condition.
26659 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
26660 to create a {1,2,3,...} vector.
26661 (vectorizable_reduction): Treat the number of units as polynomial.
26662 Choose vectype_in based on the largest scalar element size rather
26663 than the smallest number of units. Enforce the restrictions
26666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26667 Alan Hayward <alan.hayward@arm.com>
26668 David Sherwood <david.sherwood@arm.com>
26670 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
26671 number of units as polynomial.
26673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26674 Alan Hayward <alan.hayward@arm.com>
26675 David Sherwood <david.sherwood@arm.com>
26677 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
26678 * target.def (autovectorize_vector_sizes): Return the vector sizes
26679 by pointer, using vector_sizes rather than a bitmask.
26680 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
26681 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
26682 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
26684 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
26685 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
26686 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
26687 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
26688 * omp-general.c (omp_max_vf): Likewise.
26689 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26690 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
26691 * tree-vect-loop.c (vect_analyze_loop): Likewise.
26692 * tree-vect-slp.c (vect_slp_bb): Likewise.
26693 * doc/tm.texi: Regenerate.
26694 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
26696 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
26697 the vector size as a poly_uint64 rather than an unsigned int.
26698 (current_vector_size): Change from an unsigned int to a poly_uint64.
26699 (get_vectype_for_scalar_type): Update accordingly.
26700 * tree.h (build_truth_vector_type): Take the size and number of
26701 units as a poly_uint64 rather than an unsigned int.
26702 (build_vector_type): Add a temporary overload that takes
26703 the number of units as a poly_uint64 rather than an unsigned int.
26704 * tree.c (make_vector_type): Likewise.
26705 (build_truth_vector_type): Take the number of units as a poly_uint64
26706 rather than an unsigned int.
26708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26709 Alan Hayward <alan.hayward@arm.com>
26710 David Sherwood <david.sherwood@arm.com>
26712 * target.def (get_mask_mode): Take the number of units and length
26713 as poly_uint64s rather than unsigned ints.
26714 * targhooks.h (default_get_mask_mode): Update accordingly.
26715 * targhooks.c (default_get_mask_mode): Likewise.
26716 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
26717 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
26724 * omp-general.c (omp_max_vf): Likewise.
26725 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
26726 (expand_omp_simd): Handle polynomial safelen.
26727 * omp-low.c (omplow_simd_context): Add a default constructor.
26728 (omplow_simd_context::max_vf): Change from int to poly_uint64.
26729 (lower_rec_simd_input_clauses): Update accordingly.
26730 (lower_rec_input_clauses): Likewise.
26732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26733 Alan Hayward <alan.hayward@arm.com>
26734 David Sherwood <david.sherwood@arm.com>
26736 * tree-vectorizer.h (vect_nunits_for_cost): New function.
26737 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
26738 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
26739 (vect_analyze_slp_cost): Likewise.
26740 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
26741 (vect_model_load_cost): Likewise.
26743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26744 Alan Hayward <alan.hayward@arm.com>
26745 David Sherwood <david.sherwood@arm.com>
26747 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
26748 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
26749 from an unsigned int * to a poly_uint64_pod *.
26750 (calculate_unrolling_factor): New function.
26751 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
26753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26754 Alan Hayward <alan.hayward@arm.com>
26755 David Sherwood <david.sherwood@arm.com>
26757 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
26758 from an unsigned int to a poly_uint64.
26759 (_loop_vec_info::slp_unrolling_factor): Likewise.
26760 (_loop_vec_info::vectorization_factor): Change from an int
26762 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
26763 (vect_get_num_vectors): New function.
26764 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
26765 (vect_get_num_copies): Use vect_get_num_vectors.
26766 (vect_analyze_data_ref_dependences): Change max_vf from an int *
26767 to an unsigned int *.
26768 (vect_analyze_data_refs): Change min_vf from an int * to a
26770 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26771 than an unsigned HOST_WIDE_INT.
26772 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
26773 (vect_analyze_data_ref_dependence): Change max_vf from an int *
26774 to an unsigned int *.
26775 (vect_analyze_data_ref_dependences): Likewise.
26776 (vect_compute_data_ref_alignment): Handle polynomial vf.
26777 (vect_enhance_data_refs_alignment): Likewise.
26778 (vect_prune_runtime_alias_test_list): Likewise.
26779 (vect_shift_permute_load_chain): Likewise.
26780 (vect_supportable_dr_alignment): Likewise.
26781 (dependence_distance_ge_vf): Take the vectorization factor as a
26782 poly_uint64 rather than an unsigned HOST_WIDE_INT.
26783 (vect_analyze_data_refs): Change min_vf from an int * to a
26785 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
26786 vfm1 as a poly_uint64 rather than an int. Make the same change
26787 for the returned bound_scalar.
26788 (vect_gen_vector_loop_niters): Handle polynomial vf.
26789 (vect_do_peeling): Likewise. Update call to
26790 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
26791 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
26793 * tree-vect-loop.c (vect_determine_vectorization_factor)
26794 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
26795 (vect_get_known_peeling_cost): Likewise.
26796 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
26797 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
26798 (vect_transform_loop): Likewise. Use the lowest possible VF when
26799 updating the upper bounds of the loop.
26800 (vect_min_worthwhile_factor): Make static. Return an unsigned int
26801 rather than an int.
26802 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
26803 polynomial unroll factors.
26804 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
26805 (vect_make_slp_decision): Likewise.
26806 (vect_supported_load_permutation_p): Likewise, and polynomial
26808 (vect_analyze_slp_cost): Handle polynomial vf.
26809 (vect_slp_analyze_node_operations): Likewise.
26810 (vect_slp_analyze_bb_1): Likewise.
26811 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26812 than an unsigned HOST_WIDE_INT.
26813 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
26814 (vectorizable_load): Handle polynomial vf.
26815 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
26817 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
26819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26820 Alan Hayward <alan.hayward@arm.com>
26821 David Sherwood <david.sherwood@arm.com>
26823 * match.pd: Handle bit operations involving three constants
26824 and try to fold one pair.
26826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26828 * tree-vect-loop-manip.c: Include gimple-fold.h.
26829 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
26830 niters_maybe_zero parameters. Handle other cases besides a step of 1.
26831 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
26832 Add a path that uses a step of VF instead of 1, but disable it
26834 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
26835 and niters_no_overflow parameters. Update calls to
26836 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
26837 Create a new SSA name if the latter choses to use a ste other
26838 than zero, and return it via niters_vector_mult_vf_var.
26839 * tree-vect-loop.c (vect_transform_loop): Update calls to
26840 vect_do_peeling, vect_gen_vector_loop_niters and
26841 slpeel_make_loop_iterate_ntimes.
26842 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
26843 (vect_gen_vector_loop_niters): Update declarations after above changes.
26845 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
26847 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
26848 128-bit round to integer instructions.
26849 (ceil<mode>2): Likewise.
26850 (btrunc<mode>2): Likewise.
26851 (round<mode>2): Likewise.
26853 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26855 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
26856 unaligned VSX load/store on P8/P9.
26857 (expand_block_clear): Allow the use of unaligned VSX
26858 load/store on P8/P9.
26860 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26862 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
26864 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
26865 swap associated with both a load and a store.
26867 2018-01-02 Andrew Waterman <andrew@sifive.com>
26869 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
26870 * config/riscv/riscv.md (clear_cache): Use it.
26872 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
26874 * web.c: Remove out-of-date comment.
26876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26878 * expr.c (fixup_args_size_notes): Check that any existing
26879 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
26880 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
26881 (emit_single_push_insn): ...here.
26883 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26885 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
26886 (const_vector_encoded_nelts): New function.
26887 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
26888 (const_vector_int_elt, const_vector_elt): Declare.
26889 * emit-rtl.c (const_vector_int_elt_1): New function.
26890 (const_vector_elt): Likewise.
26891 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
26892 of CONST_VECTOR_ELT.
26894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26896 * expr.c: Include rtx-vector-builder.h.
26897 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
26898 directly on the tree encoding.
26899 (const_vector_from_tree): Likewise.
26900 * optabs.c: Include rtx-vector-builder.h.
26901 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
26902 sequence of "u" values.
26903 * vec-perm-indices.c: Include rtx-vector-builder.h.
26904 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
26905 directly on the vec_perm_indices encoding.
26907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26909 * doc/rtl.texi (const_vector): Describe new encoding scheme.
26910 * Makefile.in (OBJS): Add rtx-vector-builder.o.
26911 * rtx-vector-builder.h: New file.
26912 * rtx-vector-builder.c: Likewise.
26913 * rtl.h (rtx_def::u2): Add a const_vector field.
26914 (CONST_VECTOR_NPATTERNS): New macro.
26915 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
26916 (CONST_VECTOR_DUPLICATE_P): Likewise.
26917 (CONST_VECTOR_STEPPED_P): Likewise.
26918 (CONST_VECTOR_ENCODED_ELT): Likewise.
26919 (const_vec_duplicate_p): Check for a duplicated vector encoding.
26920 (unwrap_const_vec_duplicate): Likewise.
26921 (const_vec_series_p): Check for a non-duplicated vector encoding.
26922 Say that the function only returns true for integer vectors.
26923 * emit-rtl.c: Include rtx-vector-builder.h.
26924 (gen_const_vec_duplicate_1): Delete.
26925 (gen_const_vector): Call gen_const_vec_duplicate instead of
26926 gen_const_vec_duplicate_1.
26927 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
26928 (gen_const_vec_duplicate): Use rtx_vector_builder.
26929 (gen_const_vec_series): Likewise.
26930 (gen_rtx_CONST_VECTOR): Likewise.
26931 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
26932 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26933 Build a new vector rather than modifying a CONST_VECTOR in-place.
26934 (handle_special_swappables): Update call accordingly.
26935 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
26936 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26937 Build a new vector rather than modifying a CONST_VECTOR in-place.
26938 (handle_special_swappables): Update call accordingly.
26940 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26942 * simplify-rtx.c (simplify_const_binary_operation): Use
26943 CONST_VECTOR_ELT instead of XVECEXP.
26945 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26947 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
26948 the selector elements to be different from the data elements
26949 if the selector is a VECTOR_CST.
26950 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
26951 ssizetype for the selector.
26953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26955 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
26956 before testing each element individually.
26957 * tree-vect-generic.c (lower_vec_perm): Likewise.
26959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26961 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
26962 * selftest-run-tests.c (selftest::run_tests): Call it.
26963 * vector-builder.h (vector_builder::operator ==): New function.
26964 (vector_builder::operator !=): Likewise.
26965 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
26966 (vec_perm_indices::all_from_input_p): New function.
26967 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26968 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
26969 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
26970 instead of reading the VECTOR_CST directly. Detect whether both
26971 vector inputs are the same before constructing the vec_perm_indices,
26972 and update the number of inputs argument accordingly. Use the
26973 utility functions added above. Only construct sel2 if we need to.
26975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26977 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
26978 the broadcast of the low byte.
26979 (expand_mult_highpart): Use an explicit encoding for the permutes.
26980 * optabs-query.c (can_mult_highpart_p): Likewise.
26981 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
26982 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26983 (vectorizable_bswap): Likewise.
26984 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
26985 explicit encoding for the power-of-2 permutes.
26986 (vect_permute_store_chain): Likewise.
26987 (vect_grouped_load_supported): Likewise.
26988 (vect_permute_load_chain): Likewise.
26990 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26992 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
26993 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
26994 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
26995 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26996 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
26997 (vect_gen_perm_mask_any): Likewise.
26999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27001 * int-vector-builder.h: New file.
27002 * vec-perm-indices.h: Include int-vector-builder.h.
27003 (vec_perm_indices): Redefine as an int_vector_builder.
27004 (auto_vec_perm_indices): Delete.
27005 (vec_perm_builder): Redefine as a stand-alone class.
27006 (vec_perm_indices::vec_perm_indices): New function.
27007 (vec_perm_indices::clamp): Likewise.
27008 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27009 (vec_perm_indices::new_vector): New function.
27010 (vec_perm_indices::new_expanded_vector): Update for new
27011 vec_perm_indices class.
27012 (vec_perm_indices::rotate_inputs): New function.
27013 (vec_perm_indices::all_in_range_p): Operate directly on the
27014 encoded form, without computing elided elements.
27015 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27016 encoding. Update for new vec_perm_indices class.
27017 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27018 the given vec_perm_builder.
27019 (expand_vec_perm_var): Update vec_perm_builder constructor.
27020 (expand_mult_highpart): Use vec_perm_builder instead of
27021 auto_vec_perm_indices.
27022 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27023 vec_perm_indices instead of auto_vec_perm_indices. Use a single
27024 or double series encoding as appropriate.
27025 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27026 vec_perm_indices instead of auto_vec_perm_indices.
27027 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27028 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27029 (vect_permute_store_chain): Likewise.
27030 (vect_grouped_load_supported): Likewise.
27031 (vect_permute_load_chain): Likewise.
27032 (vect_shift_permute_load_chain): Likewise.
27033 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27034 (vect_transform_slp_perm_load): Likewise.
27035 (vect_schedule_slp_instance): Likewise.
27036 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27037 (vectorizable_mask_load_store): Likewise.
27038 (vectorizable_bswap): Likewise.
27039 (vectorizable_store): Likewise.
27040 (vectorizable_load): Likewise.
27041 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27042 vec_perm_indices instead of auto_vec_perm_indices. Use
27043 tree_to_vec_perm_builder to read the vector from a tree.
27044 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27045 vec_perm_builder instead of a vec_perm_indices.
27046 (have_whole_vector_shift): Use vec_perm_builder and
27047 vec_perm_indices instead of auto_vec_perm_indices. Leave the
27048 truncation to calc_vec_perm_mask_for_shift.
27049 (vect_create_epilog_for_reduction): Likewise.
27050 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27051 from auto_vec_perm_indices to vec_perm_indices.
27052 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27053 instead of changing individual elements.
27054 (aarch64_vectorize_vec_perm_const): Use new_vector to install
27055 the vector in d.perm.
27056 * config/arm/arm.c (expand_vec_perm_d::perm): Change
27057 from auto_vec_perm_indices to vec_perm_indices.
27058 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27059 instead of changing individual elements.
27060 (arm_vectorize_vec_perm_const): Use new_vector to install
27061 the vector in d.perm.
27062 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27063 Update vec_perm_builder constructor.
27064 (rs6000_expand_interleave): Likewise.
27065 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27066 (rs6000_expand_interleave): Likewise.
27068 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27070 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27071 to qimode could truncate the indices.
27072 * optabs.c (expand_vec_perm_var): Likewise.
27074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27076 * Makefile.in (OBJS): Add vec-perm-indices.o.
27077 * vec-perm-indices.h: New file.
27078 * vec-perm-indices.c: Likewise.
27079 * target.h (vec_perm_indices): Replace with a forward class
27081 (auto_vec_perm_indices): Move to vec-perm-indices.h.
27082 * optabs.h: Include vec-perm-indices.h.
27083 (expand_vec_perm): Delete.
27084 (selector_fits_mode_p, expand_vec_perm_var): Declare.
27085 (expand_vec_perm_const): Declare.
27086 * target.def (vec_perm_const_ok): Replace with...
27087 (vec_perm_const): ...this new hook.
27088 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27089 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27090 * doc/tm.texi: Regenerate.
27091 * optabs.def (vec_perm_const): Delete.
27092 * doc/md.texi (vec_perm_const): Likewise.
27093 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27094 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27095 expand_vec_perm for constant permutation vectors. Assert that
27096 the mode of variable permutation vectors is the integer equivalent
27097 of the mode that is being permuted.
27098 * optabs-query.h (selector_fits_mode_p): Declare.
27099 * optabs-query.c: Include vec-perm-indices.h.
27100 (selector_fits_mode_p): New function.
27101 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27102 is defined, instead of checking whether the vec_perm_const_optab
27103 exists. Use targetm.vectorize.vec_perm_const instead of
27104 targetm.vectorize.vec_perm_const_ok. Check whether the indices
27105 fit in the vector mode before using a variable permute.
27106 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27107 vec_perm_indices instead of an rtx.
27108 (expand_vec_perm): Replace with...
27109 (expand_vec_perm_const): ...this new function. Take the selector
27110 as a vec_perm_indices rather than an rtx. Also take the mode of
27111 the selector. Update call to shift_amt_for_vec_perm_mask.
27112 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27113 Use vec_perm_indices::new_expanded_vector to expand the original
27114 selector into bytes. Check whether the indices fit in the vector
27115 mode before using a variable permute.
27116 (expand_vec_perm_var): Make global.
27117 (expand_mult_highpart): Use expand_vec_perm_const.
27118 * fold-const.c: Includes vec-perm-indices.h.
27119 * tree-ssa-forwprop.c: Likewise.
27120 * tree-vect-data-refs.c: Likewise.
27121 * tree-vect-generic.c: Likewise.
27122 * tree-vect-loop.c: Likewise.
27123 * tree-vect-slp.c: Likewise.
27124 * tree-vect-stmts.c: Likewise.
27125 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27127 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27128 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27129 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27130 (aarch64_vectorize_vec_perm_const): ...this new function.
27131 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27132 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27133 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27134 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27135 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27136 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27137 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27139 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
27140 check for NEON modes.
27141 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27142 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27143 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27144 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27146 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
27147 the old VEC_PERM_CONST conditions.
27148 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27149 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27150 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27151 (ia64_vectorize_vec_perm_const_ok): Merge into...
27152 (ia64_vectorize_vec_perm_const): ...this new function.
27153 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27154 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27155 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27156 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27157 * config/mips/mips.c (mips_expand_vec_perm_const)
27158 (mips_vectorize_vec_perm_const_ok): Merge into...
27159 (mips_vectorize_vec_perm_const): ...this new function.
27160 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27161 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27162 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27163 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27164 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27165 (rs6000_expand_vec_perm_const): Delete.
27166 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27168 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27169 (altivec_expand_vec_perm_const_le): Take each operand individually.
27170 Operate on constant selectors rather than rtxes.
27171 (altivec_expand_vec_perm_const): Likewise. Update call to
27172 altivec_expand_vec_perm_const_le.
27173 (rs6000_expand_vec_perm_const): Delete.
27174 (rs6000_vectorize_vec_perm_const_ok): Delete.
27175 (rs6000_vectorize_vec_perm_const): New function.
27176 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27177 an element count and rtx array.
27178 (rs6000_expand_extract_even): Update call accordingly.
27179 (rs6000_expand_interleave): Likewise.
27180 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27181 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27182 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27183 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27184 (rs6000_expand_vec_perm_const): Delete.
27185 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27186 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27187 (altivec_expand_vec_perm_const_le): Take each operand individually.
27188 Operate on constant selectors rather than rtxes.
27189 (altivec_expand_vec_perm_const): Likewise. Update call to
27190 altivec_expand_vec_perm_const_le.
27191 (rs6000_expand_vec_perm_const): Delete.
27192 (rs6000_vectorize_vec_perm_const_ok): Delete.
27193 (rs6000_vectorize_vec_perm_const): New function. Remove stray
27194 reference to the SPE evmerge intructions.
27195 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27196 an element count and rtx array.
27197 (rs6000_expand_extract_even): Update call accordingly.
27198 (rs6000_expand_interleave): Likewise.
27199 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27200 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27202 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27206 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27207 vector mode and that that mode matches the mode of the data
27209 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27210 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
27211 directly using expand_vec_perm_1 when forcing selectors into
27213 (expand_vec_perm_var): New function, split out from expand_vec_perm.
27215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27217 * optabs-query.h (can_vec_perm_p): Delete.
27218 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27219 * optabs-query.c (can_vec_perm_p): Split into...
27220 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27221 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27222 particular selector is valid.
27223 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27224 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27225 (vect_grouped_load_supported): Likewise.
27226 (vect_shift_permute_load_chain): Likewise.
27227 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27228 (vect_transform_slp_perm_load): Likewise.
27229 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27230 (vectorizable_bswap): Likewise.
27231 (vect_gen_perm_mask_checked): Likewise.
27232 * fold-const.c (fold_ternary_loc): Likewise. Don't take
27233 implementations of variable permutation vectors into account
27234 when deciding which selector to use.
27235 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27236 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27237 with a false third argument.
27238 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27239 to test whether the constant selector is valid and can_vec_perm_var_p
27240 to test whether a variable selector is valid.
27242 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27244 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27245 * optabs-query.c (can_vec_perm_p): Likewise.
27246 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27247 instead of vec_perm_indices.
27248 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27249 (vect_gen_perm_mask_checked): Likewise,
27250 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27251 (vect_gen_perm_mask_checked): Likewise,
27253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27255 * optabs-query.h (qimode_for_vec_perm): Declare.
27256 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27257 (qimode_for_vec_perm): ...this new function.
27258 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27260 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27262 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27263 does not have a conditional at the top.
27265 2018-01-02 Richard Biener <rguenther@suse.de>
27267 * ipa-inline.c (big_speedup_p): Fix expression.
27269 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27272 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27275 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27279 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27280 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27281 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27282 cond_taken_branch_cost 3->4.
27284 2018-01-01 Jakub Jelinek <jakub@redhat.com>
27286 PR tree-optimization/83581
27287 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27288 TODO_cleanup_cfg if any changes have been made.
27290 PR middle-end/83608
27291 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27292 convert_modes if target mode has the right side, but different mode
27295 PR middle-end/83609
27296 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27297 last argument when extracting from CONCAT. If either from_real or
27298 from_imag is NULL, use expansion through memory. If result is not
27299 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27300 the parts directly to inner mode, if even that fails, use expansion
27303 PR middle-end/83623
27304 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27305 check for bswap in mode rather than HImode and use that in expand_unop
27308 Copyright (C) 2018 Free Software Foundation, Inc.
27310 Copying and distribution of this file, with or without modification,
27311 are permitted in any medium without royalty provided the copyright
27312 notice and this notice are preserved.