1 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5 and cond temporaries don't have reference type if iterator has
6 pointer type. For init use &for_pre_body instead of pre_p if
7 for_pre_body is non-empty.
9 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
11 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12 double-double modes to SFmode directly directly.
13 (trunc<mode>sf2_fprs): Delete.
15 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
17 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
18 for conversions between IFmode and the decimal floating point modes.
19 (init_float128_ieee): Use the correct names for conversions between
20 KFmode and the decimal floating point modes.
22 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
24 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
25 for the conversions between TDmode and IFmode.
26 (init_float128_ieee): Use more correct names for the conversions
27 between TDmode and KFmode.
29 2018-07-16 Jakub Jelinek <jakub@redhat.com>
31 PR tree-optimization/86526
32 * builtins.c (expand_builtin_memcmp): Formatting fixes.
33 (inline_expand_builtin_string_cmp): Likewise.
34 (inline_string_cmp): Likewise. Use c_readstr instead of
35 builtin_memcpy_read_str. Add unit_mode temporary.
37 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
40 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
41 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
43 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
45 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
46 Alphabetize prototypes of built-in functions, separating out
47 built-in functions that are listed in this section but should be
50 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
53 * expmed.c (emit_store_flag): Do not emit setcc followed by a
54 conditional move when trapping comparison was split to a
55 non-trapping one (and vice versa).
57 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
59 * config/s390/s390.c (s390_function_profiler): Generate nops
60 instead of profiler call sequences.
61 * config/s390/s390.opt: Add the new option.
63 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
65 * config/s390/s390.c (s390_function_profiler): Generate
67 * config/s390/s390.opt: Add the new option.
69 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
71 * common.opt: Add the new warning.
72 * config/s390/s390.c (s390_function_profiler): Emit "brasl
73 %r0,__fentry__" when -mfentry is specified.
74 (s390_option_override_internal): Disallow -mfentry for 31-bit
76 * config/s390/s390.opt: Add the new option.
78 2018-07-16 Richard Biener <rguenther@suse.de>
81 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
82 for function-local FUNCTION_DECL and RESULT_DECL immediately.
84 2018-07-16 Martin Liska <mliska@suse.cz>
87 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
90 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
92 * config/arc/arcHS.md: Update ARCHS scheduling rules.
94 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
96 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
98 * config/arc/arc-cpus.def (hs4x): New cpu.
100 * config/arc/arc-tables.opt: Regenerate.
101 * config/arc/arc.c (arc_sched_issue_rate): New function.
102 (TARGET_SCHED_ISSUE_RATE): Define.
103 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
104 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
106 (attr tune): Add ARCHS4x tune values.
107 (attr tune_dspmpy): Define.
108 (*tst): Correct instruction type.
109 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
110 * config/arc/arcHS4x.md: New file.
111 * config/arc/fpu.md: Update instruction type attributes.
112 * config/arc/t-multilib: Regenerate.
114 2018-07-16 Tom de Vries <tdevries@suse.de>
117 * var-tracking.c (vt_initialize): Fix pre_dec handling.
119 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
121 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
124 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
126 PR tree-optimization/86514
127 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
128 conversion to a boolean type from a type with greater precision.
130 2018-07-16 Tom de Vries <tdevries@suse.de>
132 * var-tracking.c (vt_initialize): Print adjusted insn slim if
133 dump_flags request TDF_SLIM.
135 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
137 * fold-const.c (int_const_binop_1): Abstract...
138 (wide_int_binop): ...wide int code here.
139 (poly_int_binop): ...poly int code here.
140 Abstract the rest of int_const_binop_1 into int_const_binop.
141 * fold-const.h (wide_int_binop): New.
142 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
143 Remove useless PLUS/MINUS_EXPR case.
144 (zero_nonzero_bits_from_vr): Move wide int code...
145 (zero_nonzero_bits_from_bounds): ...here.
146 (extract_range_from_binary_expr_1): Move mask optimization code...
147 (range_easy_mask_min_max): ...here.
148 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
149 (range_easy_mask_min_max): New.
151 2018-07-15 Jeff Law <law@redhat.com>
154 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
157 2018-07-14 Jim Wilson <jimw@sifive.com>
159 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
161 2018-07-14 Paul Koning <ni1d@arrl.net>
163 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
165 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
167 * lto-streamer-out.c (copy_function_or_variable): Dump info about
170 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
171 Steve Munroe <munroesj52@gmail.com>
173 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
174 (_mm_andnot_si128): Likewise.
175 (_mm_or_si128): Likewise.
176 (_mm_xor_si128): Likewise.
178 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
181 * builtins.c (expand_builtin_memcmp): Inline the calls first
182 when result_eq is false.
183 (expand_builtin_strcmp): Inline the calls first.
184 (expand_builtin_strncmp): Likewise.
185 (inline_string_cmp): New routine. Expand a string compare
186 call by using a sequence of char comparison.
187 (inline_expand_builtin_string_cmp): New routine. Inline expansion
188 a call to str(n)cmp/memcmp.
189 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
191 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
193 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
195 * config/arm/driver-arm.c: Include arm-native.h.
196 (host_detect_local_cpu): Use auto-generated data tables.
197 (vendors, arm_cpu_table): Delete. Move part information to ...
198 * config/arm/arm-cpus.in: ... here.
199 * config/arm/parsecpu.awk (gen_native): New function.
200 (vendor, part): New CPU fields.
201 (END): Add support for building the native CPU detection tables.
202 * config/arm/t-arm (arm-native.h): Add build rule.
203 (driver-arm.o): Add dependency on arm-native.h.
205 2018-07-13 Richard Biener <rguenther@suse.de>
208 * match.pd (addr1 - addr2): Allow either of the operand to
211 2018-07-13 Tom de Vries <tdevries@suse.de>
213 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
216 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
218 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
219 arrays instead of numbered variables.
221 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
223 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
224 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
226 2018-07-13 Richard Biener <rguenther@suse.de>
229 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
230 instead of get_context_die.
232 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
233 Richard Biener <rguenther@suse.de>
236 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
237 that the loop latch destination where phi is defined.
239 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
241 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
242 (riscv_expand_epilogue): Add assertion to check interrupt mode.
243 (riscv_set_current_function): Extract getting interrupt type to new
245 (riscv_get_interrupt_type): New function.
246 (riscv_merge_decl_attributes): New function, checking interrupt type is
248 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
250 2018-07-12 Paul Koning <ni1d@arrl.net>
252 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
255 2018-07-12 Paul Koning <ni1d@arrl.net>
257 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
258 zero reference, add doloop_end instead.
259 * doc/md.texi (decrement_and_branch_until_zero): Remove.
260 (Looping patterns): Remove decrement_and_branch_until_zero. Add
261 detail for doloop_end.
263 2018-07-12 Martin Sebor <msebor@redhat.com>
266 * attribs.c (decl_attributes): Reject conflicting attributes before
267 calling attribute handlers.
269 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
271 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
273 (gcc::dump_manager::get_dump_file_name): likewise.
274 (dump_begin): Likewise.
275 * dumpfile.h (dump_begin): Update prototype.
276 (gcc::dump_manager::get_dump_file_name,
277 gcc::dump_manager::get_dump_file_name): Update prototype.
279 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
281 * internal-fn.h (vectorizable_internal_fn_p): New function.
282 * tree-vect-slp.c (compatible_calls_p): Likewise.
283 (vect_build_slp_tree_1): Remove nops argument. Handle calls
284 to internal functions.
285 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
287 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
289 * fold-const.h (inverse_conditions_p): Declare.
290 * fold-const.c (inverse_conditions_p): New function.
291 * match.pd: Use inverse_conditions_p. Add folds of view_converts
292 that test the inverse condition of a conditional internal function.
293 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
294 * internal-fn.c (internal_fn_mask_index): Handle conditional
296 (vectorized_internal_fn_supported_p): New function.
297 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
298 (any_pred_load_store): Replace with...
299 (need_to_predicate): ...this new variable.
300 (redundant_ssa_names): New variable.
301 (ifcvt_can_use_mask_load_store): Move initial checks to...
302 (ifcvt_can_predicate): ...this new function. Handle tree codes
303 for which a conditional internal function exists.
304 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
305 instead of ifcvt_can_use_mask_load_store. Update after variable
307 (predicate_load_or_store): New function, split out from
308 predicate_mem_writes.
309 (check_redundant_cond_expr): New function.
310 (value_available_p): Likewise.
311 (predicate_rhs_code): Likewise.
312 (predicate_mem_writes): Rename to...
313 (predicate_statements): ...this. Use predicate_load_or_store
314 and predicate_rhs_code.
315 (combine_blocks, tree_if_conversion): Update after above name changes.
316 (ifcvt_local_dce): Handle redundant_ssa_names.
317 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
318 general conditional functions.
319 * tree-vect-stmts.c (vectorizable_call): Likewise.
321 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
322 Alan Hayward <alan.hayward@arm.com>
323 David Sherwood <david.sherwood@arm.com>
325 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
326 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
327 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
328 plus and minus and convert them into IFN_COND_FMA-based sequences.
329 (convert_mult_to_fma): Handle conditional plus and minus.
331 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
333 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
334 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
335 (cond_fnms_optab): New optabs.
336 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
338 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
339 * internal-fn.h (get_conditional_internal_fn): Declare.
340 (get_unconditional_internal_fn): Likewise.
341 * internal-fn.c (cond_ternary_direct): New macro.
342 (expand_cond_ternary_optab_fn): Likewise.
343 (direct_cond_ternary_optab_supported_p): Likewise.
344 (FOR_EACH_COND_FN_PAIR): Likewise.
345 (get_conditional_internal_fn): New function.
346 (get_unconditional_internal_fn): Likewise.
347 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
348 (gimple_match_op::gimple_match_op): Add a new overload for 5
350 (gimple_match_op::set_op): Likewise.
351 (gimple_resimplify5): Declare.
352 * genmatch.c (decision_tree::gen): Generate simplifications for
354 * gimple-match-head.c (gimple_simplify): Define an overload for
355 5 operands. Handle calls with 5 arguments in the top-level overload.
356 (convert_conditional_op): Handle conversions from unconditional
357 internal functions to conditional ones.
358 (gimple_resimplify5): New function.
359 (build_call_internal): Pass a fifth operand.
360 (maybe_push_res_to_seq): Likewise.
361 (try_conditional_simplification): Try converting conditional
362 internal functions to unconditional internal functions.
363 Handle 3-operand unconditional forms.
364 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
365 Define ternary equivalents of the current rules for binary conditional
367 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
369 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
370 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
371 (optab): Handle them.
372 (SVE_COND_FP_TERNARY): New int iterator.
373 (sve_fmla_op, sve_fmad_op): New int attributes.
374 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
375 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
376 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
378 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
380 * target.def (preferred_else_value): New target hook.
381 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
382 * doc/tm.texi: Regenerate.
383 * targhooks.h (default_preferred_else_value): Declare.
384 * targhooks.c (default_preferred_else_value): New function.
385 * internal-fn.h (conditional_internal_fn_code): Declare.
386 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
387 (get_conditional_internal_fn): Use it.
388 (conditional_internal_fn_code): New function.
389 * gimple-match.h (gimple_match_cond): New struct.
390 (gimple_match_op): Add a cond member function.
391 (gimple_match_op::gimple_match_op): Update all forms to take a
393 * genmatch.c (expr::gen_transform): Use the same condition as res_op
394 for the suboperation, but don't specify a particular else_value.
395 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
396 (visit_nary_op, visit_reference_op_load): Pass
397 gimple_match_cond::UNCOND to the gimple_match_op constructor.
398 * gimple-match-head.c: Include tree-eh.h
399 (convert_conditional_op): New function.
400 (maybe_resimplify_conditional_op): Likewise.
401 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
402 (gimple_resimplify2): Likewise.
403 (gimple_resimplify3): Likewise.
404 (gimple_resimplify4): Likewise.
405 (maybe_push_res_to_seq): Return null for conditional operations.
406 (try_conditional_simplification): New function.
407 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
409 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
411 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
413 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
415 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
417 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
419 (hash_tree): Do not hash DECL_FCONTEXT
420 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
421 Do not stream DECL_FCONTEXT.
422 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
423 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
425 2018-07-12 Richard Biener <rguenther@suse.de>
428 * dwarf2out.c (gen_block_die): Only output blocks when they have
429 at least one !DECL_IGNORED_P variable.
431 2018-07-12 Richard Biener <rguenther@suse.de>
434 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
435 Remove -mieee-fp handling.
437 2018-07-12 Richard Biener <rguenther@suse.de>
439 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
440 left-over from last patch.
442 2018-07-12 Jakub Jelinek <jakub@redhat.com>
444 PR tree-optimization/86492
445 * gimple-ssa-store-merging.c
446 (imm_store_chain_info::coalesce_immediate_stores): Call
447 check_no_overlap even for the merge_overlapping case. Formatting fix.
449 2018-07-12 Richard Biener <rguenther@suse.de>
452 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
453 move possibly trapping operations into the conditional.
455 2018-07-12 Richard Biener <rguenther@suse.de>
457 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
458 (vn_lookup_simplify_result): Remove recursion limit applied
460 (vn_nary_build_or_lookup_1): Adjust.
461 (try_to_simplify): Likewise.
462 * gimple-match-head.c (gimple_resimplify1): Instead apply one
464 (gimple_resimplify2): Likewise.
465 (gimple_resimplify3): Likewise.
466 (gimple_resimplify4): Likewise.
468 2018-07-11 Jakub Jelinek <jakub@redhat.com>
470 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
471 Use __mmask64 type instead of __mmask8 for __M argument.
472 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
473 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
475 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
476 __mmask16 for __M argument.
477 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
478 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
479 to __mmask16 instead of __mmask8.
480 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
481 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
482 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
483 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
484 instead of __mmask16 for __U argument.
485 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
486 __mmask16 instead of __mmask8 for __U argument.
487 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
489 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
491 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
493 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
495 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
497 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
499 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
501 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
503 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
505 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
507 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
508 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
509 return type as well as __M argument type and all casts from __mmask8
511 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
512 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
513 return type as well as __M argument type and all casts from __mmask8
515 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
516 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
517 return type as well as __M argument type and all casts from __mmask8
519 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
520 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
521 return type as well as __M argument type and all casts from __mmask8
523 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
524 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
527 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
529 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
530 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
533 2018-07-11 Paul Koning <ni1d@arrl.net>
535 * doc/md.texi (define_subst): Document how multiple occurrences of
536 the same argument in the replacement pattern are handled.
538 2018-07-11 Paul Koning <ni1d@arrl.net>
540 * doc/extend.texi (Common Variable Attributes): Move "mode" into
542 (Common Type Attributes): Add "mode" attribute.
544 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
546 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
547 stream DECL_ORIGINAL_TYPE.
548 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
549 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
550 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
551 Do not walk original type.
552 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
554 (write_ts_decl_non_common_tree_pointers): Do not stream
556 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
557 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
559 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
561 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
562 thread twice from the same starting edge.
564 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
566 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
567 * gimple.c (gimple_call_nonnull_result_p): ...here...
568 (gimple_call_nonnull_arg): ...and here.
569 * gimple.h (gimple_call_nonnull_result_p): New.
570 (gimple_call_nonnull_arg): New.
572 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
574 * config/arm/arm-cpus.in: Move information from fpu field of each
575 cpu definition to the isa field.
576 * config/arm/parsecpu.awk (fpu): Delete match rule.
577 (gen_comm_data): Don't add bits from the CPU's FPU entry.
579 2018-07-11 Richard Biener <rguenther@suse.de>
582 * dwarf2out.c (init_sections_and_labels): Use
583 output_asm_line_debug_info consistently.
584 (dwarf2out_early_finish): Likewise.
585 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
588 2018-07-11 Richard Biener <rguenther@suse.de>
590 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
591 Rework father_bb setting in a way to avoid propagating constants
592 multiple times on a loop body.
594 2018-07-10 Mark Wielaard <mark@klomp.org>
597 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
599 2018-07-10 Richard Biener <rguenther@suse.de>
601 * hash-map.h (hash_map::iterator::operator*): Return
602 references to key and value.
604 2018-07-10 Jakub Jelinek <jakub@redhat.com>
607 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
608 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
609 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
610 (gimplify_omp_for): For composite loops, move outer
611 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
612 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
613 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
614 TREE_LIST for both the original class iterator and the "last" helper
615 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
616 loop, remember has_decl_expr from outer composite loops for the
617 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
619 2018-07-09 Martin Sebor <msebor@redhat.com>
623 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
624 accessing implicitly initialized array elements.
625 * expr.c (string_constant): Handle string initializers of
626 character arrays within aggregates.
627 * gimple-fold.c (fold_array_ctor_reference): Add argument.
628 Store element offset. As a special case, handle zero size.
629 (fold_nonarray_ctor_reference): Same.
630 (fold_ctor_reference): Add argument. Store subobject offset.
631 * gimple-fold.h (fold_ctor_reference): Add argument.
633 2018-07-09 Paul Koning <ni1d@arrl.net>
635 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
636 (pdp11_insn_cost): New function.
637 (pdp11_md_asm_adjust): New function.
638 (TARGET_INVALID_WITHIN_DOLOOP): Define.
639 (pdp11_rtx_costs): Update to match machine better.
640 (output_addr_const_pdp11): Correct format mismatch warnings.
641 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
642 * config/pdp11/pdp11.md: General change to add base_cost and/or
643 length attributes for use by new pdp11_insn_cost function.
644 (MIN_BRANCH): Correct definition.
646 (doloop_end): Use standard pattern name for looping pattern.
647 (doloop_end_nocc): New.
648 (movsf): Add another constraint alternative.
649 (zero_extendqihi2): Add constraint alternatives for not in place
651 (zero_extendhisi2): Remove.
652 (shift patterns): Add CC handling variants.
656 (define_peephole2): New peephole to recognize mov that sets CC for
659 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
661 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
662 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
665 2018-07-09 Jakub Jelinek <jakub@redhat.com>
668 * real.c (real_nextafter): Return true if result is denormal.
670 2018-07-09 Martin Liska <mliska@suse.cz>
672 * common.opt: Add back wrongly removed attribute.
674 2018-07-09 Richard Biener <rguenther@suse.de>
677 * dwarf2out.c (gen_block_die): For an early generated DIE
678 always output high/low PC attributes.
680 2018-07-09 Tom de Vries <tdevries@suse.de>
682 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
683 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
684 onto VAR_DECL with abstract origin.
686 2018-07-07 Jim Wilson <jimw@sifive.com>
688 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
690 2018-07-07 Tom de Vries <tdevries@suse.de>
692 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
694 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
696 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
698 (combine_bound): Use wide-int overflow calculation instead of
700 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
702 * fold-const.c (int_const_binop_2): Same.
703 (extract_muldiv_1): Same.
704 (fold_div_compare): Same.
705 (fold_abs_const): Same.
707 * poly-int.h (add): Same.
711 * predict.c (predict_iv_comparison): Same.
712 * profile-count.c (slow_safe_scale_64bit): Same.
713 * simplify-rtx.c (simplify_const_binary_operation): Same.
714 * tree-chrec.c (tree_fold_binomial): Same.
715 * tree-data-ref.c (split_constant_offset_1): Same.
716 * tree-if-conv.c (idx_within_array_bound): Same.
717 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
718 * tree-ssa-phiopt.c (minmax_replacement): Same.
719 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
720 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
721 * vr-values.c (vr_values::adjust_range_with_scev): Same.
722 * wide-int.cc (wi::add_large): Same.
723 (wi::mul_internal): Same.
724 (wi::sub_large): Same.
725 (wi::divmod_internal): Same.
726 * wide-int.h: Change overflow type to overflow_type for neg, add,
727 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
728 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
729 mul_internal, divmod_internal.
730 (overflow_type): New enum.
731 (accumulate_overflow): New.
733 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
735 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
736 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
738 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
740 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
741 argument is checked for zero before entering loop, avoid checking again.
743 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
745 * gimplify.h (generic_expr_could_trap_p): Set as global function.
746 * gimplify.h (generic_expr_could_trap_p): Likwise.
747 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
749 2018-07-06 Jakub Jelinek <jakub@redhat.com>
751 PR tree-optimization/86401
752 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
753 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
754 (fold_bit_and_mask): ... here. New helper function for match.pd.
755 * fold-const.h (fold_bit_and_mask): Declare.
756 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
758 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
761 * target.def (translate_mode_attribute): New hook.
762 * targhooks.h (default_translate_mode_attribute): Declare.
763 * targhooks.c (default_translate_mode_attribute): New function.
764 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
765 * doc/tm.texi: Regenerate.
766 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
767 (rs6000_translate_mode_attribute): New function.
769 2018-07-06 Paul Koning <ni1d@arrl.net>
771 * doc/md.texi (define_split): Document DONE and FAIL.
772 (define_peephole2): Ditto.
774 2018-07-05 Jeff Law <law@redhat.com>
776 PR tree-optimization/86010
777 * tree-ssa-dse.c (compute_trims): More aggressively trim at
778 both the head and tail of mem* and str* calls.
780 2018-07-05 Jim Wilson <jimw@sifive.com>
782 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
784 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
786 * config/aarch64/aarch64-simd.md: correct flags text for
787 MIN_EXPR replacement.
789 2018-07-05 James Clarke <jrtc27@jrtc27.com>
791 * configure: Regenerated.
793 2018-07-05 Carl Love <cel@us.ibm.com>
795 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
796 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
797 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
798 VSX_BUILTIN_DOUBLEL_V4SF.
800 2018-07-05 Martin Sebor <msebor@redhat.com>
803 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
804 than its domain to compute its the upper bound of a char array.
806 2018-07-05 Nathan Sidwell <nathan@acm.org>
808 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
809 * doc/cpp.texi: Update comment.
810 * doc/tm.texi: Rebuilt.
811 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
812 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
813 * doc/extend.texi (Backwards Compatibility): Clarify it is system
814 headers affected by extern "C".
815 * system.h: Poison NO_IMPLICIT_EXTERN_C.
816 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
817 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
818 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
819 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
820 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
821 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
822 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
823 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
824 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
825 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
826 config/sparc/sp64-elf.h, config/spu/spu.h,
827 config/stormy16/stormy16.h, config/v850/v850.h,
828 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
829 define NO_IMPLICIT_EXTERN_C.
830 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
832 2018-07-05 Tamar Christina <tamar.christina@arm.com>
835 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
836 instead of GET_MODE_SIZE when comparing Units.
838 2018-07-05 Tamar Christina <tamar.christina@arm.com>
841 * rtlanal.c (set_noop_p): Constrain on mode change,
842 include hard-reg-set.h
844 2018-07-05 Tamar Christina <tamar.christina@arm.com>
846 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
848 2018-07-05 Jakub Jelinek <jakub@redhat.com>
851 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
854 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
856 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
858 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
861 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
863 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
865 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
868 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
870 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
872 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
875 2018-07-04 Martin Liska <mliska@suse.cz>
877 * tree-switch-conversion.c: Define
878 max_ratio_for_speed and max_ratio_for_size constants.
880 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
881 Martin Liska <mliska@suse.cz>
886 * common.opt: Rename align options with 'str_' prefix.
887 * common/config/i386/i386-common.c (set_malign_value): New
889 (ix86_handle_option): Use it to set -falign-* options/
890 * config/aarch64/aarch64-protos.h (struct tune_params): Change
891 type from int to string.
892 * config/aarch64/aarch64.c: Update default values from int
894 * config/alpha/alpha.c (alpha_override_options_after_change):
896 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
897 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
898 max skip conditionally.
899 * config/i386/freebsd.h (SUBALIGN_LOG): New.
900 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
901 max skip conditionally.
902 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
903 max skip conditionally.
904 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
905 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
906 max skip conditionally.
907 * config/i386/i386.c (struct ptt): Change type from int to
909 (ix86_default_align): Set default values.
910 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
911 max skip conditionally.
912 * config/i386/iamcu.h (SUBALIGN_LOG): New.
913 (ASM_OUTPUT_MAX_SKIP_ALIGN):
914 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
915 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
916 max skip conditionally.
917 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
918 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
919 * config/i386/x86-64.h (SUBALIGN_LOG): New.
920 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
921 max skip conditionally.
922 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
923 * config/ia64/ia64.c (ia64_option_override): Set default values
924 for alignment options.
925 * config/m68k/m68k.c: Handle new str_align_* options.
926 * config/mips/mips.c (mips_set_compression_mode): Change
928 (mips_option_override): Set default values for options.
929 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
931 * config/rs6000/rs6000.c (rs6000_option_override_internal):
933 * config/rx/rx.c (rx_option_override): Likewise.
934 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
935 (LABEL_ALIGN): Use align_labels_log.
936 (LOOP_ALIGN): Use align_loops_align.
937 * config/s390/s390.c (s390_asm_output_function_label): Use new
939 * config/sh/sh.c (sh_override_options_after_change):
940 Change type of constants.
941 * config/spu/spu.c (spu_sched_init): Likewise.
942 * config/sparc/sparc.c (sparc_option_override): Set default
944 * config/visium/visium.c (visium_option_override): Likewise.
945 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
946 emit p2align format with last argument if it's not needed.
947 * doc/invoke.texi: Document extended format of -falign-*.
948 * final.c: Use align_labels alignment.
949 * flags.h (struct target_flag_state): Change type to use
951 (struct align_flags_tuple): New.
952 (struct align_flags): Likewise.
953 (align_loops_log): Redefine macro to use new types.
954 (align_loops_max_skip): Redefine macro to use new types.
955 (align_jumps_log): Redefine macro to use new types.
956 (align_jumps_max_skip): Redefine macro to use new types.
957 (align_labels_log): Redefine macro to use new types.
958 (align_labels_max_skip): Redefine macro to use new types.
959 (align_functions_log): Redefine macro to use new types.
960 (align_loops): Redefine macro to use new types.
961 (align_jumps): Redefine macro to use new types.
962 (align_labels): Redefine macro to use new types.
963 (align_functions): Redefine macro to use new types.
964 (align_functions_max_skip): Redefine macro to use new types.
965 (align_loops_value): New macro.
966 (align_jumps_value): New macro.
967 (align_labels_value): New macro.
968 (align_functions_value): New macro.
969 * function.c (invoke_set_current_function_hook): Propagate
970 alignment values from flags to global variables default in
972 * ipa-icf.c (sem_function::equals_wpa): Use
973 cl_optimization_option_eq instead of memcmp.
974 * lto-streamer.h (cl_optimization_stream_out): Support streaming
976 (cl_optimization_stream_in): Likewise.
977 * optc-save-gen.awk: Support strings in cl_optimization.
978 * opth-gen.awk: Likewise.
979 * opts.c (finish_options): Remove error checking of invalid
981 (MAX_CODE_ALIGN): Remove.
982 (MAX_CODE_ALIGN_VALUE): Likewise.
983 (parse_and_check_align_values): New function.
984 (check_alignment_argument): Likewise.
985 (common_handle_option): Use check_alignment_argument.
986 * opts.h (parse_and_check_align_values): Declare.
987 * toplev.c (init_alignments): Remove.
988 (read_log_maxskip): New.
989 (parse_N_M): Likewise.
990 (parse_alignment_opts): Likewise.
991 (backend_init_target): Remove usage of init_alignments.
992 * toplev.h (parse_alignment_opts): Declare.
993 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
995 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
996 * tree.c (cl_option_hasher::equal): New.
997 * varasm.c: Use new global macros.
999 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1001 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1002 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1003 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1004 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1005 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1006 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1007 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1008 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1009 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1010 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1011 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1012 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1013 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1015 2018-07-04 Martin Liska <mliska@suse.cz>
1016 Jonathan Wakely <jwakely@redhat.com>
1018 * coverage.c: Use correct type.
1019 * doc/invoke.texi: Language correction.
1021 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1024 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1025 ENDBRANCH for non-tail call which may return via indirect branch.
1026 * doc/extend.texi: Document indirect_return attribute.
1028 2018-07-03 Martin Sebor <msebor@redhat.com>
1030 PR tree-optimization/86274
1031 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1033 (format_floating): Correct handling of infinities and NaNs.
1035 2018-07-03 Martin Sebor <msebor@redhat.com>
1037 * print-tree.c (print_real_cst): New function.
1038 (print_node_brief): Call it.
1039 (print_node): Ditto.
1041 2018-07-03 Jeff Law <law@redhat.com>
1043 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1044 into a single pattern.
1046 * config/h8300/h8300.md (ors code_iterator): New.
1047 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1048 a single pattern and single splitter.
1049 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1050 (iorqi3_1, xorqi3_1): Likewise.
1051 (iorqi3, xorqi3 expanders): Similarly.
1053 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1054 (movmd_internal) into a single pattern using the P mode iterator.
1055 (movmd splitters): Similarly.
1056 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1057 (movsd splitters): Similarly.
1059 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1060 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1061 right table based on the size of the operand.
1062 * config/h8300/h8300.md (length_table): Corresponding changes. All
1063 references to "addb", "addw" and "addl" changed to "add".
1064 (btst patterns): Merge two variants into a single pattern.
1065 (tstqi, tsthi): Likewise.
1066 (addhi3_incdec, addsi3_incdec): Likewise.
1067 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1068 (mulhi3, mulsi3): Likewise.
1069 (udivhi3, udivsi3): Likewise.
1070 (divhi3, divsi3): Likewise.
1071 (andorqi3, andorhi3, andorsi3): Likewise.
1073 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1076 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1077 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1079 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1081 PR tree-optimization/85694
1082 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1083 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1085 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1086 (<u>avg<mode>3_ceil): New patterns.
1088 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1090 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1091 scan-tree-dump directive.
1092 * gcc.dg/vect/slp-perm-2.c: Likewise.
1093 * gcc.dg/vect/slp-perm-3.c: Likewise.
1094 * gcc.dg/vect/slp-perm-5.c: Likewise.
1095 * gcc.dg/vect/slp-perm-6.c: Likewise.
1096 * gcc.dg/vect/slp-perm-7.c: Likewise.
1097 * gcc.dg/vect/slp-perm-8.c: Likewise.
1099 2018-07-03 Marek Polacek <polacek@redhat.com>
1102 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1104 2018-07-03 Richard Biener <rguenther@suse.de>
1107 * tree-ssa-structalias.c (find_func_clobbers): Properly
1108 handle indirect calls.
1110 2018-07-03 Jeff Law <law@redhat.com>
1112 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1113 (shifts): New code iterator.
1114 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1115 expander. Fix HImode handling on H8/SX.
1116 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1117 (subqi3, subhi3, subsi3 expanders): Likewise.
1118 (andqi3, andhi3, andsi3 expanders): Likewise.
1119 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1120 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1121 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1122 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1123 (zero_extendqihi2, zero_extendqisi2): Likewise.
1124 (extendqihi2, extendqisi2): Likewise.
1125 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1126 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1127 (rotlqi3_1, rotlhi3_1): Likewise.
1128 (logicalhi3_sn, logicalsi3_sn): Likewise.
1129 (logicalhi3, logicalsi3): Likewise.
1131 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1133 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1134 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1135 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1136 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1137 type to append_pattern_def_seq instead of creating a stmt_vec_info
1139 (build_mask_conversion): Likewise. Remove vinfo argument.
1140 (vect_add_conversion_to_patterm): Likewise, renaming to...
1141 (vect_add_conversion_to_pattern): ...this.
1142 (vect_recog_mask_conversion_pattern): Update call to
1143 build_mask_conversion. Pass the vector type to
1144 append_pattern_def_seq here too.
1145 (vect_recog_gather_scatter_pattern): Update call to
1146 vect_add_conversion_to_pattern.
1148 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1150 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1151 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1152 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1153 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1154 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1155 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1156 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1157 append_pattern_def_seq instead of new_pattern_def_seq.
1158 (vect_recog_divmod_pattern): Do both of the above.
1159 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1162 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1164 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1165 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1166 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1167 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1168 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1169 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1170 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1171 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1172 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1173 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1174 parameter with a single stmt_vec_info.
1175 (vect_recog_func_ptr): Likewise.
1176 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1177 (vect_try_gather_scatter_pattern): ...this.
1178 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1179 the stmt_vec_info of the statement to be matched. Don't clear
1180 STMT_VINFO_RELATED_STMT.
1181 (vect_pattern_recog): Update call accordingly.
1183 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1185 PR tree-optimization/85694
1186 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1187 (uavgM3_ceil): Document new optabs.
1188 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1189 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1191 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1192 (savg_ceil_optab): New optabs.
1193 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1194 (vect_vect_recog_func_ptrs): Add it.
1195 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1196 constant directly from the associated lhs.
1198 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1200 * tree-vect-patterns.c (vect_split_statement): New function.
1201 (vect_convert_input): Use it to try to split an existing cast.
1203 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1205 * poly-int.h (print_hex): New function.
1206 * dumpfile.h (dump_dec, dump_hex): Declare.
1207 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1208 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1209 min_input_precision, operation_precision and operation_sign.
1210 * tree-vect-patterns.c (vect_get_range_info): New function.
1211 (vect_same_loop_or_bb_p, vect_single_imm_use)
1212 (vect_operation_fits_smaller_type): Delete.
1213 (vect_look_through_possible_promotion): Add an optional
1214 single_use_p parameter.
1215 (vect_recog_over_widening_pattern): Rewrite to use new
1216 stmt_vec_info infomration. Handle one operation at a time.
1217 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1218 (vect_truncatable_operation_p, vect_set_operation_type)
1219 (vect_set_min_input_precision): New functions.
1220 (vect_determine_min_output_precision_1): Likewise.
1221 (vect_determine_min_output_precision): Likewise.
1222 (vect_determine_precisions_from_range): Likewise.
1223 (vect_determine_precisions_from_users): Likewise.
1224 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1225 (vect_vect_recog_func_ptrs): Put over_widening first.
1227 (vect_pattern_recog): Call vect_determine_precisions.
1229 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1231 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1232 statements that have been replaced by further pattern statements.
1233 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1235 2018-07-03 Richard Biener <rguenther@suse.de>
1237 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1238 always set *dt. Dump vectype in vectype overload.
1239 * dumpfile.h (dump_gimple_expr): New function.
1240 (dump_gimple_expr_loc): Likewise.
1241 * dumpfile.c (dump_gimple_expr): New function.
1242 (dump_gimple_expr_loc): Likewise.
1244 2018-07-02 Jeff Law <law@redhat.com>
1246 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1247 the H8/300, H8/300H and H8/S variants into a single pattern.
1248 (movhi_h8300, movqi_h8300hs): Similarly.
1249 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1250 (QHI mode iterator): New.
1252 * config/h8300/h8300.md: Remove trailing whitespace.
1254 2018-07-02 Jim Wilson <jimw@sifive.com>
1256 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1257 instead of emit_insn for interrupt returns.
1258 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1259 (riscv_sret, riscv_uret): Likewise.
1261 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1263 * pretty-print.c (selftest::test_pp_format): Move save and restore
1264 of quotes to class auto_fix_quotes, and add an instance.
1265 * selftest.c: Include "intl.h".
1266 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1267 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1268 * selftest.h (selftest::auto_fix_quotes): New class.
1270 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1272 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1273 (aarch64_sve_prepare_conditional_op): Remove.
1274 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1275 Allow aarch64_simd_reg_or_zero as select operand; remove
1276 the aarch64_sve_prepare_conditional_op call.
1277 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1278 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1279 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1280 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1281 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1282 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1283 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1284 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1285 and a splitters to match all of the *_any patterns.
1286 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1288 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1289 (SVE_COND_FP_BINARY_REV): Remove.
1290 (sve_int_op_rev, sve_fp_op_rev): New.
1291 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1292 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1293 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1294 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1295 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1296 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1297 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1298 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1299 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1301 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1302 Remove match_dup 1 from the inner unspec.
1303 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1305 * config/aarch64/aarch64.md (movprfx): New attr.
1306 (length): Default movprfx to 8.
1307 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1308 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1309 (*<su>mul<SVE_I>3_highpart): Likewise.
1310 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1311 (*v<ASHIFT><SVE_I>3): Likewise.
1312 (*<su><MAXMIN><SVE_I>3): Likewise.
1313 (*<su><MAXMIN><SVE_F>3): Likewise.
1314 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1315 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1316 (*div<SVE_F>4): Likewise.
1318 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1320 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1323 2018-07-02 Richard Biener <rguenther@suse.de>
1325 PR tree-optimization/86363
1326 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1327 memset argument refers to a non-variable address.
1329 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1331 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1332 {PLUS,MINUS}_EXPR code to...
1333 (adjust_symbolic_bound): ...here,
1334 (combine_bound): ...here,
1335 (set_value_range_with_overflow): ...and here.
1337 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1339 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1341 (extract_range_from_abs_expr): ...here.
1343 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1345 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1346 -fno-omit-frame-pointer when not optimizing.
1348 2018-07-02 Martin Liska <mliska@suse.cz>
1351 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1352 (propagate_nothrow): Likewise.
1354 2018-07-02 Martin Liska <mliska@suse.cz>
1357 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1359 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1361 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1362 function in r262149, changing "loc" param from source_location to
1363 const dump_location_t &.
1364 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1365 declaration, as above.
1367 2018-07-01 Paul Koning <ni1d@arrl.net>
1369 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1370 -munit-asm, -mgnu-asm, -mdec-asm.
1371 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1372 (pdp11_output_labelref): New.
1373 (pdp11_output_def): New.
1374 (pdp11_output_addr_vec_elt): New.
1375 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1376 %# and %@ format codes.
1377 (pdp11_option_override): New.
1378 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1379 (pdp11_output_ident): New.
1380 (pdp11_asm_named_section): New.
1381 (pdp11_asm_init_sections): New.
1382 (pdp11_file_start): New.
1383 (pdp11_file_end): New.
1384 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1385 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1386 %o, like %c but octal.
1387 (pdp11_option_override): New.
1388 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1390 (DATA_SECTION_ASM_OP): Ditto.
1391 (READONLY_DATA_SECTION_ASM_OP): New.
1392 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1393 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1394 (ASM_OUTPUT_LABELREF): Ditto.
1395 (ASM_OUTPUT_DEF): Ditto.
1396 (ASM_OUTPUT_EXTERNAL): New.
1397 (ASM_OUTPUT_SOURCE_FILENAME): New.
1398 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1399 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1400 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1401 %# and %@ format codes.
1402 * config/pdp11/pdp11.opt (mgnu-asm): New.
1403 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1404 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1405 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1407 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1409 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1410 dereferencing path[] beyond its length.
1412 (debug_all_paths): New.
1413 (rewire_first_differing_edge): New.
1414 (adjust_paths_after_duplication): New.
1415 (duplicate_thread_path): Call adjust_paths_after_duplication.
1417 (thread_through_all_blocks): Add new argument to
1418 duplicate_thread_path.
1420 2018-06-30 Jim Wilson <jimw@sifive.com>
1422 * config/riscv/predicates.md (p2m1_shift_operand): New.
1423 (high_mask_shift_operand): New.
1424 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1425 pattern using p2m1_shift_operand.
1426 (lshsi3_zero_extend_3+2): New combiner pattern using
1427 high_mask_shift_operand.
1429 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1431 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1433 (vect_recog_rotate_pattern): ...here.
1434 (vect_convert_input): Try to insert casts of invariants in the
1436 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1437 preheader to be empty.
1439 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1441 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1442 vector type. If given, install it in the new statement's
1444 (vect_element_precision): New function.
1445 (vect_unpromoted_value): New struct.
1446 (vect_unpromoted_value::vect_unpromoted_value): New function.
1447 (vect_unpromoted_value::set_op): Likewise.
1448 (vect_look_through_possible_promotion): Likewise.
1449 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1450 (vect_widened_op_tree, vect_convert_input): Likewise.
1451 (vect_convert_inputs, vect_convert_output): Likewise.
1452 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1453 to handle the optional cast of the multiplication result and
1454 vect_widened_op_tree to detect the widened multiplication itself.
1455 Do not require the input and output of promotion casts to have
1456 the same sign, but base the signedness of the operation on the
1457 input rather than the result. If the pattern includes two
1458 promotions, check that those promotions have the same sign.
1459 Do not restrict the MULT_EXPR handling to a double-width result;
1460 handle quadruple-width results and wider. Use vect_convert_inputs
1461 to convert the inputs to the common type.
1462 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1463 to handle the optional cast of the ABS result. Also allow a sign
1464 change or a sign extension between the ABS and MINUS.
1465 Use vect_widened_op_tree to detect the widened subtraction and use
1466 vect_convert_inputs to convert the inputs to the common type.
1467 (vect_handle_widen_op_by_const): Delete.
1468 (vect_recog_widen_op_pattern): New function.
1469 (vect_recog_widen_mult_pattern): Use it.
1470 (vect_recog_widen_shift_pattern): Likewise.
1471 (vect_recog_widen_sum_pattern): Use
1472 vect_look_through_possible_promotion to handle the promoted
1475 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1477 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1478 the containing gimple_seq *.
1479 * gimple-iterator.h (gsi_for_stmt): Declare it.
1480 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1481 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1482 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1483 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1484 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1486 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1488 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1489 statement being replaced is part of an existing pattern
1490 definition sequence, inserting the new pattern statements before
1492 (vect_pattern_recog_1): Don't return a bool. If the statement
1493 is already part of a pattern, instead apply pattern matching
1494 to the pattern definition statements. Don't clear the
1495 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1496 (vect_pattern_recog): Don't break after the first match;
1497 continue processing the pattern definition statements instead.
1498 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1500 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1502 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1503 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1504 (vect_recog_widen_sum_pattern): Use it.
1506 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1508 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1509 phi is not a pattern statement and has not been replaced by
1510 a pattern statement.
1511 * tree-vect-patterns.c (type_conversion_p): Don't check
1512 STMT_VINFO_IN_PATTERN_P.
1513 (vect_recog_vector_vector_shift_pattern): Likewise.
1514 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1515 the pattern statement rather than the original statement; check
1516 directly for a WIDEN_MULT_EXPR here.
1517 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1518 vect_is_simple_use to return the pattern statement rather
1519 than the original statement; use is_pattern_stmt_p to check
1520 for such a pattern statement.
1521 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1522 to return the pattern statement rather than the original statement;
1523 don't do the same transformation here.
1524 (vect_is_simple_use): If the defining statement has been replaced
1525 by a pattern statement, return the pattern statement instead.
1526 Remove the corresponding (local) transformation from the vectype
1529 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1531 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1532 end and default to null.
1533 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1534 (vectorizable_reduction): Update calls accordingly, dropping the
1535 gimple ** argument if the passed-back statement isn't needed.
1536 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1537 (vect_recog_rotate_pattern): Likewise.
1538 (vect_recog_mask_conversion_pattern): Likewise.
1539 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1540 (vect_mask_constant_operand_p): Likewise.
1541 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1542 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1543 (get_group_load_store_type, get_load_store_type): Likewise.
1544 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1545 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1546 (vectorizable_conversion, vectorizable_assignment): Likewise.
1547 (vectorizable_shift, vectorizable_operation): Likewise.
1548 (vectorizable_store, vect_is_simple_cond): Likewise.
1549 (vectorizable_condition, vectorizable_comparison): Likewise.
1550 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1551 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1552 and move it to the end. Cope with null def_stmt_outs.
1554 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1556 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1558 2018-06-29 Jeff Law <law@redhat.com>
1560 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1561 displacements for TARGET_V850E2V3 and newer.
1562 (TARGET_LRA_P): Remove. Defaults to LRA now.
1563 * config/v850/v850.md (sign23byte_load): Remove.
1564 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1565 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1567 2018-06-29 Martin Liska <mliska@suse.cz>
1570 * coverage.c (coverage_init): Mangle full path name.
1571 * doc/invoke.texi: Document the change.
1572 * gcov-io.c (mangle_path): New.
1573 * gcov-io.h (mangle_path): Likewise.
1574 * gcov.c (mangle_name): Use mangle_path for path mangling.
1576 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1578 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1579 if starting source register is not even.
1581 2018-06-29 Martin Liska <mliska@suse.cz>
1583 PR tree-optimization/86263
1584 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1585 Make edge redirection.
1587 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1589 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1590 (dump_scope_depth): New variable.
1591 (get_dump_scope_depth): New function.
1592 (dump_begin_scope): New function.
1593 (dump_end_scope): New function.
1594 * dumpfile.h (get_dump_scope_depth): New declaration.
1595 (dump_begin_scope): New declaration.
1596 (dump_end_scope): New declaration.
1597 (class auto_dump_scope): New class.
1598 (AUTO_DUMP_SCOPE): New macro.
1599 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1602 2018-06-29 Richard Biener <rguenther@suse.de>
1604 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1605 compute_all_dependences succeeds.
1606 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1607 exceed --param loop-max-datarefs-for-datadeps.
1609 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1611 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1613 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1616 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1617 alternative 0 in preferred_for_speed attribute.
1619 2018-06-28 Paul Koning <ni1d@arrl.net>
1621 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1622 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1623 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1624 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1625 for shift insn patterns.
1627 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1629 * cgraph.c (cgraph_node::get_body): Replace assignments to
1630 "dump_file" with calls to set_dump_file.
1631 * dumpfile.c (alt_dump_file): Make static, and group with...
1632 (alt_flags): ...this definition.
1633 (dumps_are_enabled): New variable.
1634 (refresh_dumps_are_enabled): New function.
1635 (set_dump_file): New function.
1636 (set_alt_dump_file): New function.
1637 (gcc::dump_manager::dump_start): Replace assignments to
1638 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1640 (gcc::dump_manager::dump_finish): Likewise.
1641 * dumpfile.h (alt_dump_file): Delete decl.
1642 (dumps_are_enabled): New variable decl.
1643 (set_dump_file): New function decl.
1644 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1646 * tree-nested.c (lower_nested_functions): Replace assignments to
1647 "dump_file" with calls to set_dump_file.
1649 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1651 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1652 goto_locus of each outgoing edge of each basic block.
1654 2018-06-28 Richard Biener <rguenther@suse.de>
1656 * dwarf2out.c (decl_scope_table): Remove.
1657 (push_decl_scope): Likewise.
1658 (pop_decl_scope): Likewise.
1659 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1660 (gen_struct_or_union_type_die): Likewise.
1661 (gen_tagged_type_die): Likewise.
1662 (dwarf2out_init): Do not initialize decl_scope_table.
1663 (dwarf2out_c_finalize): Do not free it.
1665 2018-06-28 Richard Biener <rguenther@suse.de>
1667 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1668 deciding whether to not re-use a DIE.
1670 2018-06-28 Richard Biener <rguenther@suse.de>
1672 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1673 DW_AT_abstract_origin attribute.
1675 2018-06-28 Martin Liska <mliska@suse.cz>
1677 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1678 Use newly introduced constants.
1679 * tree-switch-conversion.h (struct jump_table_cluster):
1680 Define max_ratio_for_size and max_ratio_for_speed.
1682 2018-06-28 Martin Liska <mliska@suse.cz>
1684 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1685 Add new checking assert to catch invalid state.
1686 (jump_table_cluster::can_be_handled): Handle single case
1688 (jump_table_cluster::is_beneficial): Bail out for such case.
1689 (bit_test_cluster::find_bit_tests):
1690 Add new checking assert to catch invalid state.
1691 (bit_test_cluster::can_be_handled): Handle single case
1693 (bit_test_cluster::is_beneficial): Bail out for such case.
1694 (switch_decision_tree::analyze_switch_statement):
1697 2018-06-28 Martin Liska <mliska@suse.cz>
1699 * common.opt: Introduce -completion option.
1700 * gcc.c (driver_handle_option): Handle it.
1701 (driver::main): Print completions if completion
1703 * opt-suggestions.c (option_proposer::get_completions):
1705 (option_proposer::suggest_completion): Likewise.
1706 (option_proposer::find_param_completions): Likewise.
1707 (verify_autocompletions): Likewise.
1708 (test_completion_valid_options): Likewise.
1709 (test_completion_valid_params): Likewise.
1710 (in_completion_p): Likewise.
1711 (empty_completion_p): Likewise.
1712 (test_completion_partial_match): Likewise.
1713 (test_completion_garbage): Likewise.
1714 (opt_proposer_c_tests): Likewise.
1715 * opt-suggestions.h: Declare new functions.
1716 * opts.c (common_handle_option): Handle OPT__completion_.
1717 * selftest-run-tests.c (selftest::run_tests): Add
1718 opt_proposer_c_tests.
1719 * selftest.c (assert_str_startswith): New.
1720 * selftest.h (assert_str_startswith): Likewise.
1721 (opt_proposer_c_tests): New.
1722 (ASSERT_STR_STARTSWITH): Likewise.
1724 2018-06-28 Martin Liska <mliska@suse.cz>
1726 * Makefile.in: Add opt-suggestions.o.
1727 * gcc-main.c: Include opt-suggestions.h.
1728 * gcc.c (driver::driver): Likewise.
1729 (driver::~driver): Remove m_option_suggestions.
1730 (driver::build_option_suggestions): Moved to option_proposer.
1731 (driver::suggest_option): Likewise.
1732 (driver::handle_unrecognized_options): Use option_proposer.
1733 * gcc.h (class driver): Add new memver m_option_proposer.
1734 * opt-suggestions.c: New file.
1735 * opt-suggestions.h: New file.
1737 2018-06-28 Martin Liska <mliska@suse.cz>
1739 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1740 (auto_string_vec::~auto_string_vec): Likewise.
1742 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1744 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1745 prevent_decl_creation_for_types fields up and add reset_location field.
1746 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1747 statement if id->reset_location is true.
1748 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1749 id->reset_location is true.
1750 (copy_phis_for_bb): Force input_location on the arguments if
1751 id->reset_location is true.
1752 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1753 is set on the function to be inlined.
1755 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1757 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1759 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1761 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1762 registers for Pmode.
1763 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1764 hard registers for the clobbered pseudo.
1766 2018-06-27 Paul Koning <ni1d@arrl.net>
1768 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1769 mutually exclusive options.
1770 * config/pdp11/constraints.md (h): New constraint.
1771 (O): Update definition to match shift code generation.
1772 (D): New constraint.
1773 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1775 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1777 (output_jump): Change arguments.
1778 (pdp11_fixed_cc_regs): New function.
1779 (pdp11_cc_mode): Ditto.
1780 (pdp11_expand_shift): Ditto.
1781 (pdp11_assemble_shift): Ditto.
1782 (pdp11_small_shift): Ditto.
1783 (pdp11_branch_cost): Remove.
1784 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1786 (pdp11_register_move_cost): Update for CC registers.
1787 (pdp11_rtx_costs): Add case for LSHIFTRT.
1788 (pdp11_output_jump): Add CCNZ mode conditional branches.
1789 (notice_update_cc_on_set): Remove.
1790 (pdp11_cc_mode): New function.
1791 (simple_memory_operand): Correct pre/post decrement case.
1792 (no_side_effect_operand): New function.
1793 (pdp11_regno_reg_class): Add CC_REGS class.
1794 (pdp11_fixed_cc_regs): New function.
1795 (pdp11_small_shift): New function.
1796 (pdp11_expand_shift): New function to expand shift insns.
1797 (pdp11_assemble_shift): New function to output shifts.
1798 (pdp11_branch_cost): Remove.
1799 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1800 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1801 (WCHAR_TYPE): Ditto.
1802 (PTRDIFF_TYPE): Ditto.
1803 (ADJUST_INSN_LENGTH): New macro.
1804 (FIXED_REGISTERS): Add CC registers.
1805 (CALL_USED_REGISTERS): Ditto.
1807 (REG_CLASS_NAMES): Ditto.
1808 (REG_CLASS_CONTENTS): Ditto.
1809 (SELECT_CC_MODE): Use new function.
1810 (TARGET_FLAGS_REGNUM): New macro.
1811 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1812 (cc0_reg_rtx): Remove.
1813 (CC_STATUS_MDEP): Remove.
1814 (CC_STATUS_MDEFP_INIT): Remove.
1815 (CC_IN_FPU): Remove.
1816 (NOTICE_UPDATE_CC): Remove.
1817 (REGISTER_NAMES): Add CC registers.
1818 (BRANCH_COST): Change to constant 1.
1819 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1821 * config/pdp11/pdp11.opt (mbcopy): Remove.
1822 (mbcopy-builtin): Remove.
1823 (mbranch-cheap): Remove.
1824 (mbranch-expensive): Remove.
1825 * config/pdp11/predicates.md (expand_shift_operand): Update to
1826 match shift code generation.
1827 (ccnz_operator): New predicate.
1828 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1829 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1830 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1832 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1833 description of O constraint.
1835 2018-06-27 Jeff Law <law@redhat.com>
1836 Austin Law <austinklaw@gmail.com>
1838 * config/v850/v850.md (addsi3_set_flags): New pattern.
1839 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1840 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1841 (zero_extendhisi2_v850_set_flags): Likewise.
1842 (zero_extendqisi2_v850_set_flags): Likewise.
1843 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1844 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1845 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1847 * config/v850/v850-protos.h (notice_update_cc): Remove.
1848 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1849 (v850_print_operand): Handle 'D' and "d".
1850 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1851 Add handling of arithmetic/logical operations compared against zero.
1852 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1853 Do not look at v850_compare_op, instead get mode from last argument.
1854 (v850_gen_compare): Remove
1855 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1856 after reload for prologue insns.
1857 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1859 (construct_save_jarl): Likewise.
1860 (TARGET_FLAGS_REGNUM): Define.
1861 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1862 (NOTICE_UPDATE_CC): Remove.
1863 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1864 than cc0. Conditionalize on reload_completed.
1865 (cmpsi_insn, setfcc_insn): Likewise.
1866 (tst1 splitter): Turn into define_and_split which sets the flags
1868 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1869 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1870 (cstoresf4, cstoredf4): Clobber the flags.
1871 (cmpsi, cmpsf, cmpdf): Remove expanders.
1872 (setf_insn): Remove pattern.
1873 (addsi3): Turn into define_and_split which clobbers the flags after
1874 reload and a suitable pattern (addsi3_clobber_flags) for use after
1876 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1877 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1878 (ashrsi3, ashrsi3_v850e2): Likewise.
1879 (bins): Clobber the flags.
1880 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1881 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1882 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1883 (call_value_internal_short, call_value_internal_long): Likewise.
1884 (callt_save_interrupt, callt_return_interrupt): Likewise.
1885 (save_interrupt, return_interrupt): Likewise.
1886 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1887 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1888 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1889 (All FP comparisons): Only allow after reload has completed.
1891 (divh, divhu): Tweak output template.
1892 (branch_z_normal, branch_z_invert): Remove
1893 (branch_nz_normal, branch_nz_invert): Likewise.
1894 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1896 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1897 * config/v850/v850.c (notice_update_cc): Remove.
1898 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1899 (CC_NO_CARRY): Likewise.
1900 (NOTICE_UPDATE_CC): Define to nothing.
1901 * config/v850/v850.md: Remove block comment on cc0 handling
1902 Remove "cc" attribute from all patterns. Remove cc_status handling
1903 from all patterns. Minor formatting fixes.
1905 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1907 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1908 (cortex-a76.cortex-a55): Likewise.
1909 * config/aarch64/aarch64-tune.md: Regenerate.
1910 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1911 cortex-a76.cortex-a55.
1913 2018-06-27 Jeff Law <law@redhat.com>
1915 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1916 (MULTILIB_DIRNAMES): Similarly.
1918 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1920 * gimple.h (gimple_return_retbnd): Delete.
1921 (gimple_return_set_retbnd): Likewise.
1922 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1923 gimple_return_set_retbnd.
1924 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1925 gimple_return_retbnd and adjust.
1926 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1927 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1928 Explicitly return NULL in a couple more cases. Move assertion
1929 on debug statements and remove unreachable code.
1930 (reset_debug_binding): Do not test id->retbnd.
1931 (expand_call_inline): Do not set it.
1933 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1935 * configure.ac: Add --disable-gcov option.
1936 * configure: Regenerate.
1937 * Makefile.in: Honour @enable_gcov@.
1938 * doc/install.texi: Document --disable-gcov.
1940 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1942 * config/arm/arm-cpus.in (cortex-a76): New entry.
1943 (cortex-a76.cortex-a55): Likewise.
1944 * config/arm/arm-tables.opt: Regenerate.
1945 * config/arm/arm-tune.md: Likewise.
1946 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1947 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1948 cortex-a76.cortex-a55.
1950 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1953 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1955 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1957 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1959 (EPILOGUE_USES): Likewise.
1961 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1963 * tree-inline.c (remap_location): New function extracted from...
1964 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1965 (copy_phis_for_bb): ...here. Call remap_location.
1966 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1968 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1970 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1971 unaligned vsx for 16B memset.
1973 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1976 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1977 ieee128_float_type_node to long_double_type_node unless
1978 TARGET_LONG_DOUBLE_128 is set.
1980 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1982 * cfgloop.c (get_loop_location): Convert return type from
1983 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1984 by implicit construction from rtx_insn *, and using
1985 dump_user_location_t::from_function_decl for the fallback case.
1986 * cfgloop.h (get_loop_location): Convert return type from
1987 location_t to dump_user_location_t.
1988 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1989 dump_printf_loc to pass in a dump_location_t rather than a
1990 location_t, via the gimple stmt.
1991 * coverage.c (get_coverage_counts): Update calls to
1992 dump_printf_loc to pass in dump_location_t rather than a
1994 * doc/optinfo.texi (Dump types): Convert example of
1995 dump_printf_loc from taking "locus" to taking "insn". Update
1996 description of the "_loc" calls to cover dump_location_t.
1997 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1999 (dump_user_location_t::dump_user_location_t): New constructors,
2000 from gimple *stmt and rtx_insn *.
2001 (dump_user_location_t::from_function_decl): New function.
2002 (dump_loc): Make static.
2003 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2004 const dump_location_t &.
2005 (dump_generic_expr_loc): Delete.
2006 (dump_printf_loc): Convert param "loc" from location_t to
2007 const dump_location_t &.
2008 (selftest::test_impl_location): New function.
2009 (selftest::dumpfile_c_tests): New function.
2010 * dumpfile.h: Include "profile-count.h".
2011 (class dump_user_location_t): New class.
2012 (struct dump_impl_location_t): New struct.
2013 (class dump_location_t): New class.
2014 (dump_printf_loc): Convert 2nd param from source_location to
2015 const dump_location_t &.
2016 (dump_generic_expr_loc): Delete.
2017 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2018 const dump_location_t &.
2019 * gimple-fold.c (fold_gimple_assign): Update call to
2020 dump_printf_loc to pass in a dump_location_t rather than a
2021 location_t, via the gimple stmt.
2022 (gimple_fold_call): Likewise.
2023 * gimple-loop-interchange.cc
2024 (loop_cand::analyze_iloop_reduction_var): Update for change to
2025 check_reduction_path.
2026 (tree_loop_interchange::interchange): Update for change to
2028 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2029 change in return-type of find_loop_location.
2030 (graphite_regenerate_ast_isl): Likewise.
2031 * graphite-optimize-isl.c (optimize_isl): Likewise.
2032 * graphite.c (graphite_transform_loops): Likewise.
2033 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2034 pass in a dump_location_t rather than a location_t, via the
2036 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2037 * ipa.c (walk_polymorphic_call_targets): Likewise.
2038 * loop-unroll.c (report_unroll): Convert "locus" param from
2039 location_t to dump_location_t.
2040 (decide_unrolling): Update for change to get_loop_location's
2042 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2043 location_t to dump_user_location_t.
2044 (grid_find_single_omp_among_assignments_1): Updates calls to
2045 dump_printf_loc to pass in a dump_location_t rather than a
2046 location_t, via the gimple stmt.
2047 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2048 location_t to dump_location_t. Updates calls to dump_printf_loc
2049 to pass in a dump_location_t rather than a location_t, via the
2051 (grid_inner_loop_gridifiable_p): Likewise.
2052 (grid_dist_follows_simple_pattern): Likewise.
2053 (grid_gfor_follows_tiling_pattern): Likewise.
2054 (grid_target_follows_gridifiable_pattern): Likewise.
2055 (grid_attempt_target_gridification): Convert initialization
2056 of local "grid" from memset to zero-initialization; FIXME: does
2057 this require C++11? Update call to dump_printf_loc to pass in a
2058 optinfo_location rather than a location_t, via the gimple stmt.
2059 * profile.c (read_profile_edge_counts): Updates call to
2060 dump_printf_loc to pass in a dump_location_t rather than a
2062 (compute_branch_probabilities): Likewise.
2063 * selftest-run-tests.c (selftest::run_tests): Call
2065 * selftest.h (dumpfile_c_tests): New decl.
2066 * tree-loop-distribution.c (pass_loop_distribution::execute):
2067 Update for change in return type of find_loop_location.
2068 * tree-parloops.c (parallelize_loops): Likewise.
2069 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2070 "locus" from location_t to dump_user_location_t.
2071 (canonicalize_loop_induction_variables): Likewise.
2072 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2073 for change in return type of find_loop_location.
2074 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2075 to dump_printf_loc to pass in a dump_location_t rather than a
2076 location_t, via the stmt.
2077 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2079 * tree-vect-loop-manip.c (find_loop_location): Convert return
2080 type from source_location to dump_user_location_t.
2081 (vect_do_peeling): Update for above change.
2082 (vect_loop_versioning): Update for change in type of
2084 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2085 from location_t to dump_user_location_t.
2086 (vect_estimate_min_profitable_iters): Update for change in type
2088 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2089 location_t to dump_location_t.
2090 (vect_slp_bb): Update for change in type of vect_location.
2091 * tree-vectorizer.c (vect_location): Convert from source_location
2092 to dump_user_location_t.
2093 (try_vectorize_loop_1): Update for change in vect_location's type.
2094 (vectorize_loops): Likewise.
2095 (increase_alignment): Likewise.
2096 * tree-vectorizer.h (vect_location): Convert from source_location
2097 to dump_user_location_t.
2098 (find_loop_location): Convert return type from source_location to
2099 dump_user_location_t.
2100 (check_reduction_path): Convert 1st param from location_t to
2101 dump_user_location_t.
2102 * value-prof.c (check_counter): Update call to dump_printf_loc to
2103 pass in a dump_user_location_t rather than a location_t; update
2104 call to error_at for change in type of "locus".
2105 (check_ic_target): Update call to dump_printf_loc to
2106 pass in a dump_user_location_t rather than a location_t, via the
2109 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2111 * config/s390/s390.h (enum processor_flags): Do not use
2112 default tune parameter when -march was specified.
2114 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2117 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2118 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2121 2018-06-26 Richard Biener <rguenther@suse.de>
2123 PR tree-optimization/86287
2125 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2126 (vect_analyze_loop): Initialize n_stmts.
2128 2018-06-26 Richard Biener <rguenther@suse.de>
2131 * fold-const.c (fold_convertible_p): Pointer extension
2134 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2137 * dwarf2out.c (loc_list_has_views): Adjust comments.
2138 (dw_loc_list): Split single cross-partition range with
2141 2018-06-25 Jeff Law <law@redhat.com>
2143 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2144 on -mbig-switch by default.
2146 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2148 (const_float_0_operand): Remove unused predicate.
2149 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2150 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2151 (recipsf2): New expander. Original pattern now called
2153 (recipdf2, recipdf2_insn): Similarly.
2154 (rsqrtsf2, rsqrtsf2_insn): Similarly
2155 (rsqrtdf2, rsqrtdf2_insn): Similarly
2157 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2159 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2160 Simplify logic for FreeBSD (twice).
2162 2018-06-25 Martin Sebor <msebor@redhat.com>
2164 PR tree-optimization/86204
2165 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2166 a strnlen result if it's less than the length of the string.
2168 2018-06-25 Martin Sebor <msebor@redhat.com>
2170 PR tree-optimization/85700
2171 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2172 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2173 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2175 2018-06-25 Martin Sebor <msebor@redhat.com>
2177 * doc/extend.texi (Zero-length arrays): Update and clarify.
2179 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2181 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2182 added IEEE/IBM long double multilib support on PowerPC little
2183 endian Linux systems.
2184 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2185 (MULTILIB_DEFAULTS): Likewise.
2186 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2188 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2189 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2190 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2192 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2195 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2196 (REORDER_45): Likewise.
2198 2018-06-25 Jeff Law <law@redhat.com>
2200 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2201 dividend to 32 bits. Adjust length.
2202 (udivmodhi4): Cleanup output template. Fix length.
2204 2018-06-25 Carl Love <cel@us.ibm.com>
2206 * config/rs6000/vsx.md: Change word selector to prefered location.
2208 2018-06-25 Richard Biener <rguenther@suse.de>
2210 PR tree-optimization/86304
2211 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2212 epilogue-if-converted loops as well.
2214 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2216 * lto-section-out.c (lto_begin_section): Do not print section
2217 name for noaddr and unnumbered dumps.
2219 2018-06-25 Richard Biener <rguenther@suse.de>
2221 * tree-vectorizer.h (struct vec_info_shared): New structure
2222 with parts split out from struct vec_info and loop_nest from
2223 struct _loop_vec_info.
2224 (struct vec_info): Adjust accordingly.
2225 (struct _loop_vec_info): Likewise.
2226 (LOOP_VINFO_LOOP_NEST): Adjust.
2227 (LOOP_VINFO_DATAREFS): Likewise.
2228 (LOOP_VINFO_DDRS): Likewise.
2229 (struct _bb_vec_info): Likewise.
2230 (BB_VINFO_DATAREFS): Likewise.
2231 (BB_VINFO_DDRS): Likewise.
2232 (struct _stmt_vec_info): Add dr_aux member.
2233 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2234 (DR_MISALIGNMENT_UNINITIALIZED): New.
2235 (set_dr_misalignment): Adjust.
2236 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2237 (vect_analyze_loop): Adjust prototype.
2238 (vect_analyze_loop_form): Likewise.
2239 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2240 Compute dependences lazily.
2241 (vect_record_base_alignments): Use shared datarefs/ddrs.
2242 (vect_verify_datarefs_alignment): Likewise.
2243 (vect_analyze_data_refs_alignment): Likewise.
2244 (vect_analyze_data_ref_accesses): Likewise.
2245 (vect_analyze_data_refs): Likewise.
2246 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2247 constructor parameter for shared part.
2248 (vect_analyze_loop_form): Pass in shared part and adjust.
2249 (vect_analyze_loop_2): Pass in storage for the number of
2250 stmts. Move loop nest finding to the caller. Compute
2252 (vect_analyze_loop): Pass in shared part.
2253 (vect_transform_loop): Verify shared datarefs are unchanged.
2254 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2255 constructor parameter for shared part.
2256 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2257 (vect_slp_bb): Verify shared datarefs are unchanged before
2259 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2261 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2262 DR_MISALIGNMENT_UNINITIALIZED.
2263 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2264 parameter for shared part.
2265 (vec_info::~vec_info): Adjust.
2266 (vec_info_shared::vec_info_shared): New.
2267 (vec_info_shared::~vec_info_shared): Likewise.
2268 (vec_info_shared::save_datarefs): Likewise.
2269 (vec_info_shared::check_datarefs): Likewise.
2270 (try_vectorize_loop_1): Construct shared part live for analyses
2271 of a single loop for multiple vector sizes.
2272 * tree-parloops.c (gather_scalar_reductions): Adjust.
2274 2018-06-25 Richard Biener <rguenther@suse.de>
2276 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2277 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2278 (vect_analyze_data_refs): Remove similar code from here and
2279 simplify accordingly.
2281 2018-06-25 Richard Biener <rguenther@suse.de>
2283 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2284 for reverse storage order accesses rather than asserting
2285 they cannot happen here.
2287 2018-06-25 Tom de Vries <tdevries@suse.de>
2290 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2291 Use data16 instead of .byte for insn prefix.
2293 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2296 * parser.c (make_char_string_pack): Pass this literal chars
2297 through cpp_interpret_string.
2298 (cp_parser_userdef_numeric_literal): Check the result of
2299 make_char_string_pack.
2301 2018-06-24 Maya Rashish <coypu@sdf.org>
2303 * ginclude/stddef.h: Simplify conditions around avoiding
2304 re-definition of __size_t.
2306 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2308 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2309 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2311 2018-06-22 Maya Rashish <coypu@sdf.org>
2313 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2314 made to OpenBSD and NetBSD through the years.
2316 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2318 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2319 behavior of vec_pack (vector double, vector double) to match
2320 behavior of vec_float2 (vector double, vector double).
2322 2018-06-22 Olivier Hainque <hainque@adacore.com>
2324 * gimplify.c (gimplify_function_tree): Prevent creation
2325 of a trampoline for the address of the current function
2326 passed to entry/exit instrumentation hooks.
2328 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2331 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2334 2018-06-22 Martin Liska <mliska@suse.cz>
2336 PR tree-optimization/86263
2337 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2338 Bail out if is_enabled is false.
2339 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2341 (jump_table_cluster::is_enabled): New function.
2343 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2345 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2346 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2347 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2348 (lto_input_ts_binfo_tree_pointers): Likewise.
2349 * tree-streamer-out.c (streamer_write_tree_bitfields,
2350 write_ts_binfo_tree_pointers): Likewise.
2351 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2353 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2355 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2357 2018-06-22 Martin Liska <mliska@suse.cz>
2359 * symbol-summary.h (get): Make it pure and inline move
2360 functionality from ::get function.
2361 (get): Remove and inline into ::get and ::get_create.
2362 (get_create): Move code from ::get function.
2364 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2367 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2368 -x assembler-with-cpp.
2370 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2372 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2373 _FILE_OFFSET_BITS=64 for C++.
2375 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2377 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2378 conversion insn that shows up when pr85657-3.c is compiled using
2379 IEEE 128-bit long double.
2380 (neg<mode>2_internal): Use the correct mode to check whether the
2381 mode is IBM extended.
2382 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2383 multiply and divide external functions from being created more
2386 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2388 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2390 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2391 the edge can be forwarded.
2392 (cfg_layout_merge_blocks): Likewise.
2394 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2396 * except.c (finish_eh_generation): Commit edge insertions only after
2397 the EH edges have been redirected from post-landing to landing pads.
2399 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2401 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2402 create_tmp_var_for to create the FRAME decl.
2403 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2405 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2407 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2408 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2409 debug statement when resetting its value.
2410 (expand_call_inline): Copy the locus of the call onto the assignment
2411 of the return value, if any. Use local variable in more cases.
2413 2018-06-21 Martin Liska <mliska@suse.cz>
2415 * ipa-pure-const.c (propagate_nothrow): Use
2416 funct_state_summaries->get.
2417 (dump_malloc_lattice): Likewise.
2418 (propagate_malloc): Likewise.
2420 2018-06-21 Richard Biener <rguenther@suse.de>
2422 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2423 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2424 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2426 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2427 BLOCK_ABSTRACT_ORIGIN unconditionally.
2429 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2431 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2433 * ipa-reference.c (ipa_reference_c_finalize): Delete
2434 ipa_ref_opt_sum_summaries and set it to NULL.
2436 2018-06-21 Tom de Vries <tdevries@suse.de>
2438 PR tree-optimization/85859
2439 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2440 test with comment from bb_no_side_effects_p.
2442 2018-06-21 Richard Biener <rguenther@suse.de>
2444 PR tree-optimization/86232
2445 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2446 max for constant niter.
2448 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2450 * config/aarch64/aarch64-simd.md
2451 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2453 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2455 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2456 Make opernads of the unspec commutative.
2458 2018-06-21 Richard Biener <rguenther@suse.de>
2460 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2461 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2462 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2463 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2465 (vect_record_base_alignments): Do not record base alignment
2466 for gathers or scatters.
2467 (vect_compute_data_ref_alignment): Drop return value that is always
2468 true. Bail out early for gathers or scatters.
2469 (vect_enhance_data_refs_alignment): Bail out early for gathers
2471 (vect_find_same_alignment_drs): Likewise.
2472 (vect_analyze_data_refs_alignment): Remove dead code.
2473 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2474 (vect_analyze_data_refs): For possible gathers or scatters do
2475 not create an alternate DR, just check their possible validity
2476 and mark them. Adjust DECL_NONALIASED handling to not rely
2478 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2479 update inits of gathers or scatters.
2480 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2481 Also copy gather/scatter flag to pattern vinfo.
2483 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2485 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2486 behavior of vec_packsu (vector unsigned long long, vector unsigned
2487 long long) to match behavior of vec_packs with same signature.
2489 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2490 Thomas Schwinge <thomas@codesourcery.com>
2491 Cesar Philippidis <cesar@codesourcery.com>
2493 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2494 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2495 (gimplify_adjust_omp_clauses): Likewise.
2496 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2497 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2498 (gimplify_omp_target_update): Update handling of acc update and
2500 * omp-low.c (install_var_field): Remove unused parameter
2501 base_pointers_restrict.
2502 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2503 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2505 (omp_target_base_pointers_restrict_p): Delete.
2506 (scan_omp_target): Update call to scan_sharing_clauses.
2507 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2509 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2510 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2511 (convert_local_omp_clauses): Likewise.
2512 * tree-pretty-print.c (dump_omp_clause): Likewise.
2513 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2515 (omp_clause_code_name): Likewise.
2517 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2520 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2523 PR tree-optimization/86231
2524 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2525 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2527 2018-06-20 Tom de Vries <tdevries@suse.de>
2529 PR tree-optimization/86097
2530 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2531 iv type if signedness of iv type is not the same as that of *nit.
2533 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2535 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2536 EDGE_EH edges, verify they are all EDGE_EH.
2538 2018-06-20 Maya Rashish <coypu@sdf.org>
2540 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2542 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2544 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2545 * config/aarch64/aarch64.c (xgene1_tunings): Add
2546 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2547 (aarch64_mode_valid_for_sched_fusion_p):
2548 Allow 16-byte modes.
2549 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2550 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2552 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2554 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2555 * config/aarch64/iterators.md (VQ2): New mode iterator.
2557 2018-06-20 Martin Liska <mliska@suse.cz>
2559 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2560 Change default ratio from 10 to 8.
2562 2018-06-20 Martin Liska <mliska@suse.cz>
2564 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2566 (bit_test_cluster::find_bit_tests): Likewise.
2567 (switch_decision_tree::analyze_switch_statement): Find clusters.
2568 * tree-switch-conversion.h (struct jump_table_cluster): Document
2571 2018-06-20 Martin Liska <mliska@suse.cz>
2573 * tree-switch-conversion.c (switch_conversion::collect):
2574 Record m_uniq property.
2575 (switch_conversion::expand): Bail out for special conditions.
2576 (group_cluster::~group_cluster): New.
2577 (group_cluster::group_cluster): Likewise.
2578 (group_cluster::dump): Likewise.
2579 (jump_table_cluster::emit): New.
2580 (switch_decision_tree::fix_phi_operands_for_edges): New.
2581 (struct case_node): Remove struct.
2582 (jump_table_cluster::can_be_handled): New.
2583 (case_values_threshold): Moved to header.
2584 (reset_out_edges_aux): Likewise.
2585 (jump_table_cluster::is_beneficial): New.
2586 (bit_test_cluster::can_be_handled): Likewise.
2587 (add_case_node): Remove.
2588 (bit_test_cluster::is_beneficial): New.
2589 (case_bit_test::cmp): New.
2590 (bit_test_cluster::emit): New.
2591 (expand_switch_as_decision_tree_p): Remove.
2592 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2593 (fix_phi_operands_for_edge): Likewise.
2594 (switch_decision_tree::analyze_switch_statement): New.
2595 (compute_cases_per_edge): Move ...
2596 (switch_decision_tree::compute_cases_per_edge): ... here.
2597 (try_switch_expansion): Likewise.
2598 (switch_decision_tree::try_switch_expansion): Likewise.
2599 (record_phi_operand_mapping): Likewise.
2600 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2601 (emit_case_decision_tree): Likewise.
2602 (switch_decision_tree::emit): Likewise.
2603 (balance_case_nodes): Likewise.
2604 (switch_decision_tree::balance_case_nodes): Likewise.
2605 (dump_case_nodes): Likewise.
2606 (switch_decision_tree::dump_case_nodes): Likewise.
2607 (emit_jump): Likewise.
2608 (switch_decision_tree::emit_jump): Likewise.
2609 (emit_cmp_and_jump_insns): Likewise.
2610 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2611 (emit_case_nodes): Likewise.
2612 (switch_decision_tree::emit_case_nodes): Likewise.
2613 (conditional_probability): Remove.
2614 * tree-switch-conversion.h (enum cluster_type): New.
2616 (struct cluster): Likewise.
2617 (cluster::cluster): Likewise.
2618 (struct simple_cluster): Likewise.
2619 (simple_cluster::simple_cluster): Likewise.
2620 (struct group_cluster): Likewise.
2621 (struct jump_table_cluster): Likewise.
2622 (struct bit_test_cluster): Likewise.
2623 (struct min_cluster_item): Likewise.
2624 (struct case_tree_node): Likewise.
2625 (case_tree_node::case_tree_node): Likewise.
2626 (jump_table_cluster::case_values_threshold): Likewise.
2627 (struct case_bit_test): Likewise.
2628 (struct switch_decision_tree): Likewise.
2629 (struct switch_conversion): Likewise.
2630 (switch_decision_tree::reset_out_edges_aux): Likewise.
2632 2018-06-20 Martin Liska <mliska@suse.cz>
2634 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2635 (hoist_edge_and_branch_if_true): Likewise.
2636 (expand_switch_using_bit_tests_p): Likewise.
2637 (struct case_bit_test): Likewise.
2638 (case_bit_test_cmp): Likewise.
2639 (emit_case_bit_tests): Likewise.
2640 (switch_conversion::switch_conversion): New class.
2641 (struct switch_conv_info): Remove old struct.
2642 (collect_switch_conv_info): More to ...
2643 (switch_conversion::collect): ... this.
2644 (check_range): Likewise.
2645 (switch_conversion::check_range): Likewise.
2646 (check_all_empty_except_final): Likewise.
2647 (switch_conversion::check_all_empty_except_final): Likewise.
2648 (check_final_bb): Likewise.
2649 (switch_conversion::check_final_bb): Likewise.
2650 (create_temp_arrays): Likewise.
2651 (switch_conversion::create_temp_arrays): Likewise.
2652 (free_temp_arrays): Likewise.
2653 (gather_default_values): Likewise.
2654 (switch_conversion::gather_default_values): Likewise.
2655 (build_constructors): Likewise.
2656 (switch_conversion::build_constructors): Likewise.
2657 (constructor_contains_same_values_p): Likewise.
2658 (switch_conversion::contains_same_values_p): Likewise.
2659 (array_value_type): Likewise.
2660 (switch_conversion::array_value_type): Likewise.
2661 (build_one_array): Likewise.
2662 (switch_conversion::build_one_array): Likewise.
2663 (build_arrays): Likewise.
2664 (switch_conversion::build_arrays): Likewise.
2665 (gen_def_assigns): Likewise.
2666 (switch_conversion::gen_def_assigns): Likewise.
2667 (prune_bbs): Likewise.
2668 (switch_conversion::prune_bbs): Likewise.
2669 (fix_phi_nodes): Likewise.
2670 (switch_conversion::fix_phi_nodes): Likewise.
2671 (gen_inbound_check): Likewise.
2672 (switch_conversion::gen_inbound_check): Likewise.
2673 (process_switch): Use the newly created class.
2674 (switch_conversion::expand): New.
2675 (switch_conversion::~switch_conversion): New.
2676 * tree-switch-conversion.h: New file.
2678 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2680 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2681 tree-vect-patterns.c.
2682 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2683 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2684 (vect_recog_sad_pattern): Likewise.
2685 (vect_recog_widen_sum_pattern): Likewise.
2686 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2687 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2688 (vect_recog_rotate_pattern): Likewise.
2689 (vect_recog_mult_pattern): Likewise.
2690 (vect_recog_vector_vector_shift_pattern): Likewise.
2691 (vect_recog_divmod_pattern): Likewise.
2692 (vect_recog_mixed_size_cond_pattern): Likewise.
2693 (vect_recog_bool_pattern): Likewise.
2694 (vect_recog_mask_conversion_pattern): Likewise.
2695 (vect_try_gather_scatter_pattern): Likewise.
2696 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2697 (vect_recog_over_widening_pattern): Likewise.
2698 (vect_recog_gather_scatter_pattern): Likewise.
2699 (vect_recog_func_ptr): Move from tree-vectorizer.h
2700 (vect_vect_recog_func_ptrs): Move further down the file.
2701 (vect_recog_func): Likewise. Remove the third argument.
2702 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2703 (vect_pattern_recog_1): Expect the pattern function to do any
2704 necessary target tests. Also expect it to provide a vector type.
2705 Remove the type_in handling.
2707 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2709 * tree-vect-patterns.c (vect_pattern_detected): New function.
2710 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2711 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2712 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2713 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2714 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2715 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2716 (vect_recog_mask_conversion_pattern)
2717 (vect_try_gather_scatter_pattern): Likewise.
2719 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2721 * tree-vect-patterns.c (vect_get_internal_def): New function.
2722 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2723 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2724 (search_type_for_mask_1): Use it.
2726 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2728 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2729 redundant WIDEN_SUM_EXPR handling.
2730 (vect_recog_sad_pattern): Likewise.
2732 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2734 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2735 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2736 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2737 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2738 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2740 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2742 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2743 use the stmt_vec_info of the original bb statement for the
2744 new zero assignment, even if the call is part of a pattern.
2746 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2748 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2749 that the sequence is attached to the original statement rather
2750 than the pattern statement.
2751 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2752 PATTERN_DEF_SEQ from the original statement rather than
2753 the main pattern statement.
2754 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2755 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2756 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2758 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2760 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2761 definition statements before the early exit for statements that aren't
2763 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2765 (vect_transform_loop): ...here. Process pattern definition
2766 statements without first checking whether the main pattern
2767 statement is live or relevant.
2769 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2771 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2772 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2774 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2776 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2777 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2778 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2779 (expand_block_compare): Change select_block_compare_mode call.
2780 (expand_strncmp_align_check): Use new functions, fix comment.
2781 (emit_final_str_compare_gpr): New function.
2782 (expand_strn_compare): Refactor and clean up code.
2783 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2785 2018-06-19 Tony Reix <tony.reix@atos.com>
2786 Damien Bergamini <damien.bergamini@atos.com>
2787 David Edelsohn <dje.gcc@gmail.com>
2789 * collect2.c (static_obj): New variable.
2790 (static_libs): New variable.
2791 (is_in_list): Uncomment declaration.
2792 (main): Track AIX libraries linked statically.
2793 (is_in_list): Uncomment definition.
2794 (scan_prog_file): Don't add AIX shared libraries initializer
2795 to constructor list if linking statically.
2797 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2799 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2801 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2803 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2805 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2808 2018-06-19 Martin Liska <mliska@suse.cz>
2810 * config/i386/i386.c (ix86_can_inline_p): Do not use
2811 ipa_fn_summaries::get_create.
2812 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2814 (devirtualization_time_bonus): Likewise.
2815 (ipcp_propagate_stage): Likewise.
2816 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2817 (edge_set_predicate): Likewise.
2818 (evaluate_conditions_for_known_args): Likewise.
2819 (evaluate_properties_for_edge): Likewise.
2820 (ipa_call_summary::reset): Tranform to ...
2821 (ipa_call_summary::~ipa_call_summary): ... this.
2822 (ipa_fn_summary::reset): Transform to ...
2823 (ipa_fn_summary::~ipa_fn_summary): ... this.
2824 (ipa_fn_summary_t::remove): Rename to ...
2825 (ipa_fn_summary_t::remove_callees): ... this.
2826 (ipa_fn_summary_t::duplicate): Use placement new
2827 instead of memory copy.
2828 (ipa_call_summary_t::duplicate): Likewise.
2829 (ipa_call_summary_t::remove): Remove.
2830 (dump_ipa_call_summary): Change get_create to get.
2831 (ipa_dump_fn_summary): Dump only when summary exists.
2832 (analyze_function_body): Use symbol_summary::get instead
2834 (compute_fn_summary): Likewise.
2835 (estimate_edge_devirt_benefit): Likewise.
2836 (estimate_edge_size_and_time): Likewise.
2837 (inline_update_callee_summaries): Likewise.
2838 (remap_edge_change_prob): Likewise.
2839 (remap_edge_summaries): Likewise.
2840 (ipa_merge_fn_summary_after_inlining): Likewise.
2841 (write_ipa_call_summary): Likewise.
2842 (ipa_fn_summary_write): Likewise.
2843 (ipa_free_fn_summary): Likewise.
2844 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2845 (struct ipa_call_summary): Likewise.
2846 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2848 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2849 (estimate_size_after_inlining): Likewise.
2850 (estimate_growth): Likewise.
2851 (growth_likely_positive): Likewise.
2852 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2853 (inline_call): Likewise.
2854 * ipa-inline.c (caller_growth_limits): Likewise.
2855 (can_inline_edge_p): Likewise.
2856 (can_inline_edge_by_limits_p): Likewise.
2857 (compute_uninlined_call_time): Likewise.
2858 (compute_inlined_call_time): Likewise.
2859 (want_inline_small_function_p): Likewise.
2860 (edge_badness): Likewise.
2861 (update_caller_keys): Likewise.
2862 (update_callee_keys): Likewise.
2863 (inline_small_functions): Likewise.
2864 (inline_to_all_callers_1): Likewise.
2865 (dump_overall_stats): Likewise.
2866 (early_inline_small_functions): Likewise.
2867 (early_inliner): Likewise.
2868 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2869 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2870 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2871 * ipa-split.c (execute_split_functions): Likewise.
2872 * symbol-summary.h: Likewise.
2873 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2875 2018-06-19 Richard Biener <rguenther@suse.de>
2877 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2878 (vectorize_loops): ... here. Fix dbgcnt handling.
2879 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2881 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2884 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2885 ieee128 argument takes up only one (vector) register, not two (floating
2888 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2890 * gimplify.c (gimplify_init_constructor): Really never clear for an
2891 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2893 2018-06-19 Richard Biener <rguenther@suse.de>
2895 PR tree-optimization/86179
2896 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2897 after failed recognition.
2899 2018-06-18 Martin Sebor <msebor@redhat.com>
2902 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2903 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2904 Handle integer subtraction.
2905 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2906 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2908 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2910 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2911 param from rtx to rtx_insn *.
2912 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2914 (frv_ifcvt_modify_insn): Likwise.
2915 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2916 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2917 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2918 as_a <rtx_insn *> cast to local "unprotected_region" once
2919 it's been established that it's not NULL or pc_rtx.
2920 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2921 param "sethi" from rtx to rtx_insn *.
2922 (nds32_group_float_insns): Likewise for param "insn".
2923 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2925 (vax_output_int_subtract): Likewise.
2926 * config/vax/vax.c (vax_output_int_add): Likewise for param
2928 (vax_output_int_subtract): Likewise.
2929 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2930 (emit_pattern_after): Likewise for param "after".
2931 (emit_insn_after): Likewise.
2932 (emit_jump_insn_after): Likewise.
2933 (emit_call_insn_after): Likewise.
2934 (emit_debug_insn_after): Likewise.
2935 (emit_pattern_before): Likewise for param "before".
2936 (emit_insn_before): Likewise.
2937 (emit_jump_insn_before): Likewise.
2938 * final.c (get_insn_template): Likewise for param "insn", removing
2940 * output.h (get_insn_template): Likewise for 2nd param.
2941 * rtl.h (emit_insn_before): Likewise.
2942 (emit_jump_insn_before): Likewise.
2943 (emit_debug_insn_before_noloc): Likewise.
2944 (emit_insn_after): Likewise.
2945 (emit_jump_insn_after): Likewise.
2946 (emit_call_insn_after): Likewise.
2947 (emit_debug_insn_after): Likewise.
2948 (set_insn_deleted): Likewise for param.
2950 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2953 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2954 floating point modes, so that IFmode is numerically greater than
2955 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2956 to declare the ordering. This prevents IFmode from being
2957 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2958 machine. Include rs6000-modes.h to share the fractional values
2959 between genmodes* and the rest of the compiler.
2963 * config/rs6000/rs6000-modes.h: New file.
2964 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2965 meaning of rs6000_long_double_size so that 126..128 selects an
2966 appropriate 128-bit floating point type.
2967 (rs6000_option_override_internal): Likewise.
2968 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2969 (TARGET_LONG_DOUBLE_128): Change the meaning of
2970 rs6000_long_double_size so that 126..128 selects an appropriate
2971 128-bit floating point type.
2972 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2973 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2974 source and destination to match the standard usage.
2975 (truncifkf2): Likewise.
2976 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2977 ISA 2.07 to use an explicit clobber, instead of passing in a
2979 (copysign<mode>3_soft): Likewise.
2981 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2983 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2984 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2985 (vect_slp_analyze_instance_dependence): Likewise.
2986 (vect_enhance_data_refs_alignment): Likewise.
2987 (vect_analyze_data_refs_alignment): Likewise.
2988 (vect_slp_analyze_and_verify_instance_alignment
2989 (vect_analyze_data_ref_accesses): Likewise.
2990 (vect_prune_runtime_alias_test_list): Likewise.
2991 (vect_analyze_data_refs): Likewise.
2992 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2993 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2994 (vect_analyze_scalar_cycles_1): Likewise.
2995 (vect_get_loop_niters): Likewise.
2996 (vect_analyze_loop_form_1): Likewise.
2997 (vect_update_vf_for_slp): Likewise.
2998 (vect_analyze_loop_operations): Likewise.
2999 (vect_analyze_loop): Likewise.
3000 (vectorizable_induction): Likewise.
3001 (vect_transform_loop): Likewise.
3002 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3003 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3004 (vect_make_slp_decision): Likewise.
3005 (vect_detect_hybrid_slp): Likewise.
3006 (vect_slp_analyze_operations): Likewise.
3007 (vect_slp_bb): Likewise.
3008 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3009 (vectorizable_bswap): Likewise.
3010 (vectorizable_call): Likewise.
3011 (vectorizable_simd_clone_call): Likewise.
3012 (vectorizable_conversion): Likewise.
3013 (vectorizable_assignment): Likewise.
3014 (vectorizable_shift): Likewise.
3015 (vectorizable_operation): Likewise.
3016 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3018 2018-06-18 Martin Sebor <msebor@redhat.com>
3020 PR tree-optimization/81384
3021 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3022 * builtins.c (expand_builtin_strnlen): New function.
3023 (expand_builtin): Call it.
3024 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3025 * builtins.def (BUILT_IN_STRNLEN): New.
3026 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3027 Warn for bounds in excess of maximum object size.
3028 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3029 single-value ranges. Handle strnlen.
3030 (handle_builtin_strlen): Handle strnlen.
3031 (strlen_check_and_optimize_stmt): Same.
3032 * doc/extend.texi (Other Builtins): Document strnlen.
3034 2018-06-18 Maya Rashish <coypu@sdf.org>
3036 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3037 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3038 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3040 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3042 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3044 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3046 * tree.c (escaped_string::escape): Replace cast to char * by
3047 const_cast<char *> (unescaped).
3049 2018-06-18 Nick Clifton <nickc@redhat.com>
3052 * tree.c (escaped_string): New class. Converts an unescaped
3053 string into its escaped equivalent.
3054 (warn_deprecated_use): Use the new class to convert the
3055 deprecation message, if present.
3056 (test_escaped_strings): New self test.
3057 (test_c_tests): Add test_escaped_strings.
3058 * doc/extend.texi (deprecated): Add a note that the
3059 deprecation message is affected by the -fmessage-length
3060 option, and that control characters will be escaped.
3061 (#pragma GCC error): Document this pragma.
3062 (#pragma GCC warning): Likewise.
3063 * doc/invoke.texi (-fmessage-length): Document this option's
3064 effect on the #warning and #error preprocessor directives and
3065 the deprecated attribute.
3067 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3069 * tree.c (decl_value_expr_lookup): Revert latest change.
3070 (decl_value_expr_insert): Likewise.
3072 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3074 * gimplify.c (nonlocal_vlas): Delete.
3075 (nonlocal_vla_vars): Likewise.
3076 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3078 (gimplify_body): Do not create and destroy nonlocal_vlas.
3079 * tree-nested.c: Include diagnostic.h.
3080 (use_pointer_in_frame): Tweak.
3081 (lookup_field_for_decl): Add assertion and declare the transformation.
3082 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3083 internal error when the reference is in a wrong context. Do not
3084 create a debug decl by default.
3085 (note_nonlocal_block_vlas): Delete.
3086 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3087 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3088 create a debug decl by default.
3089 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3090 call is in a wrong context.
3091 (fixup_vla_decls): New function.
3092 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3093 debug variables were created.
3094 * tree.c (decl_value_expr_lookup): Add checking assertion.
3095 (decl_value_expr_insert): Likewise.
3097 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3100 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3101 * tree-scalar-evolution.c (interpret_expr): Likewise.
3102 (expression_expensive_p): Likewise.
3103 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3104 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3105 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3106 (ssa_defined_by_minus_one_stmt_p): New.
3108 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3111 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3112 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3113 * dojump.c (do_jump): Likewise.
3114 * expr.c (expand_expr_real_2): Check operand type's sign.
3115 * fold-const.c (const_unop): Handle ABSU_EXPR.
3116 (fold_abs_const): Likewise.
3117 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3118 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3119 (strip_sign_op_1): Likesise.
3120 * match.pd: Add new pattern to generate ABSU_EXPR.
3121 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3122 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3123 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3124 * tree-inline.c (estimate_operator_cost): Likewise.
3125 * tree-pretty-print.c (dump_generic_node): Likewise.
3126 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3127 * tree.def (ABSU_EXPR): New.
3129 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3132 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3133 documented as preserved for backward compatibility only.
3134 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3136 PR rtl-optimization/86108
3137 * bb-reorder.c (create_forwarder_block): Renamed to ...
3138 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3139 jump from new landing pad to the second part.
3140 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3143 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3146 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3147 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3148 Only call store_expr for halves if the mode is the same.
3151 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3152 Fix up comment formatting.
3154 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3156 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3157 (typed_splay_tree::closure,
3158 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3159 (typed_splay_tree::typed_splay_tree,
3160 typed_splay_tree::operator =): Declared private.
3161 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3162 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3163 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3164 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3165 typed_splay_tree::splay_tree_splay,
3166 typed_splay_tree::splay_tree_foreach_helper,
3167 typed_splay_tree::splay_tree_insert,
3168 typed_splay_tree::splay_tree_remove,
3169 typed_splay_tree::splay_tree_lookup,
3170 typed_splay_tree::splay_tree_predecessor,
3171 typed_splay_tree::splay_tree_successor,
3172 typed_splay_tree::splay_tree_min,
3173 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3174 (typed_splay_tree::root, typed_splay_tree::comp,
3175 typed_splay_tree::delete_key,
3176 typed_splay_tree::delete_value): New data members.
3177 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3178 typed_splay_tree::remove.
3180 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3182 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3183 -mginv and -mno-ginv to the assembler.
3184 * config/mips/mips.opt (-mcrc): New option.
3186 * doc/invoke.text (-mcrc): Document.
3189 2018-06-15 Nick Clifton <nickc@redhat.com>
3192 * tree.c (escaped_string): New class. Converts an unescaped
3193 string into its escaped equivalent.
3194 (warn_deprecated_use): Use the new class to convert the
3195 deprecation message, if present.
3196 (test_escaped_strings): New self test.
3197 (test_c_tests): Add test_escaped_strings.
3198 * doc/extend.texi (deprecated): Add a note that the
3199 deprecation message is affected by the -fmessage-length
3200 option, and that control characters will be escaped.
3201 (#pragma GCC error): Document this pragma.
3202 (#pragma GCC warning): Likewise.
3203 * doc/invoke.texi (-fmessage-length): Document this option's
3204 effect on the #warning and #error preprocessor directives and
3205 the deprecated attribute.
3207 2018-06-15 Richard Biener <rguenther@suse.de>
3209 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3210 here, also noting vector size used.
3211 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3212 size used in MSG_OPTIMIZED_LOCATIONS dump.
3213 (pass_slp_vectorize::execute): Adjust.
3215 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3218 * config/arc/arc.c (arc_return_address_register): Fix
3221 2018-06-15 Richard Biener <rguenther@suse.de>
3224 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3225 leave useless conversion stripping to force_gimple_operand_gsi.
3226 (gimplify_build2): Likewise.
3227 (gimplify_build1): Likewise.
3229 2018-06-15 Richard Biener <rguenther@suse.de>
3232 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3233 before adjusting their block.
3235 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3237 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3238 multilibs for *-*-rtems*.
3239 * config/riscv/t-rtems: New file.
3241 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3244 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3245 unsigned_type_for returns NULL.
3248 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3249 subregs of multi-word pseudos unless the float mode has word size.
3251 2018-06-14 Richard Biener <rguenther@suse.de>
3254 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3255 duplicate and harmful type_hash_canon.
3256 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3258 2018-06-14 Richard Biener <rguenther@suse.de>
3261 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3264 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3266 * config/rtems.h (STDINT_LONG32): Define.
3268 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3269 Prachi Godbole <prachi.godbole@imgtec.com>
3271 * config/mips/mips-cpus.def: Define P6600.
3272 * config/mips/mips-tables.opt: Regenerate.
3273 * config/mips/mips.c (mips_ucbranch_type): New enum.
3274 (mips_rtx_cost_data): Add support for P6600.
3275 (mips_issue_rate): Likewise.
3276 (mips_multipass_dfa_lookahead): Likewise.
3277 (mips_avoid_hazard): Likewise.
3278 (mips_reorg_process_insns): Likewise.
3279 (mips_classify_branch_p6600): New function.
3280 * config/mips/mips.h (TUNE_P6600): New define.
3281 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3282 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3283 * config/mips/mips.md: Include p6600.md.
3284 (processor): Add p6600.
3285 * config/mips/p6600.md: New file.
3286 * doc/invoke.texi: Add p6600 to supported architectures.
3288 2018-06-13 Martin Sebor <msebor@redhat.com>
3290 PR tree-optimization/86114
3291 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3293 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3295 2018-06-13 Richard Biener <rguenther@suse.de>
3297 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3298 Properly set vector type of the intermediate stmt.
3299 * tree-vect-stmts.c (vectorizable_operation): The destination
3300 var always has vectype_out type.
3302 2018-06-13 Jeff Law <law@redhat.com>
3304 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3305 integer 0 for argument to print_rtl_with_bb.
3306 (rl78_reorg): Likewise.
3308 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3310 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3311 from rtx to rtx_insn *.
3312 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3314 (add_sched_insns_for_speculation): Likewise for local "target",
3315 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3316 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3317 from rtx_insn ** to rtx_code_label **.
3318 (reorg_emit_nops): Likewise.
3319 (c6x_reorg): Likewise for local "call_labels".
3320 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3322 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3323 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3324 the loops over LABEL_REFS.
3325 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3327 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3328 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3329 (split_branches): Strengthen local "olabel" from rtx to
3330 rtx_insn *, adding a safe_as_a cast.
3331 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3333 (add_insn_after): Likewise for first two params.
3334 (add_insn_before): Likewise.
3335 (remove_insn): Likewise for param.
3336 (emit_pattern_before_noloc): Likewise for second and third params.
3337 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3338 (emit_call_insn_before_noloc): Likewise.
3339 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3341 (emit_barrier_before): Likewise.
3342 (emit_label_before): Strengthen "label" param from "rtx" to
3343 "rtx_code_label *". Strengthen "before" param from "rtx" to
3345 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3347 (emit_pattern_after_noloc): Likewise.
3348 (emit_insn_after_noloc): Likewise.
3349 (emit_jump_insn_after_noloc): Likewise.
3350 (emit_call_insn_after_noloc): Likewise.
3351 (emit_debug_insn_after_noloc): Likewise.
3352 (emit_barrier_after): Likewise.
3353 (emit_label_after): Likewise for both params.
3354 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3355 "loc" param from "int" to "location_t".
3356 (emit_insn_after_setloc): Likewise.
3357 (emit_jump_insn_after_setloc): Likewise.
3358 (emit_call_insn_after_setloc): Likewise.
3359 (emit_debug_insn_after_setloc): Likewise.
3360 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3361 "loc" param from "int" to "location_t".
3362 (emit_pattern_before): Convert NULL_RTX to NULL.
3363 (emit_insn_before_setloc): Convert "loc" param from "int" to
3365 (emit_jump_insn_before_setloc): Likewise.
3366 (emit_call_insn_before_setloc): Likewise.
3367 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3368 rtx_insn *. Convert "loc" param from "int" to "location_t".
3369 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3370 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3371 Convert 3rd param from "int" to "location_t".
3372 (emit_barrier_before, emit_barrier_after, next_real_insn):
3373 Strengthen param from rtx to rtx_insn *.
3374 (emit_label_before): Strengthen 1st param from "rtx" to
3375 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3377 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3378 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3379 Strengthen 2nd param from "rtx" to "rtx_insn *".
3380 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3381 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3382 Likewise. Convert 3rd param from "int" to "location_t".
3383 (emit_label_after): Strengthen 1st param from "rtx" to
3385 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3387 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3388 from "rtx" to "rtx_insn *".
3390 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3392 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3393 bodies streamed in with -Q.
3394 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3395 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3396 * gimple-streamer-out.c: Include gimple-pretty-print.h
3397 (output_bb): Dump stmts streamed.
3398 * lto-section-out.c: Include print-tree.h
3399 (lto_begin_section): Dump sections created.
3400 (lto_output_decl_index): Dump decl encoded.
3401 * lto-streamer-out.c: Include print-tree.h
3402 (create_output_block): Dump output block created.
3403 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3404 (output_function): Dump function output.
3405 (output_constructor): Dump constructor streamed.
3406 (write_global_stream): Output indexes encoded.
3407 (produce_asm_for_decls): Dump streams encoded.
3408 * lto-streamer.c (streamer_dump_file): New global var.
3409 * lto-streamer.h (streamer_dump_file): Declare.
3410 * passes.c (ipa_write_summaries): Initialize streamer dump.
3411 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3414 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3417 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3418 offsets for register save directives. Emit a second batch of save
3419 directives, if need be, when the function accesses prior frames.
3421 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3423 * config/arc/fpu.md (fmasf4): Force operand to register.
3424 (fnmasf4): Likewise.
3426 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3428 * config/arc/arc-protos.h (arc_pad_return): Remove.
3429 * config/arc/arc.c (machine_function): Remove force_short_suffix
3431 (arc_print_operand): Adjust printing of '&'.
3432 (arc_verify_short): Remove conditional printing of short suffix.
3433 (arc_final_prescan_insn): Remove reference to size_reason.
3434 (pad_return): New function.
3435 (arc_reorg): Call pad_return.
3436 (arc_pad_return): Remove.
3437 (arc_init_machine_status): Remove reference to force_short_suffix.
3438 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3439 (attr length): When attribute iscompact is true force to 2
3440 regardless; in the case of maybe check if we want to force the
3441 instruction to have 4 bytes length.
3442 (nopv): Change it to generate 4 byte long nop as well.
3443 (blockage): New pattern.
3444 (simple_return): Remove call to arc_pad_return.
3445 (p_return_i): Likewise.
3447 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3449 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3451 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3453 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3456 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3458 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3459 for ARC700 and ARCv2.
3461 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3464 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3465 operands[2] instead of operands[1].
3468 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3470 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3471 case, check whether the outer register overlaps an unallocatable
3472 register, not just whether it fits the required class.
3474 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3476 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3478 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3479 (memrefs_conflict_p): Likewise.
3480 (init_alias_analysis): Likewise.
3481 * cfgexpand.c (expand_debug_expr): Likewise.
3482 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3483 * cse.c (fold_rtx): Likewise.
3484 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3485 * expr.c (emit_block_move_hints): Likewise.
3486 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3487 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3488 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3489 (emit_group_store): Likewise.
3490 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3491 to read the PRE/POST_MODIFY increment.
3492 * calls.c (store_one_arg): Use strip_offset.
3493 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3495 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3497 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3498 (simplify_binary_operation_1): Extend CONST_INT handling to
3500 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3501 than a HOST_WIDE_INT.
3502 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3504 (adjust_mems, add_stores): Update accodingly.
3505 (vt_canonicalize_addr): Track polynomial offsets.
3506 (emit_note_insn_var_location): Likewise.
3507 (vt_add_function_parameter): Likewise.
3508 (vt_initialize): Likewise.
3510 2018-06-12 Jeff Law <law@redhat.com>
3512 * config.gcc (alpha*-*-freebsd*): Remove.
3513 * config/alpha/freebsd.h: Remove.
3515 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3518 * spellcheck-tree.c (levenshtein_distance): Rename to...
3519 (get_edit_distance): ...this, and update for underlying renaming.
3520 * spellcheck-tree.h (levenshtein_distance): Rename to...
3521 (get_edit_distance): ...this.
3522 * spellcheck.c (levenshtein_distance): Rename to...
3523 (get_edit_distance): ...this. Convert from Levenshtein distance
3524 to Damerau-Levenshtein distance by supporting transpositions of
3525 adjacent characters. Rename "v1" to "v_next" and "v0" to
3527 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3528 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3529 update for underlying renaming.
3530 (selftest::levenshtein_distance_unit_test): Rename to...
3531 (selftest::test_get_edit_distance_unit): ...this, and update for
3532 underlying renaming.
3533 (selftest::test_find_closest_string): Add example from PR 69968
3534 where transposition helps
3535 (selftest::test_metric_conditions): Update for renaming.
3536 (selftest::test_metric_conditions): Likewise.
3537 (selftest::spellcheck_c_tests): Likewise.
3538 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3539 (get_edit_distance): ...this.
3540 (best_match::consider): Update for renaming.
3542 2018-06-12 Martin Sebor <msebor@redhat.com>
3544 PR tree-optimization/85259
3545 * builtins.c (compute_objsize): Handle constant offsets.
3546 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3547 true iff a warning has been issued.
3548 * gimple.h (gimple_nonartificial_location): New function.
3549 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3550 gimple_nonartificial_location and handle -Wno-system-headers.
3551 (handle_builtin_stxncpy): Same.
3553 2018-06-12 Martin Sebor <msebor@redhat.com>
3556 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3558 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3560 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3561 BUILTIN_VEC_XST entries for pointer to double and long long.
3563 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3566 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3568 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3571 2018-06-12 Martin Liska <mliska@suse.cz>
3573 * doc/options.texi: Document IntegerRange.
3575 2018-06-12 Martin Liska <mliska@suse.cz>
3577 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3578 * opt-functions.awk: Handle Deprecated flag.
3579 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3581 (read_cmdline_option): Report warning for a deprecated option.
3582 * opts.h (struct cl_option): Add new field cl_deprecated.
3583 (CL_ERR_DEPRECATED): New.
3585 2018-06-12 Martin Liska <mliska@suse.cz>
3587 * doc/options.texi: Document Deprecated option flag.
3589 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3591 * config/arc/arc-arch.h (arc_extras): New enum.
3592 (arc_cpu_t):Add field extra.
3593 (arc_cpu_types): Consider the extras.
3594 * config/arc/arc-cpus.def: Add extras info.
3595 * config/arc/arc-opts.h (processor_type): Consider extra field.
3596 * config/arc/arc.c (arc_override_options): Handle extra field.
3598 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3600 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3601 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3603 * config/arc/arc.c (arc_init): Update pic warning.
3604 (irq_range): Update irq range parsing warnings.
3605 (arc_override_options): Update various warning messages.
3606 (arc_handle_aux_attribute): Likewise.
3608 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3610 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3612 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3614 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3615 compiler messages directives.
3617 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3619 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3620 * config/mips/mips-tables.opt: Regenerate.
3621 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3623 * doc/invoke.texi: Document -march=i6500.
3625 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3627 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3628 (i6400_gpmul): Add cpu_unit.
3629 (i6400_gpdiv): Likewise.
3630 (i6400_msa_add_d): Update reservations.
3631 (i6400_msa_int_add) Likewise.
3632 (i6400_msa_short_logic3) Likewise.
3633 (i6400_msa_short_logic2) Likewise.
3634 (i6400_msa_short_logic) Likewise.
3635 (i6400_msa_move) Likewise.
3636 (i6400_msa_cmp) Likewise.
3637 (i6400_msa_short_float2) Likewise.
3638 (i6400_msa_div_d) Likewise.
3639 (i6400_msa_long_logic1) Likewise.
3640 (i6400_msa_long_logic2) Likewise.
3641 (i6400_msa_mult) Likewise.
3642 (i6400_msa_long_float2) Likewise.
3643 (i6400_msa_long_float4) Likewise.
3644 (i6400_msa_long_float5) Likewise.
3645 (i6400_msa_long_float8) Likewise.
3646 (i6400_fpu_fadd): Include frint type.
3647 (i6400_fpu_store): New define_insn_reservation.
3648 (i6400_fpu_load): Likewise.
3649 (i6400_fpu_move): Likewise.
3650 (i6400_fpu_fcmp): Likewise.
3651 (i6400_fpu_fmadd): Likewise.
3652 (i6400_int_mult): Include imul3nc type and update reservation.
3653 (i6400_int_div): Include idiv3 type and update reservation.
3654 (i6400_int_load): Update to check type not move_type.
3655 (i6400_int_store): Likewise.
3656 (i6400_int_prefetch): Set zero latency.
3658 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3660 * gcc.c: Document new %@{...} sequence.
3661 (LINK_COMMAND_SPEC): Use it for the -L switches.
3662 (cpp_unique_options): Use it for the -I switches.
3663 (at_file_argbuf): New global variable.
3664 (in_at_file): Likewise.
3665 (alloc_args): Create at_file_argbuf.
3666 (clear_args): Truncate at_file_argbuf.
3667 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3668 (open_at_file): New function.
3669 (close_at_file): Likewise.
3670 (create_at_file): Delete.
3671 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3674 (validate_switches_from_spec): Deal with %@{...} sequence.
3675 (validate_switches): Likewise.
3676 (driver::finalize): Call clear_args.
3678 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3680 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3682 2018-06-11 Martin Sebor <msebor@redhat.com>
3684 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3685 (Wno-class-memaccess): Add @opindex.
3686 (Wno-templates, Wno-multiple-inheritance): Same.
3687 (Wno-virtual-inheritance, Wno-namespaces): Same.
3688 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3689 (Wno-format-overflow, Wno-format-truncation): Same.
3690 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3691 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3692 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3693 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3694 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3695 (Wno-misspelled-isr): Same.
3697 2018-06-11 Martin Sebor <msebor@redhat.com>
3699 * PR tree-optimization/86083
3700 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3702 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3704 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3706 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3709 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3710 on the correct operand.
3711 (*movdi_internal64): Ditto.
3713 2018-06-11 Martin Liska <mliska@suse.cz>
3715 PR tree-optimization/86089
3716 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3718 2018-06-11 Julia Koval <julia.koval@intel.com>
3720 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3721 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3722 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3724 2018-06-11 Olivier Hainque <hainque@adacore.com>
3726 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3727 for Ada with strict dwarf2.
3729 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3732 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3735 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3737 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3739 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3741 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3744 2018-06-08 Martin Liska <mliska@suse.cz>
3746 * tree-cfg.h (debug_function): Fix argument type to match
3749 2018-06-08 Martin Liska <mliska@suse.cz>
3751 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3752 Remove usage of MPX-related (and removed) fields.
3753 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3755 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3757 * cfg.c (debug): Use TDF_NONE rather than 0.
3758 * cfghooks.c (debug): Likewise.
3759 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3760 (struct dump_option_value_info): Convert to...
3761 (struct kv_pair): ...this template type.
3762 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3764 (optinfo_verbosity_options): Likewise.
3765 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3767 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3768 than int for "optgroup_flags" param.
3769 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3771 (dump_dec): Likewise.
3772 (dump_finish): Use TDF_NONE rather than 0.
3773 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3774 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3775 than 0. Update for change to option_ptr.
3776 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3777 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3778 0. Update for changes to optinfo_verbosity_options and
3780 (opt_info_switch_p): Convert optgroup_flags from int to
3782 (dump_basic_block): Use dump_flags_t rather than int
3783 for "dump_kind" param.
3784 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3785 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3786 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3787 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3788 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3789 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3790 TDF_NONE): Convert from macros to...
3791 (enum dump_flag): ...this new enum.
3792 (dump_flags_t): Update to use enum.
3793 (operator|, operator&, operator~, operator|=, operator&=):
3794 Implement for dump_flags_t.
3795 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3796 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3797 Convert from macros to...
3798 (enum optgroup_flag): ...this new enum.
3799 (optgroup_flags_t): New typedef.
3800 (operator|, operator|=): Implement for optgroup_flags_t.
3801 (struct dump_file_info): Convert field "alt_flags" to
3802 dump_flags_t. Convert field "optgroup_flags" to
3804 (dump_basic_block): Use dump_flags_t rather than int for param.
3805 (dump_generic_expr_loc): Likewise.
3806 (dump_dec): Likewise.
3807 (dump_register): Convert param "optgroup_flags" to
3809 (opt_info_enable_passes): Likewise.
3810 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3812 * gimple-pretty-print.c (debug): Likewise.
3813 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3814 (merged_store_group::apply_stores): Likewise.
3815 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3816 * gimple.c (verify_gimple_pp): Likewise.
3817 * graphite-poly.c (print_pbb_body): Likewise.
3818 * passes.c (pass_manager::register_one_dump_file): Convert
3819 local "optgroup_flags" to optgroup_flags_t.
3820 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3822 (debug_body): Likewise.
3823 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3824 to optgroup_flags_t.
3825 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3827 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3828 (convert_mult_to_fma): Likewise.
3829 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3830 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3831 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3832 "dump_kind" to dump_flags_t.
3834 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3836 * config/rs6000/rs6000.c (min, max): Delete.
3838 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3840 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3841 -mabi=spe and -mabi=no-spe.
3843 2018-06-08 Martin Liska <mliska@suse.cz>
3845 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3846 where we expect an existing summary.
3848 2018-06-08 Martin Liska <mliska@suse.cz>
3850 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3851 * ipa-inline.h (estimate_edge_growth): Likewise.
3853 2018-06-08 Martin Liska <mliska@suse.cz>
3855 * cgraph.c (function_version_hasher::hash): Use
3856 cgraph_node::get_uid ().
3857 (function_version_hasher::equal):
3858 * cgraph.h (cgraph_node::get_uid): New method.
3859 * ipa-inline.c (update_caller_keys): Use
3860 cgraph_node::get_uid ().
3861 (update_callee_keys): Likewise.
3862 * ipa-utils.c (searchc): Likewise.
3863 (ipa_reduced_postorder): Likewise.
3864 * lto-cgraph.c (input_node): Likewise.
3865 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3866 * symbol-summary.h (symtab_insertion): Likewise.
3867 (symtab_removal): Likewise.
3868 (symtab_duplication): Likewise.
3869 * tree-pretty-print.c (dump_function_header): Likewise.
3870 * tree-sra.c (convert_callers_for_node): Likewise.
3872 2018-06-08 Martin Liska <mliska@suse.cz>
3874 * cgraph.c (symbol_table::create_edge): Always assign a new
3876 (symbol_table::free_edge): Do not recycle numbers.
3877 * cgraph.h (cgraph_edge::get): New method.
3878 * symbol-summary.h (symtab_removal): Use it.
3879 (symtab_duplication): Likewise.
3880 (call_summary::hashable_uid): Remove.
3882 2018-06-08 Martin Liska <mliska@suse.cz>
3884 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3885 (initialize_growth_caches): Remove.
3886 (free_growth_caches): Likewise.
3887 (do_estimate_edge_time): Use edge_growth_cache.
3888 (do_estimate_edge_size): Likewise.
3889 (do_estimate_edge_hints): Likewise.
3890 * ipa-inline.c (reset_edge_caches): Likewise.
3891 (recursive_inlining): Likewise.
3892 (inline_small_functions): Likewise.
3893 * ipa-inline.h (initialize_growth_caches): Remove.
3894 (estimate_edge_size): Likewise.
3895 (estimate_edge_time): Likewise.
3896 (estimate_edge_hints): Likewise.
3897 (reset_edge_growth_cache): Likewise.
3898 * symbol-summary.h (call_summary::remove): New method.
3900 2018-06-08 Martin Liska <mliska@suse.cz>
3902 * ipa-cp.c (class edge_clone_summary): New summary.
3903 (grow_edge_clone_vectors): Remove.
3904 (ipcp_edge_duplication_hook): Remove.
3905 (class edge_clone_summary_t): New call_summary class.
3906 (ipcp_edge_removal_hook): Remove.
3907 (edge_clone_summary_t::duplicate): New function.
3908 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3909 (create_specialized_node): Likewise.
3910 (ipcp_driver): Initialize edge_clone_summaries and do not
3913 2018-06-08 Martin Liska <mliska@suse.cz>
3915 * symbol-summary.h (get): New function.
3916 (call_summary::m_initialize_when_cloning): New class member.
3918 2018-06-08 Martin Liska <mliska@suse.cz>
3920 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3921 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3922 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3923 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3924 (remove_cgraph_node_from_order): Use the removed_nodes set.
3925 (do_per_function_toporder): Likwise.
3926 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3927 instead of summary_uid.
3928 (symtab_removal): Likewise.
3929 (symtab_duplication): Likewise.
3931 2018-06-08 Martin Liska <mliska@suse.cz>
3933 * ipa-cp.c (ipcp_store_bits_results): Use
3934 ipcp_transformation_sum.
3935 (ipcp_store_vr_results): Likewise.
3936 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3938 (ipcp_transformation_initialize): ... this.
3939 (ipa_set_node_agg_value_chain):
3940 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3941 (write_ipcp_transformation_info): Likewise.
3942 (read_ipcp_transformation_info): Likewise.
3943 (ipcp_update_bits): Likewise.
3944 (ipcp_update_vr): Likewise.
3945 (ipcp_transform_function): Likewise.
3946 * ipa-prop.h: Rename ipcp_transformation_summary to
3947 ipcp_transformation.
3948 (class ipcp_transformation_t): New function summary.
3949 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3950 (ipa_get_agg_replacements_for_node): Likewise.
3952 2018-06-08 Martin Liska <mliska@suse.cz>
3954 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3956 (class funct_state_summary_t): New function_summary class.
3957 (has_function_state): Remove.
3958 (get_function_state): Likewise.
3959 (set_function_state): Likewise.
3960 (add_new_function): Likewise.
3961 (funct_state_summary_t::insert): New function.
3962 (duplicate_node_data): Remove.
3963 (remove_node_data): Remove.
3964 (funct_state_summary_t::duplicate): New function.
3965 (register_hooks): Create new funct_state_summaries.
3966 (pure_const_generate_summary): Use it.
3967 (pure_const_write_summary): Likewise.
3968 (pure_const_read_summary): Likewise.
3969 (propagate_pure_const): Likewise.
3970 (propagate_nothrow): Likewise.
3971 (dump_malloc_lattice): Likewise.
3972 (propagate_malloc): Likewise.
3973 (execute): Do not register hooks, just remove summary
3975 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3978 2018-06-08 Martin Liska <mliska@suse.cz>
3980 * ipa-reference.c (remove_node_data): Remove.
3981 (duplicate_node_data): Likewise.
3982 (class ipa_ref_var_info_summary_t): New class.
3983 (class ipa_ref_opt_summary_t): Likewise.
3984 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3985 (get_reference_optimization_summary): Use
3986 ipa_ref_opt_sum_summaries.
3987 (set_reference_vars_info): Remove.
3988 (set_reference_optimization_summary): Likewise.
3989 (ipa_init): Create summaries.
3990 (init_function_info): Use function summary.
3991 (ipa_ref_opt_summary_t::duplicate): New function.
3992 (ipa_ref_opt_summary_t::remove): New function.
3993 (get_read_write_all_from_node): Fix GNU coding style.
3994 (propagate): Use function summary.
3995 (write_node_summary_p): Fix GNU coding style.
3996 (stream_out_bitmap): Likewise.
3997 (ipa_reference_read_optimization_summary): Use function summary.
3998 (ipa_reference_c_finalize): Do not release hooks.
4000 2018-06-08 Martin Liska <mliska@suse.cz>
4002 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4003 (analyze_function_body): Extract multiple calls of get_create.
4004 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4005 * ipa-inline.c (recursive_inlining): Use ::get method.
4006 * ipa-inline.h (estimate_edge_growth): Likewise.
4008 2018-06-08 Martin Liska <mliska@suse.cz>
4010 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4012 (hsa_function_summary::hsa_function_summary): Use the new enum
4014 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4015 * hsa-gen.c (hsa_get_host_function): Likewise.
4016 (get_brig_function_name): Likewise.
4017 * ipa-hsa.c (process_hsa_functions): Likewise.
4018 (ipa_hsa_write_summary): Likewise.
4019 * symbol-summary.h (symtab_duplication): Use ::get function/
4020 (get): New function.
4022 2018-06-08 Martin Liska <mliska@suse.cz>
4024 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4026 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4027 (hsa_register_kernel): Likewise.
4028 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4029 * hsa-gen.c (hsa_get_host_function): Likewise.
4030 (get_brig_function_name): Likewise.
4031 (generate_hsa): Likewise.
4032 (pass_gen_hsail::execute): Likewise.
4033 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4034 (devirtualization_time_bonus): Likewise.
4035 (ipcp_propagate_stage): Likewise.
4036 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4037 (edge_set_predicate): Likewise.
4038 (evaluate_conditions_for_known_args): Likewise.
4039 (evaluate_properties_for_edge): Likewise.
4040 (ipa_fn_summary::reset): Likewise.
4041 (ipa_fn_summary_t::duplicate): Likewise.
4042 (dump_ipa_call_summary): Likewise.
4043 (ipa_dump_fn_summary): Likewise.
4044 (analyze_function_body): Likewise.
4045 (compute_fn_summary): Likewise.
4046 (estimate_edge_devirt_benefit): Likewise.
4047 (estimate_edge_size_and_time): Likewise.
4048 (estimate_calls_size_and_time): Likewise.
4049 (estimate_node_size_and_time): Likewise.
4050 (inline_update_callee_summaries): Likewise.
4051 (remap_edge_change_prob): Likewise.
4052 (remap_edge_summaries): Likewise.
4053 (ipa_merge_fn_summary_after_inlining): Likewise.
4054 (ipa_update_overall_fn_summary): Likewise.
4055 (read_ipa_call_summary): Likewise.
4056 (inline_read_section): Likewise.
4057 (write_ipa_call_summary): Likewise.
4058 (ipa_fn_summary_write): Likewise.
4059 (ipa_free_fn_summary): Likewise.
4060 * ipa-hsa.c (process_hsa_functions): Likewise.
4061 (ipa_hsa_write_summary): Likewise.
4062 (ipa_hsa_read_section): Likewise.
4063 * ipa-icf.c (sem_function::merge): Likewise.
4064 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4065 (do_estimate_edge_time): Likewise.
4066 (estimate_size_after_inlining): Likewise.
4067 (estimate_growth): Likewise.
4068 (growth_likely_positive): Likewise.
4069 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4070 (inline_call): Likewise.
4071 * ipa-inline.c (caller_growth_limits): Likewise.
4072 (can_inline_edge_p): Likewise.
4073 (can_inline_edge_by_limits_p): Likewise.
4074 (compute_uninlined_call_time): Likewise.
4075 (compute_inlined_call_time): Likewise.
4076 (want_inline_small_function_p): Likewise.
4077 (edge_badness): Likewise.
4078 (update_caller_keys): Likewise.
4079 (update_callee_keys): Likewise.
4080 (recursive_inlining): Likewise.
4081 (inline_small_functions): Likewise.
4082 (inline_to_all_callers_1): Likewise.
4083 (dump_overall_stats): Likewise.
4084 (early_inline_small_functions): Likewise.
4085 (early_inliner): Likewise.
4086 * ipa-inline.h (estimate_edge_growth): Likewise.
4087 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4088 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4089 * ipa-prop.h (IPA_NODE_REF): Likewise.
4090 (IPA_EDGE_REF): Likewise.
4091 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4092 (propagate_malloc): Likewise.
4093 * ipa-split.c (execute_split_functions): Likewise.
4094 * symbol-summary.h: Rename get to get_create.
4096 (get_create): Likewise.
4097 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4099 2018-06-08 Martin Liska <mliska@suse.cz>
4101 * symbol-summary.h (release): Move definition out of class
4103 (symtab_removal): Likewise.
4104 (symtab_duplication): Likewise.
4106 2018-06-08 Martin Liska <mliska@suse.cz>
4108 * symbol-summary.h (function_summary): Move constructor
4109 implementation out of class declaration.
4110 (release): Likewise.
4111 (symtab_insertion): Likewise.
4112 (symtab_removal): Likewise.
4113 (symtab_duplication): Likewise.
4116 2018-06-08 Martin Liska <mliska@suse.cz>
4118 * Makefile.in: Remove support for MPX (macros, related functions,
4119 fields in cgraph_node, ...).
4120 * builtin-types.def (BT_BND): Likewise.
4121 (BT_FN_BND_CONST_PTR): Likewise.
4122 (BT_FN_CONST_PTR_BND): Likewise.
4123 (BT_FN_VOID_PTR_BND): Likewise.
4124 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4125 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4126 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4127 (expand_builtin_mempcpy_with_bounds): Likewise.
4128 (expand_builtin_memset_with_bounds): Likewise.
4129 (expand_builtin_memset_args): Likewise.
4130 (std_expand_builtin_va_start): Likewise.
4131 (expand_builtin): Likewise.
4132 (expand_builtin_with_bounds): Likewise.
4133 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4134 (DEF_LIB_BUILTIN_CHKP): Likewise.
4135 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4136 (DEF_CHKP_BUILTIN): Likewise.
4137 (BUILT_IN_MEMCPY): Likewise.
4138 (BUILT_IN_MEMMOVE): Likewise.
4139 (BUILT_IN_MEMPCPY): Likewise.
4140 (BUILT_IN_MEMSET): Likewise.
4141 (BUILT_IN_STPCPY): Likewise.
4142 (BUILT_IN_STRCAT): Likewise.
4143 (BUILT_IN_STRCHR): Likewise.
4144 (BUILT_IN_STRCPY): Likewise.
4145 (BUILT_IN_STRLEN): Likewise.
4146 (BUILT_IN_MEMCPY_CHK): Likewise.
4147 (BUILT_IN_MEMMOVE_CHK): Likewise.
4148 (BUILT_IN_MEMPCPY_CHK): Likewise.
4149 (BUILT_IN_MEMSET_CHK): Likewise.
4150 (BUILT_IN_STPCPY_CHK): Likewise.
4151 (BUILT_IN_STRCAT_CHK): Likewise.
4152 (BUILT_IN_STRCPY_CHK): Likewise.
4153 * calls.c (store_bounds): Likewise.
4154 (emit_call_1): Likewise.
4155 (special_function_p): Likewise.
4156 (maybe_warn_nonstring_arg): Likewise.
4157 (initialize_argument_information): Likewise.
4158 (finalize_must_preallocate): Likewise.
4159 (compute_argument_addresses): Likewise.
4160 (expand_call): Likewise.
4161 * cfgexpand.c (expand_call_stmt): Likewise.
4162 (expand_return): Likewise.
4163 (expand_gimple_stmt_1): Likewise.
4164 (pass_expand::execute): Likewise.
4165 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4166 (cgraph_node::remove): Likewise.
4167 (cgraph_node::dump): Likewise.
4168 (cgraph_node::verify_node): Likewise.
4169 * cgraph.h (chkp_function_instrumented_p): Likewise.
4170 (symtab_node::get_alias_target): Likewise.
4171 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4172 (cgraph_local_p): Likewise.
4173 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4174 (cgraph_edge::rebuild_references): Likewise.
4175 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4176 (walk_polymorphic_call_targets): Likewise.
4177 (cgraph_node::expand_thunk): Likewise.
4178 (symbol_table::output_weakrefs): Likewise.
4179 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4180 (ix86_handle_option): Likewise.
4181 * config/i386/constraints.md: Likewise.
4182 * config/i386/i386-builtin-types.def (BND): Likewise.
4186 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4187 (BDESC_FIRST): Likewise.
4189 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4190 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4191 * config/i386/i386.c (enum reg_class): Likewise.
4192 (ix86_target_string): Likewise.
4193 (ix86_option_override_internal): Likewise.
4194 (ix86_conditional_register_usage): Likewise.
4195 (ix86_valid_target_attribute_inner_p): Likewise.
4196 (ix86_set_indirect_branch_type): Likewise.
4197 (ix86_set_current_function): Likewise.
4198 (ix86_function_arg_regno_p): Likewise.
4199 (init_cumulative_args): Likewise.
4200 (ix86_function_arg_advance): Likewise.
4201 (ix86_function_arg): Likewise.
4202 (ix86_pass_by_reference): Likewise.
4203 (ix86_function_value_regno_p): Likewise.
4204 (ix86_function_value_1): Likewise.
4205 (ix86_function_value_bounds): Likewise.
4206 (ix86_return_in_memory): Likewise.
4207 (ix86_setup_incoming_vararg_bounds): Likewise.
4208 (ix86_va_start): Likewise.
4209 (indirect_thunk_need_prefix): Likewise.
4210 (print_reg): Likewise.
4211 (ix86_print_operand): Likewise.
4212 (ix86_expand_call): Likewise.
4213 (ix86_output_function_return): Likewise.
4214 (reg_encoded_number): Likewise.
4215 (BDESC_VERIFYS): Likewise.
4216 (ix86_init_mpx_builtins): Likewise.
4217 (ix86_init_builtins): Likewise.
4218 (ix86_emit_cmove): Likewise.
4219 (ix86_emit_move_max): Likewise.
4220 (ix86_expand_builtin): Likewise.
4221 (ix86_builtin_mpx_function): Likewise.
4222 (ix86_get_arg_address_for_bt): Likewise.
4223 (ix86_load_bounds): Likewise.
4224 (ix86_store_bounds): Likewise.
4225 (ix86_load_returned_bounds): Likewise.
4226 (ix86_store_returned_bounds): Likewise.
4227 (ix86_class_likely_spilled_p): Likewise.
4228 (ix86_hard_regno_mode_ok): Likewise.
4229 (x86_order_regs_for_local_alloc): Likewise.
4230 (ix86_mitigate_rop): Likewise.
4231 (ix86_bnd_prefixed_insn_p): Likewise.
4232 (ix86_mpx_bound_mode): Likewise.
4233 (ix86_make_bounds_constant): Likewise.
4234 (ix86_initialize_bounds): Likewise.
4235 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4236 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4237 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4238 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4239 (TARGET_CHKP_BOUND_MODE): Likewise.
4240 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4241 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4242 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4243 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4244 * config/i386/i386.h (TARGET_MPX): Likewise.
4245 (TARGET_MPX_P): Likewise.
4246 (VALID_BND_REG_MODE): Likewise.
4247 (FIRST_BND_REG): Likewise.
4248 (LAST_BND_REG): Likewise.
4249 (enum reg_class): Likewise.
4250 (BND_REG_P): Likewise.
4251 (BND_REGNO_P): Likewise.
4252 (BNDmode): Likewise.
4253 (ADJUST_INSN_LENGTH): Likewise.
4254 * config/i386/i386.md: Likewise.
4255 * config/i386/i386.opt: Likewise.
4256 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4257 (defined): Likewise.
4258 (LINK_MPX): Likewise.
4259 (MPX_SPEC): Likewise.
4260 (LIBMPX_SPEC): Likewise.
4261 (LIBMPXWRAPPERS_SPEC): Likewise.
4262 (CHKP_SPEC): Likewise.
4263 * config/i386/predicates.md: Likewise.
4264 * dbxout.c (dbxout_type): Likewise.
4265 * doc/extend.texi: Likewise.
4266 * doc/invoke.texi: Likewise.
4267 * doc/md.texi: Likewise.
4268 * doc/tm.texi: Likewise.
4269 * doc/tm.texi.in: Likewise.
4270 * dwarf2out.c (is_base_type): Likewise.
4271 (gen_formal_types_die): Likewise.
4272 (gen_subprogram_die): Likewise.
4273 (gen_type_die_with_usage): Likewise.
4274 (gen_decl_die): Likewise.
4275 (dwarf2out_late_global_decl): Likewise.
4276 * expr.c (expand_assignment): Likewise.
4277 (emit_storent_insn): Likewise.
4278 (store_expr_with_bounds): Likewise.
4279 (store_expr): Likewise.
4280 (expand_expr_real_1): Likewise.
4281 * expr.h (store_expr_with_bounds): Likewise.
4282 * function.c (use_register_for_decl): Likewise.
4283 (struct bounds_parm_data): Likewise.
4284 (assign_parms_augmented_arg_list): Likewise.
4285 (assign_parm_find_entry_rtl): Likewise.
4286 (assign_parm_is_stack_parm): Likewise.
4287 (assign_parm_load_bounds): Likewise.
4288 (assign_bounds): Likewise.
4289 (assign_parms): Likewise.
4290 (expand_function_start): Likewise.
4291 * gcc.c (CHKP_SPEC): Likewise.
4292 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4293 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4294 (wrestrict_dom_walker::check_call): Likewise.
4295 * gimple.c (gimple_build_call_from_tree): Likewise.
4296 * gimple.h (enum gf_mask): Likewise.
4297 (gimple_call_with_bounds_p): Likewise.
4298 (gimple_call_set_with_bounds): Likewise.
4299 * gimplify.c (gimplify_init_constructor): Likewise.
4300 * ipa-cp.c (initialize_node_lattices): Likewise.
4301 (propagate_constants_across_call): Likewise.
4302 (find_more_scalar_values_for_callers_subset): Likewise.
4303 * ipa-hsa.c (process_hsa_functions): Likewise.
4304 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4305 * ipa-icf.c (sem_function::merge): Likewise.
4306 * ipa-inline.c (early_inliner): Likewise.
4307 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4308 (warn_function_cold): Likewise.
4309 (propagate_pure_const): Likewise.
4310 * ipa-ref.h (enum GTY): Likewise.
4311 * ipa-split.c (find_retbnd): Likewise.
4312 (consider_split): Likewise.
4313 (split_function): Likewise.
4314 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4315 * ipa.c (walk_polymorphic_call_targets): Likewise.
4316 (symbol_table::remove_unreachable_nodes): Likewise.
4317 (process_references): Likewise.
4318 (cgraph_build_static_cdtor_1): Likewise.
4319 * lto-cgraph.c (lto_output_node): Likewise.
4320 (output_refs): Likewise.
4321 (compute_ltrans_boundary): Likewise.
4322 (input_overwrite_node): Likewise.
4323 (input_node): Likewise.
4324 (input_cgraph_1): Likewise.
4325 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4326 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4327 (class pass_chkp_instrumentation_passes): Likewise.
4328 (make_pass_chkp_instrumentation_passes): Likewise.
4329 * passes.def: Likewise.
4330 * rtl.h (struct GTY): Likewise.
4331 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4332 * stor-layout.c (layout_type): Likewise.
4333 * symtab.c: Likewise.
4334 * target.def: Likewise.
4335 * targhooks.c (default_chkp_bound_type): Likewise.
4336 (default_chkp_bound_mode): Likewise.
4337 (default_builtin_chkp_function): Likewise.
4338 (default_chkp_function_value_bounds): Likewise.
4339 (default_chkp_make_bounds_constant): Likewise.
4340 (default_chkp_initialize_bounds): Likewise.
4341 * targhooks.h (default_chkp_bound_type): Likewise.
4342 (default_chkp_bound_mode): Likewise.
4343 (default_builtin_chkp_function): Likewise.
4344 (default_chkp_function_value_bounds): Likewise.
4345 (default_chkp_make_bounds_constant): Likewise.
4346 (default_chkp_initialize_bounds): Likewise.
4347 * toplev.c (compile_file): Likewise.
4348 (process_options): Likewise.
4349 * tree-core.h (DEF_BUILTIN): Likewise.
4350 (DEF_BUILTIN_CHKP): Likewise.
4351 * tree-inline.c (declare_return_variable): Likewise.
4352 (remap_gimple_stmt): Likewise.
4353 (copy_bb): Likewise.
4354 (initialize_inlined_parameters): Likewise.
4355 (expand_call_inline): Likewise.
4356 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4357 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4358 (make_pass_ipa_chkp_produce_thunks): Likewise.
4359 (make_pass_chkp): Likewise.
4360 (make_pass_chkp_opt): Likewise.
4361 (make_pass_chkp_instrumentation_passes): Likewise.
4362 * tree-pretty-print.c (dump_generic_node): Likewise.
4363 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4364 * tree-ssa-dce.c (propagate_necessity): Likewise.
4365 (eliminate_unnecessary_stmts): Likewise.
4366 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4367 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4368 * tree-ssa-sccvn.h: Likewise.
4369 * tree-ssa-strlen.c (get_string_length): Likewise.
4370 (valid_builtin_call): Likewise.
4371 (adjust_last_stmt): Likewise.
4372 (handle_builtin_strchr): Likewise.
4373 (handle_builtin_strcpy): Likewise.
4374 (handle_builtin_stxncpy): Likewise.
4375 (handle_builtin_memcpy): Likewise.
4376 (handle_builtin_strcat): Likewise.
4377 (strlen_check_and_optimize_stmt): Likewise.
4378 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4379 * tree-streamer-in.c: Likewise.
4380 * tree-streamer.c (record_common_node): Likewise.
4381 * tree.c (tree_code_size): Likewise.
4382 (wide_int_to_tree_1): Likewise.
4383 (type_contains_placeholder_1): Likewise.
4384 (build_common_tree_nodes): Likewise.
4385 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4386 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4387 (POINTER_BOUNDS_P): Likewise.
4388 (BOUNDED_TYPE_P): Likewise.
4389 (BOUNDED_P): Likewise.
4390 (CALL_WITH_BOUNDS_P): Likewise.
4391 (pointer_bounds_type_node): Likewise.
4392 * value-prof.c (gimple_ic): Likewise.
4393 * var-tracking.c (vt_add_function_parameters): Likewise.
4394 * varasm.c (make_decl_rtl): Likewise.
4395 (assemble_start_function): Likewise.
4396 (output_constant): Likewise.
4397 (maybe_assemble_visibility): Likewise.
4398 * varpool.c (ctor_for_folding): Likewise.
4399 * chkp-builtins.def: Remove.
4400 * ipa-chkp.c: Remove.
4401 * ipa-chkp.h: Remove.
4402 * rtl-chkp.c: Remove.
4403 * rtl-chkp.h: Remove.
4404 * tree-chkp-opt.c: Remove.
4405 * tree-chkp.c: Remove.
4406 * tree-chkp.h: Remove.
4408 2018-06-07 Carl Love <cel@us.ibm.com>
4410 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4411 vextract_fp_from_shortl): Add BE support.
4413 2018-06-07 Paul Koning <ni1d@arrl.net>
4415 * compare-elim.c (try_merge_compare): Don't merge compare if
4416 address contains a side effect.
4417 (try_eliminate_compare): Likewise.
4419 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4421 * config.gcc: Support "tremont".
4422 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4423 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4425 * config/i386/i386.c (m_TREMONT): Define.
4426 (processor_target_table): Add "tremont".
4427 (PTA_TREMONT): Define.
4428 (ix86_lea_outperforms): Add TARGET_TREMONT.
4429 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4430 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4431 and M_INTEL_GOLDMONT_PLUS.
4432 (fold_builtin_cpu): Add "tremont".
4433 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4434 (ix86_option_override_internal): Add "tremont".
4435 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4436 (processor_type): Add PROCESSOR_TREMONT.
4437 * config/i386/x86-tune.def: Add m_TREMONT.
4438 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4440 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4442 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4443 symbol defined for msp430i* devices to be lower case.
4445 gcc/testsuite/gcc.target/msp430/
4446 * msp430i-device-symbol.c: New test.
4447 * msp430f-device-symbol.c: New test.
4448 * msp430.h: New test header file.
4450 2018-06-07 Richard Biener <rguenther@suse.de>
4452 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4453 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4454 Properly wrap signed arithmetic if overflow wraps.
4456 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4458 PR tree-optimization/69615
4459 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4460 of a cast from a same precision integral SSA_NAME in a bb dominated
4461 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4462 cast to utype if rhs2 has already a compatible type.
4464 2018-06-07 Richard Biener <rguenther@suse.de>
4466 PR tree-optimization/85935
4467 * graphite-scop-detection.c (find_params_in_bb): Analyze
4468 condition operands with respect to the correct loop. Assert
4469 the analysis doesn't fail.
4471 2018-06-04 Carl Love <cel@us.ibm.com>
4473 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4474 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4475 as it is slightly cheaper.
4476 (first_match_or_eos_index_<mode>):
4477 Calculate index using natural element order.
4478 (first_match_index_<mode>):
4479 Calculate index using natural element order.
4480 (first_match_or_eos_index_<mode>):
4481 Calculate index using natural order.
4482 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4483 for BE and LE modes.
4484 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4485 P9V_BUILTIN_VCLZLSBB_V16QI.
4486 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4489 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4491 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4492 indentation and line wrap for many prototypes. Add missing
4493 @smallexample directives around block of prototypes for vec_xl and
4496 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4498 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4499 track if we pass or return IEEE 128-bit floating point.
4500 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4501 C++ mangling that is compatible with GCC 8.1.
4502 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4503 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4504 floating point types.
4505 (rs6000_function_arg_advance_1): Likewise.
4506 (rs6000_mangle_type): Optionally generate mangled names that match
4507 what GCC 8.1 generated for IEEE 128-bit floating point types.
4508 (rs6000_globalize_decl_name): If we have an external function that
4509 passes or returns IEEE 128-bit types, generate a weak reference
4510 from the mangled name used in GCC 8.1 to the current mangled
4512 (rs6000_init_builtins): Make __ibm128 use the long double type if
4513 long double is IBM extended double. Make __float128 use the long
4514 double type if long double is IEEE 128-bit.
4517 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4518 macro for __ibm128 built-in functions.
4519 (PACK_IF): Add __ibm128 pack/unpack functions.
4520 (UNPACK_IF): Likewise.
4521 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4522 enable long double built-in functions if long double is IEEE
4523 128-bit floating point.
4524 (rs6000_invalid_builtin): Update long double built-in function
4526 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4527 functions, adjust the built-in function to use the long double
4528 built-in function if __ibm128 and long double are the same type.
4529 * doc/extend.texi (PowerPC builtins): Update documention for
4530 __builtin_{,un}pack_longdouble. Add documentation for
4531 __builtin_{,un}pack_ibm128.
4533 2018-06-06 Jim Wilson <jimw@sifive.com>
4535 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4536 (struct machine_function): New field interrupt_mode.
4537 (riscv_handle_type_attribute): New function. Add forward declaration.
4538 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4539 (riscv_expand_epilogue): Check interrupt_mode field.
4540 (riscv_set_current_function): Check interrupt attribute args and
4541 set interrupt_mode field.
4542 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4543 (riscv_sret, riscv_uret): New.
4544 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4545 new arguments to interrupt attribute.
4547 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4550 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4551 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4553 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4555 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4556 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4557 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4558 several redundant entries.
4560 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4562 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4563 type from "rtx" to "rtx_insn *".
4564 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4565 for local "call_insn", removing cast.
4566 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4568 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4570 PR tree-optimization/86066
4571 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4572 for BIT_INSERT_EXPR stores.
4574 2018-06-06 Richard Biener <rguenther@suse.de>
4576 PR tree-optimization/86062
4577 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4578 component refs ontop
4579 of to be offsetted base.
4581 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4583 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4584 to be static and remove check on interrupt attribute name.
4586 gcc/testsuite/gcc.target/msp430/
4587 * function-attributes-4.c: New test.
4588 * static-interrupts.c: New test.
4590 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4592 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4593 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4595 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4598 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4600 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4601 Remove second argument, change how error is called.
4602 (aarch64_layout_arg): Remove second argument from
4603 aarch64_err_no_fpadvsimd call.
4604 (aarch64_init_cumulative_args): Ditto.
4605 (aarch64_gimplify_va_arg_expr): Ditto.
4606 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4608 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4610 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4611 (*simple_return_indirect_internal<mode>): Rename from
4612 simple_return_indirect_internal. Use W mode iterator.
4613 (rstorssp): New expander.
4614 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4615 (clrssbsy): New expander.
4616 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4618 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4620 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4622 (cmse_check_pointed_object): Likewise.
4624 2018-06-05 Martin Liska <mliska@suse.cz>
4626 PR gcov-profile/47618
4627 * doc/invoke.texi: Document how -fprofile-dir format
4630 2018-06-05 Richard Biener <rguenther@suse.de>
4632 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4633 removal pretend DOM info isn't available so we do not update
4634 it and only remove edges, not dominated blocks. Actually free
4635 DOM info in case we removed something. Remove unreachable blocks.
4636 (mfb_keep_latches): Work with either DOM info or marked backedges.
4637 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4638 first. Mark backedges if DOM info isn't available.
4639 (Re-)compute DOM info after cleanup_control_flow_pre.
4641 2018-06-05 Richard Biener <rguenther@suse.de>
4643 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4644 (locus_discrim_hasher::hash): Adjust.
4645 (locus_discrim_hasher::equal): Likewise.
4646 (next_discriminator_for_locus): Work on line directly.
4647 (same_line_p): Pass in expanded locus1 as well.
4648 (assign_discriminators): Avoid redundant location expansions.
4650 2018-06-05 Richard Biener <rguenther@suse.de>
4652 PR tree-optimization/86046
4653 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4654 if required after clearing TREE_ADDRESSABLE.
4656 2018-06-05 Richard Biener <rguenther@suse.de>
4658 PR tree-optimization/86047
4659 * tree-ssa-loop.c (for_each_index): Glob handling of all
4660 decls and constants and really handle all of them.
4662 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4665 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4666 qualifier_void_pointer and qualifier_const_void_pointer.
4667 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4668 (arm_init_builtins): Handle the above.
4669 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4670 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4671 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4674 2018-06-05 Martin Liska <mliska@suse.cz>
4676 * auto-profile.c (read_autofdo_file): Do not use
4677 gcov_ctr_summary struct.
4678 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4679 * coverage.c (struct counts_entry): Likewise.
4680 (read_counts_file): Read just single summary entry.
4681 (get_coverage_counts): Use gcov_summary struct.
4682 * coverage.h (get_coverage_counts): Likewise.
4683 * gcov-dump.c (dump_working_sets): Likewise.
4684 (tag_summary): Dump just single summary.
4685 * gcov-io.c (gcov_write_summary): Write just histogram
4687 (gcov_read_summary): Read just single summary.
4688 (compute_working_sets): Use gcov_summary struct.
4689 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4690 of GCOV_COUNTERS_SUMMABLE.
4691 (GCOV_COUNTERS_SUMMABLE): Remove.
4692 (GCOV_FIRST_VALUE_COUNTER): Replace with
4693 GCOV_COUNTER_V_INTERVAL.
4694 (struct gcov_ctr_summary): Remove.
4695 (struct gcov_summary): Directly use fields of former
4697 (compute_working_sets): Use gcov_summary struct.
4698 * gcov.c (read_count_file): Do not use ctrs fields.
4699 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4701 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4703 * profile.c: Likewise.
4704 * profile.h: Likewise.
4706 2018-06-05 Martin Liska <mliska@suse.cz>
4708 PR gcov-profile/84846
4709 * gcov.c (output_lines): Print working directory only
4710 in intermediate format.
4712 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4714 * config/s390/s390-builtin-types.def: Add void function type.
4715 * config/s390/s390-builtins.def: Use the function type for the
4718 2018-06-04 Jim Wilson <jimw@sifive.com>
4720 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4722 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4723 and maybe_eh_return. Change regno to unsigned int. Use new args to
4724 handle EH_RETURN_DATA_REGNO registers properly.
4725 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4726 (riscv_expand_epilogue): Update comment. Change argument name and
4727 type. Update code to use new name and type. Pass new args to
4728 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4730 * config/riscv/riscv.md (NORMAL_RETURN): New.
4731 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4732 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4733 (eh_return): Call gen_eh_return_internal and emit barrier.
4734 (eh_return_internal): Call riscv_expand_epilogue.
4736 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4738 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4739 bit_insertion field and declare can_be_merged_into method.
4740 (merged_store_group::can_be_merged_into): New method.
4741 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4742 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4743 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4745 2018-06-04 Richard Biener <rguenther@suse.de>
4747 PR tree-optimization/85955
4748 * builtins.c (fold_builtin_sincos): Convert pointers to
4749 destination to appropriate type before dereferencing.
4751 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4753 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4755 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4757 * expr.c (expand_expr_real_1): Force the operand into memory if
4758 its TYPE_MODE is BLKmode and if there is no integer mode for
4759 the number of bits being extracted.
4761 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4765 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4766 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4768 2018-06-04 Richard Biener <rguenther@suse.de>
4770 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4771 (cleanup_tree_cfg_noloop): ... single caller. Do
4772 start_recording_case_labels later.
4774 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4776 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4777 to _IMMINTRIN_H_INCLUDED.
4778 * config/i386/pconfigintrin.h: Ditto.
4779 * config/i386/waitpkgintrin.h: Ditto.
4780 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4781 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4782 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4783 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4784 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4785 waitpkgintrin.h and cldemoteintrin.h.
4787 2018-06-04 Richard Biener <rguenther@suse.de>
4789 PR tree-optimization/86038
4790 * tracer.c (find_best_successor): Check probability for
4791 being initialized, bail out if not.
4793 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4796 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4797 of bits to ignore when comparing architectures.
4799 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4801 PR tree-optimization/69615
4802 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4803 maximum or minimum of the type, try to merge it also as if
4804 range1 is + [-, x - 1] or + [x + 1, -].
4807 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4809 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4811 PR tree-optimization/86034
4812 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4813 the unsigned bitfield type in a bit insertion sequence if it does not
4814 have a larger precision than the bitfield size.
4815 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4817 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4819 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4821 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4823 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4824 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4825 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4826 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4828 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4830 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4831 Disable -fdelete-null-pointer-checks for ELF toolchain.
4833 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4834 Kito Cheng <kito.cheng@gmail.com>
4836 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4837 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4838 (nds32*-*-*): Add glibc and uclibc conditions.
4839 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4840 (TARGET_EXCEPT_UNWIND_INFO): Define.
4841 * config/nds32/elf.h: New file.
4842 * config/nds32/linux.h: New file.
4843 * config/nds32/nds32-elf.opt: New file.
4844 * config/nds32/nds32-linux.opt: New file.
4845 * config/nds32/nds32-fp-as-gp.c
4846 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4847 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4849 (nds32_asm_file_end): Ditto.
4850 (nds32_print_operand): Ditto.
4851 (nds32_insert_attributes): Ditto.
4852 (nds32_init_libfuncs): New function.
4853 (TARGET_HAVE_TLS): Define.
4854 (TARGET_INIT_LIBFUNCS): Define.
4855 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4857 (TARGET_ELF): Apply different mcmodel setting.
4858 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4859 been migrated into elf.h and linux.h files.
4860 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4861 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4862 (mcmodel): The content has been migrated into nds32-elf.opt and
4863 nds32-linux.opt files.
4864 * config/nds32/t-elf: New file.
4865 * config/nds32/t-linux: New file.
4867 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4868 Shiva Chen <shiva0217@gmail.com>
4870 * config/nds32/constants.md (unspec_volatile_element): Add
4871 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4872 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4874 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4875 (make_pass_nds32_fp_as_gp): Declare.
4876 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4878 (nds32_asm_function_end_prologue): Remove unused asm output.
4879 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4880 (nds32_asm_file_start): Output necessary fp_as_gp information.
4881 (nds32_option_override): Adjust register usage.
4882 (nds32_expand_prologue): Consider fp_as_gp situation.
4883 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4884 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4887 (simple_return): Ditto.
4888 (omit_fp_begin): Output special directive for fp_as_gp.
4889 (omit_fp_end): Output special directive for fp_as_gp.
4890 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4891 mforbid-fp-as-gp): New options.
4893 2018-06-01 Mark Wielaard <mark@klomp.org>
4895 * dwarf2out.c (dwarf2out_finish): Remove generation of
4896 DW_AT_loclists_base.
4898 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4900 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4901 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4902 (struct merged_store_group): Add bit_insertion field.
4903 (dump_char_array): Use standard hexadecimal format.
4904 (merged_store_group::merged_store_group): Set bit_insertion to false.
4905 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4906 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4907 also print the mask in the dump file.
4908 (pass_store_merging::gate): Minor tweak.
4909 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4910 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4911 stores with INTEGER_CST stores.
4912 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4913 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4914 and use it throughout. Generate bit insertion sequences if need be.
4915 (pass_store_merging::process_store): Remove redundant condition.
4916 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4918 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4920 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4921 the 128-bit floating point types. Fix function comment.
4923 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4925 * config/aarch64/aarch64-simd.md
4926 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4928 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4931 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4933 PR tree-optimization/85989
4934 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4936 (backprop::intersect_uses): Check it when deciding whether this
4937 is a backedge reference.
4938 (backprop::process_block): Add each phi to m_visited_phis
4939 after visiting it, then clear it at the end.
4941 2018-06-01 Richard Biener <rguenther@suse.de>
4943 * tree-vectorizer.h (vect_dr_stmt): New function.
4944 (vect_get_load_cost): Adjust.
4945 (vect_get_store_cost): Likewise.
4946 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4947 Use vect_dr_stmt instead of DR_SMTT.
4948 (vect_record_base_alignments): Likewise.
4949 (vect_calculate_target_alignment): Likewise.
4950 (vect_compute_data_ref_alignment): Likewise and make static.
4951 (vect_update_misalignment_for_peel): Likewise.
4952 (vect_verify_datarefs_alignment): Likewise.
4953 (vector_alignment_reachable_p): Likewise.
4954 (vect_get_data_access_cost): Likewise. Pass down
4955 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4956 (vect_get_peeling_costs_all_drs): Likewise.
4957 (vect_peeling_hash_get_lowest_cost): Likewise.
4958 (vect_enhance_data_refs_alignment): Likewise.
4959 (vect_find_same_alignment_drs): Likewise.
4960 (vect_analyze_data_refs_alignment): Likewise.
4961 (vect_analyze_group_access_1): Likewise.
4962 (vect_analyze_group_access): Likewise.
4963 (vect_analyze_data_ref_access): Likewise.
4964 (vect_analyze_data_ref_accesses): Likewise.
4965 (vect_vfa_segment_size): Likewise.
4966 (vect_small_gap_p): Likewise.
4967 (vectorizable_with_step_bound_p): Likewise.
4968 (vect_prune_runtime_alias_test_list): Likewise.
4969 (vect_analyze_data_refs): Likewise.
4970 (vect_supportable_dr_alignment): Likewise.
4971 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4972 (vect_gen_prolog_loop_niters): Likewise.
4973 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4974 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4976 (vect_recog_mask_conversion_pattern): Likewise.
4977 (vect_try_gather_scatter_pattern): Likewise.
4978 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4979 to vect_get_store_cost.
4980 (vect_get_store_cost): Get stmt_info instead of DR.
4981 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4982 (vect_get_load_cost): Get stmt_info instead of DR.
4984 2018-06-01 Richard Biener <rguenther@suse.de>
4987 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4988 vars wrapped in MEM_REFs.
4990 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4992 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4993 Fix subreg tests so that we only return a choice between
4994 GENERAL_REGS and FP_REGS if the original classes included both.
4996 2018-06-01 Richard Biener <rguenther@suse.de>
4999 * tree-ssa-structalias.c (get_function_part_constraint):
5000 Handle NULL fi->decl.
5001 (find_func_aliases_for_call): Properly handle indirect
5002 fi from direct call.
5003 (find_func_clobbers): Likewise.
5004 (ipa_pta_execute): Likewise.
5005 (create_variable_info_for): For functions that are ifunc_resolver
5006 resolve to a varinfo that contains the result of the resolver call.
5007 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5010 2018-05-31 Michael Collison <michael.collison@arm.com>
5012 * config/aarch64/aarch64.md:
5013 (*fix_to_zero_extenddfdi2): New pattern.
5014 * gcc.target/aarch64/fix_extend1.c: New testcase.
5016 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5020 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5021 and BUILT_IN_STRNCMP_EQ.
5022 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5023 BUILT_IN_STRNCMP_EQ.
5024 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5025 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5026 (gimple_fold_builtin): Likewise.
5027 * tree-ssa-strlen.c (compute_string_length): New function.
5028 (determine_min_obsize): New function.
5029 (handle_builtin_string_cmp): New function to handle calls to
5030 string compare functions.
5031 (strlen_optimize_stmt): Add handling to builtin string compare
5033 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5034 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5035 * tree.c (build_common_builtin_nodes): Add new defines of
5036 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5038 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5041 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5042 functions with naked attribute.
5044 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5046 * config/i386/sse.md (avx_vec_concat<mode>):
5047 Substitute concat_tg_mode mode attribute with xtg_mode.
5048 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5049 (concat_tg_mode): Remove mode attribute.
5051 2018-05-31 Martin Sebor <msebor@redhat.com>
5054 * calls.c (alloc_max_size): Correct a logic error/typo.
5055 Treat excessive arguments as infinite. Warn for invalid arguments.
5056 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5058 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5061 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5062 and movx for Haswell.
5064 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5065 Cesar Philippidis <cesar@codesourcery.com>
5068 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5069 when emitting error on private/firstprivate reductions.
5070 * omp-low.c (lower_omp_target): Avoid reference-type processing
5071 on pointers for firstprivate clause.
5073 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5075 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5078 * config/aarch64/aarch64-simd.md
5079 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5080 (aarch64_ld1_x3_<mode>): Likewise
5081 (aarch64_st1x2<VALLDIF:mode>): Likewise
5082 (aarch64_st1_x2_<mode>): Likewise
5083 (aarch64_st1x3<VALLDIF:mode>): Likewise
5084 (aarch64_st1_x3_<mode>): Likewise
5085 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5086 (vld1_s8_x3): Likewise.
5087 (vld1_u16_x3): Likewise.
5088 (vld1_s16_x3): Likewise.
5089 (vld1_u32_x3): Likewise.
5090 (vld1_s32_x3): Likewise.
5091 (vld1_u64_x3): Likewise.
5092 (vld1_s64_x3): Likewise.
5093 (vld1_f16_x3): Likewise.
5094 (vld1_f32_x3): Likewise.
5095 (vld1_f64_x3): Likewise.
5096 (vld1_p8_x3): Likewise.
5097 (vld1_p16_x3): Likewise.
5098 (vld1_p64_x3): Likewise.
5099 (vld1q_u8_x3): Likewise.
5100 (vld1q_s8_x3): Likewise.
5101 (vld1q_u16_x3): Likewise.
5102 (vld1q_s16_x3): Likewise.
5103 (vld1q_u32_x3): Likewise.
5104 (vld1q_s32_x3): Likewise.
5105 (vld1q_u64_x3): Likewise.
5106 (vld1q_s64_x3): Likewise.
5107 (vld1q_f16_x3): Likewise.
5108 (vld1q_f32_x3): Likewise.
5109 (vld1q_f64_x3): Likewise.
5110 (vld1q_p8_x3): Likewise.
5111 (vld1q_p16_x3): Likewise.
5112 (vld1q_p64_x3): Likewise.
5113 (vst1_s64_x2): Likewise.
5114 (vst1_u64_x2): Likewise.
5115 (vst1_f64_x2): Likewise.
5116 (vst1_s8_x2): Likewise.
5117 (vst1_p8_x2): Likewise.
5118 (vst1_s16_x2): Likewise.
5119 (vst1_p16_x2): Likewise.
5120 (vst1_s32_x2): Likewise.
5121 (vst1_u8_x2): Likewise.
5122 (vst1_u16_x2): Likewise.
5123 (vst1_u32_x2): Likewise.
5124 (vst1_f16_x2): Likewise.
5125 (vst1_f32_x2): Likewise.
5126 (vst1_p64_x2): Likewise.
5127 (vst1q_s8_x2): Likewise.
5128 (vst1q_p8_x2): Likewise.
5129 (vst1q_s16_x2): Likewise.
5130 (vst1q_p16_x2): Likewise.
5131 (vst1q_s32_x2): Likewise.
5132 (vst1q_s64_x2): Likewise.
5133 (vst1q_u8_x2): Likewise.
5134 (vst1q_u16_x2): Likewise.
5135 (vst1q_u32_x2): Likewise.
5136 (vst1q_u64_x2): Likewise.
5137 (vst1q_f16_x2): Likewise.
5138 (vst1q_f32_x2): Likewise.
5139 (vst1q_f64_x2): Likewise.
5140 (vst1q_p64_x2): Likewise.
5141 (vst1_s64_x3): Likewise.
5142 (vst1_u64_x3): Likewise.
5143 (vst1_f64_x3): Likewise.
5144 (vst1_s8_x3): Likewise.
5145 (vst1_p8_x3): Likewise.
5146 (vst1_s16_x3): Likewise.
5147 (vst1_p16_x3): Likewise.
5148 (vst1_s32_x3): Likewise.
5149 (vst1_u8_x3): Likewise.
5150 (vst1_u16_x3): Likewise.
5151 (vst1_u32_x3): Likewise.
5152 (vst1_f16_x3): Likewise.
5153 (vst1_f32_x3): Likewise.
5154 (vst1_p64_x3): Likewise.
5155 (vst1q_s8_x3): Likewise.
5156 (vst1q_p8_x3): Likewise.
5157 (vst1q_s16_x3): Likewise.
5158 (vst1q_p16_x3): Likewise.
5159 (vst1q_s32_x3): Likewise.
5160 (vst1q_s64_x3): Likewise.
5161 (vst1q_u8_x3): Likewise.
5162 (vst1q_u16_x3): Likewise.
5163 (vst1q_u32_x3): Likewise.
5164 (vst1q_u64_x3): Likewise.
5165 (vst1q_f16_x3): Likewise.
5166 (vst1q_f32_x3): Likewise.
5167 (vst1q_f64_x3): Likewise.
5168 (vst1q_p64_x3): Likewise.
5170 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5172 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5173 user_label_prefix to name.
5175 * tree-core.h: Update comment about the format of NAME string
5176 passed to handler in attribute_spec.
5178 * config/msp430/msp430.md: Remove erroneous subreg expression from
5179 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5180 zero_extend{q,h}isi2.
5182 2018-05-30 Borislav Petkov <bp@suse.de>
5184 * doc/extend.texi: Document some architecture specific
5185 constraints and sort entries.
5187 2018-05-30 Martin Sebor <msebor@redhat.com>
5190 * builtins.c (expand_builtin_stpcpy_1): New function.
5191 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5192 only if the former succeeds.
5194 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5196 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5199 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5201 * doc/invoke.texi (-flinker-output): Document
5203 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5205 * passes.c (ipa_write_summaries): Only modify statements if body
5207 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5208 incrementally linking.
5209 (ipa_passes): Likewise.
5210 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5211 pass down resolution info.
5212 * common.opt (flag_incremental_link): Update info.
5213 * gcc.c (plugin specs): Turn flinker-output=* to
5214 -plugin-opt=-linker-output-known
5215 * toplev.c (compile_file): Also cut compilation when doing incremental
5217 * flag-types. (enum lto_partition_model): Add
5218 LTO_LINKER_OUTPUT_NOLTOREL.
5219 (invoke.texi): Add -flinker-output docs.
5220 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5221 link same way as WPA; do not stream in dead initializers.
5223 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5224 darf2out_register_external_decl): Support incremental link.
5226 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5228 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5230 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5232 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5233 it down to simple_object_copy_lto_debug_sections.
5234 (run_gcc): Determine incremental LTO link time and configure
5235 lto1 into non-wpa mode, disable renaming of debug sections.
5237 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5239 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5240 descriptions of various incorrectly documented functions.
5242 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5245 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5246 address check not strict.
5248 2018-05-30 Richard Biener <rguenther@suse.de>
5250 PR tree-optimization/85964
5251 * tracer.c (better_p): Drop initialized count check, we only
5252 call the function with initialized counts now.
5253 (find_best_successor): Do find a best edge if one
5254 has uninitialized count.
5255 (find_best_predecessor): Likewise. Do BB frequency check only
5256 if count is initialized.
5258 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5260 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5261 (aarch64_ldrstr_offset_compare): New.
5262 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5263 load/store orderings.
5264 (aarch64_gen_adjusted_ldpstp): Likewise.
5266 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5268 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5269 Check for subset of GENERAL_REGS and FP_REGS.
5270 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5273 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5275 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5276 and wi::to_poly_offset. Add the current offset and then check
5277 whether the sum fits, rather than using an unchecked addition of
5278 a checked term. Check for a shwi rather than a uhwi.
5279 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5280 (store_constructor): Use poly_int_tree_p.
5281 (expand_expr_real_1): Likewise.
5282 * function.c (assign_temp): Likewise.
5283 * fold-const.c (const_binop): Use poly_int_tree_p and
5285 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5287 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5288 to_poly_offset for MEM offsets.
5289 * ipa-icf.c (sem_variable::equals): Likewise.
5290 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5291 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5292 wi::to_poly_offset for BIT_FIELD_REF offsets.
5293 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5295 * var-tracking.c (emit_note_insn_var_location): Use
5296 tree_to_poly_uint64.
5298 2018-05-29 Jim Wilson <jimw@sifive.com>
5300 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5302 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5305 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5306 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5308 (sse4_1_round<mode>2): Use nonimmediate_operand
5309 for operand 1 predicate.
5311 2018-05-29 Martin Sebor <msebor@redhat.com>
5312 Richard Biener <rguenther@suse.de>
5315 * calls.c (get_size_range): Call determine_value_range instead
5316 of get_value_range..
5317 * tree-vrp.h (determine_value_range): Declared new function.
5318 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5320 2018-05-29 Richard Biener <rguenther@suse.de>
5322 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5323 sure to use non-pattern stmts for get_earlier_stmt arguments.
5324 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5325 called on pattern stmts.
5326 (get_later_stmt): Likewise.
5328 2018-05-29 Martin Liska <mliska@suse.cz>
5330 PR gcov-profile/85759
5331 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5334 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5336 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5338 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5342 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5343 VEC_PACK_FLOAT_EXPR): New tree codes.
5344 * tree-pretty-print.c (op_code_prio): Handle
5345 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5346 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5347 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5348 * tree-inline.c (estimate_operator_cost): Likewise.
5349 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5350 * fold-const.c (const_binop): Likewise.
5351 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5352 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5353 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5354 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5355 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5356 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5357 * expr.c (expand_expr_real_2): Likewise.
5358 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5359 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5360 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5362 * optabs.c (expand_widen_pattern_expr): For
5363 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5364 sign from result type rather than operand's type.
5365 (expand_binop_directly): For vec_packu_float_optab and
5366 vec_packs_float_optab allow result type to be different from operand's
5368 * optabs-tree.c (optab_for_tree_code): Handle
5369 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5370 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5371 * tree-vect-generic.c (expand_vector_operations_1): Handle
5372 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5373 VEC_PACK_FLOAT_EXPR.
5374 * tree-vect-stmts.c (supportable_widening_operation): Handle
5376 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5377 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5378 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5379 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5380 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5382 (vec_pack<floatprefix>_float_<mode>): New expander.
5383 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5385 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5386 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5387 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5388 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5389 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5391 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5392 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5393 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5394 VEC_PACK_FLOAT_EXPR): Document.
5396 2018-05-29 Richard Biener <rguenther@suse.de>
5398 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5400 (stmt_vec_info_vec): Make pointer.
5401 (init_stmt_vec_info_vec): Remove.
5402 (free_stmt_vec_info_vec): Likewise.
5403 (set_stmt_vec_info_vec): New function.
5404 (free_stmt_vec_infos): Likewise.
5405 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5406 (set_vinfo_for_stmt): Likewise.
5407 (get_earlier_stmt): Likewise.
5408 (get_later_stmt): Likewise.
5409 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5410 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5411 (vec_info::~vec_info): Free stmt_vec_infos.
5412 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5413 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5414 (pass_slp_vectorize::execute): Likewise.
5415 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5416 (free_stmt_vec_info_vec): Likewise.
5417 (set_stmt_vec_info_vec): New function.
5418 (free_stmt_vec_infos): Likewise.
5419 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5420 the global stmt_vec_info_vec.
5421 * tree-parloops.c (gather_scalar_reductions): Use
5422 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5425 2018-05-29 Richard Biener <rguenther@suse.de>
5427 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5429 2018-05-29 Martin Liska <mliska@suse.cz>
5430 David Malcolm <dmalcolm@redhat.com>
5432 * vec.c (test_reverse): New.
5433 (vec_c_tests): Add new test.
5434 * vec.h (vl_ptr>::reverse): New function.
5436 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5438 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5440 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5443 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5445 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5447 2018-05-28 Richard Biener <rguenther@suse.de>
5449 PR tree-optimization/85933
5450 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5451 look at stmts marked as vectorizable.
5453 2018-05-28 Richard Biener <rguenther@suse.de>
5455 PR tree-optimization/85934
5456 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5457 vector boolean check before scalar optimization.
5459 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5461 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5464 2018-05-28 Mark Wielaard <mark@klomp.org>
5466 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5467 if it is an expression containing a minus sign.
5469 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5471 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5473 2018-05-27 Paul Koning <ni1d@arrl.net>
5475 * config/pdp11/pdp11.md (truncsihi2): Remove.
5477 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5478 Chung-Ju Wu <jasonwucj@gmail.com>
5480 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5482 (unaligned_store_dw): Ditto.
5483 * config/nds32/nds32-memory-manipulation.c
5484 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5485 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5486 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5487 (emit_setmem_word_loop): Rename to ...
5488 (emit_setmem_doubleword_loop): ... this.
5489 (nds32_gen_dup_4_byte_to_word_value): New function.
5490 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5491 (nds32_expand_setmem_loop): Refine implementation.
5492 (nds32_expand_setmem_loop_v3m): Ditto.
5493 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5496 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5498 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5500 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5502 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5503 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5504 machine->attr_no_prologue_p.
5505 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5506 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5507 (nds32_expand_epilogue): Consider attr_naked_p.
5508 (nds32_expand_epilogue_v3pop): Likewise.
5509 (nds32_can_use_return_insn): Likewise.
5510 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5511 attr_no_prologue_p fields.
5512 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5514 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5517 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5519 * config/i386/sse.md
5520 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5522 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5524 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5526 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5528 (*<floatsuffix>floatv2div2sf2): Rename to ...
5529 (*float<floatunssuffix>v2div2sf2): ... this.
5530 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5531 (float<floatunssuffix>v2div2sf2_mask): ... this.
5532 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5533 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5534 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5536 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5538 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5540 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5542 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5544 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5546 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5547 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5548 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5549 gen_ufix_truncv8dfv8si2.
5550 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5551 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5552 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5553 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5554 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5555 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5556 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5557 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5559 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5563 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5565 2018-05-25 Jim Wilson <jimw@sifive.com>
5567 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5568 * config/riscv/riscv.c (struct machine_function): Add
5569 interrupt_handler_p and attribute_checked_p fields.
5570 (riscv_attribute_table): Add interrupt.
5571 (riscv_interrupt_type_p): New.
5572 (riscv_save_reg_p): Save extra regs for interrupt handler.
5573 (riscv_use_save_libcall): Return false for interrupt handler.
5574 (riscv_first_stack_step): Add forward declaration.
5575 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5576 for interrupt handler with large frame. Use it for saved reg list.
5577 (riscv_expand_prologue): Move flag_stack_usage_info support to
5578 eliminate duplication.
5579 (riscv_expand_epilogue): Generate mret for interrupt handler.
5580 (riscv_epilogue_uses): New.
5581 (riscv_can_use_return_insn): Return false for interrupt handler.
5582 (riscv_function_ok_for_sibcall): Likewise.
5583 (riscv_set_current_function): Add interrupt handler support.
5584 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5585 * config/riscv/riscv.md (UNSPECV_MRET): New.
5587 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5589 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5591 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5593 PR tree-optimization/85712
5594 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5595 this candidate has already been replaced in-situ by a copy.
5597 2018-05-25 Jason Merrill <jason@redhat.com>
5599 PR c++/80485 - inline function non-zero address.
5600 * symtab.c (nonzero_address): Check DECL_COMDAT.
5602 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5605 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5606 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5607 extension of SImode operation. Use const123_operand predicate.
5608 (*saddsi_1): Remove.
5609 (*saddl_se_1): Ditto.
5611 (*ssubl_se_1): Ditto.
5612 * config/alpha/predicates.md (const123_operand): New predicate.
5613 * config/alpha/constraints.md (P): Use IN_RANGE.
5615 2018-05-25 Richard Biener <rguenther@suse.de>
5617 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5619 (ref_maybe_used_by_stmt_p): Likewise.
5620 (stmt_may_clobber_ref_p): Likewise.
5621 (stmt_may_clobber_ref_p_1): Likewise.
5622 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5624 (ref_maybe_used_by_stmt_p): Likewise.
5625 (stmt_may_clobber_ref_p): Likewise.
5626 (stmt_may_clobber_ref_p_1): Likewise.
5627 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5628 the alias oracle to disambiguate DRs with stmts DR analysis
5630 (vect_analyze_data_refs): Do not give up on not analyzable
5631 DRs for BB vectorization. Remove code truncating the dataref
5634 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5637 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5638 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5639 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5641 2018-05-25 Richard Biener <rguenther@suse.de>
5643 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5644 function, combining stmt data ref gathering and fatal analysis
5646 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5647 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5648 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5649 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5650 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5652 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5654 PR tree-optimization/85720
5655 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5656 SCC if all partitions are builtins.
5657 (version_loop_by_alias_check): New parameter. Generate cancelable
5658 runtime alias check if all partitions are builtins.
5659 (distribute_loop): Update call to above function.
5661 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5663 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5664 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5665 (parm_default_def_partition_arg): Ditto.
5666 (set_parm_default_def_partition): Ditto.
5667 (get_parm_default_def_partitions): Ditto and make it static.
5668 (get_undefined_value_partitions): Ditto and make it static.
5669 (remove_ssa_form): Refactor call to init_var_map here.
5670 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5671 computation for loop region.
5672 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5673 (register_default_def): Delete.
5674 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5675 (parm_default_def_partition_arg): Ditto.
5676 (set_parm_default_def_partition): Ditto.
5677 (get_parm_default_def_partitions): Ditto and make it static.
5678 (get_undefined_value_partitions): Ditto and make it static.
5679 (coalesce_with_default, coalesce_with_default): Update comment.
5680 (create_coalesce_list_for_region): New func factored out from
5681 create_outofssa_var_map.
5682 (populate_coalesce_list_for_outofssa): New func factored out from
5683 create_outofssa_var_map and coalesce_ssa_name.
5684 (create_outofssa_var_map): Delete.
5685 (coalesce_ssa_name): Refactor to support live range computation.
5686 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5687 (get_parm_default_def_partitions): Delete.
5688 (get_undefined_value_partitions): Ditto.
5689 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5690 computation for loop region.
5691 (new_tree_live_info, loe_visit_block): Ditto.
5692 (live_worklist, set_var_live_on_entry): Ditto.
5693 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5694 * tree-ssa-live.h (struct _var_map): New fields.
5695 (init_var_map): Change decl.
5696 (region_contains_p): New.
5698 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5700 * tree-ssa-live.h (live_merge_and_clear): Delete.
5702 2018-05-25 Richard Biener <rguenther@suse.de>
5705 * tree-dump.c (dequeue_and_dump): Remove access to removed
5706 operand 2 of a SWITCH_EXPR.
5708 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5710 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5711 multiplication and division.
5712 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5713 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5714 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5715 (cond_udiv_optab, cond_umod_optab): New optabs.
5716 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5717 (IFN_COND_RDIV): New internal functions.
5718 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5719 TRUNC_MOD_EXPR and RDIV_EXPR.
5720 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5721 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5723 (SVE_INT_BINARY): Include mult.
5724 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5725 (optab, sve_int_op): Handle mult.
5726 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5728 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5729 for SVE_INT_BINARY_SD.
5731 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5733 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5734 (optab, sve_int_op): Handle div and udiv.
5735 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5736 for SVE_INT_BINARY_SD.
5737 (*<optab><mode>3): New insn for the same.
5739 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5741 * tree-vect-patterns.c: Include predict.h.
5742 (vect_recog_divmod_pattern): Restrict check for division support
5743 to when optimizing for size.
5745 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5747 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5748 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5749 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5750 (gimple_match_op::set_op): Likewise.
5751 (gimple_resimplify4): Declare.
5752 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5753 (expr::gen_transform): Likewise.
5754 (decision_tree::gen): Generate a simplification routine for 4 operands.
5755 * gimple-match-head.c (gimple_simplify): Add an overload for
5756 4 operands. In the top-level function, handle up to 4 call
5757 arguments and call gimple_resimplify4.
5758 (gimple_resimplify4): New function.
5759 (build_call_internal): Pass a fourth operand.
5760 (maybe_push_to_seq): Likewise.
5761 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5762 Fold VEC_COND_EXPRs of an operation and a default value into
5763 an IFN_COND_* function if possible.
5764 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5766 (SVE_COND_FP_BINARY): Include them.
5767 (optab, sve_fp_op): Handle them.
5768 (SVE_INT_BINARY_REV): New code iterator.
5769 (SVE_COND_FP_BINARY_REV): New int iterator.
5770 (commutative): New int attribute.
5771 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5773 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5775 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5776 (*cond_<optab><mode>): New patterns for reversed operands.
5778 2018-05-25 Richard Biener <rguenther@suse.de>
5780 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5781 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5782 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5783 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5784 * tree-vect-data-refs.c (everywhere): Adjust users.
5785 * tree-vect-loop.c (everywhere): Likewise.
5786 * tree-vect-slp.c (everywhere): Likewise.
5787 * tree-vect-stmts.c (everywhere): Likewise.
5788 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5790 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5792 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5794 (gcc_cv_as_section_exclude): ... this.
5795 Try Solaris as #exclude syntax.
5796 * configure: Regenerate.
5797 * config.in: Regenerate.
5798 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5800 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5801 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5803 * varasm.c (default_elf_asm_named_section): Don't check if
5804 HAVE_GAS_SECTION_EXCLUDE is defined.
5806 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5808 * doc/md.texi: Update the documentation of the cond_* optabs
5809 to mention the new final operand. Fix GET_MODE_NUNITS call.
5810 Describe the scalar case too.
5811 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5812 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5814 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5815 (get_conditional_internal_fn): Update comment.
5816 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5817 accumulator value as a final argument to conditional functions.
5818 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5819 a define_expand and add an "else" operand. Assert for now that
5820 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5821 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5822 (*cond_<optab><mode>): New patterns.
5823 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5824 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5825 (UNSPEC_COND_EOR): Delete.
5826 (optab): Remove associated mappings.
5827 (SVE_INT_BINARY): New code iterator.
5828 (sve_int_op): Remove int attribute and add "minus" to the code
5830 (SVE_COND_INT_OP): Delete.
5831 (SVE_COND_FP_OP): Rename to...
5832 (SVE_COND_FP_BINARY): ...this.
5834 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5836 * optabs.c (can_reuse_operands_p): New function.
5837 (maybe_legitimize_operands): Try to reuse the results for
5840 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5842 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5843 Add {q} suffix to insn mnemonic.
5845 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5847 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5848 (msp430_warn_func_return): New.
5850 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5852 * fold-const.c (tree_nonzero_bits): New function.
5853 * fold-const.h (tree_nonzero_bits): Likewise.
5854 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5855 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5857 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5861 * varasm.c (assemble_alias): Check ifunc_resolver only on
5864 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5867 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5868 when memory input operand is handled.
5870 2018-05-24 Luis Machado <luis.machado@linaro.org>
5872 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5874 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5876 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5878 * match.pd: Delay FMA folds until after vectorization.
5880 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5883 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5884 address check not strict.
5886 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5888 * gimple-match.h (gimple_match_op): New class.
5889 (mprts_hook): Replace parameters with a gimple_match_op *.
5890 (maybe_build_generic_op): Likewise.
5891 (gimple_simplified_result_is_gimple_val): Replace parameters with
5892 a const gimple_match_op *.
5893 (gimple_simplify): Replace code_helper * and tree * parameters with
5894 a gimple_match_op * parameter.
5895 (gimple_resimplify1): Replace code_helper *, tree and tree *
5896 parameters with a gimple_match_op * parameter.
5897 (gimple_resimplify2): Likewise.
5898 (gimple_resimplify3): Likewise.
5899 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5900 parameters with a gimple_match_op * parameter.
5901 * gimple-match-head.c (gimple_simplify): Change prototypes of
5902 auto-generated functions to take a gimple_match_op * instead of
5903 separate code_helper * and tree * parameters. Make the same
5904 change in the top-level overload and update calls to the
5905 gimple_resimplify routines. Update calls to the auto-generated
5906 functions and to maybe_push_res_to_seq in the publicly-facing
5907 operation-specific gimple_simplify overloads.
5908 (gimple_match_op::MAX_NUM_OPS): Define.
5909 (gimple_resimplify1): Replace rcode and ops with a single res_op
5910 parameter. Update call to gimple_simplify.
5911 (gimple_resimplify2): Likewise.
5912 (gimple_resimplify3): Likewise.
5913 (mprts_hook): Replace parameters with a gimple_match_op *.
5914 (maybe_build_generic_op): Likewise.
5915 (build_call_internal): Replace type, nargs and ops with
5916 a gimple_match_op *.
5917 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5918 with a single gimple_match_op *. Update calls to mprts_hook,
5919 build_call_internal and gimple_simplified_result_is_gimple_val.
5920 Factor out code that is common to the tree_code and combined_fn cases.
5921 * genmatch.c (expr::gen_transform): Replace tem_code and
5922 tem_ops with a gimple_match_op called tem_op. Update calls
5923 to the gimple_resimplify functions and maybe_push_res_to_seq.
5924 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5925 res_ops. Update call to the gimple_resimplify functions.
5926 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5927 (decision_tree::gen): Make the functions take a gimple_match_op *
5928 called res_op instead of separate res_code and res_ops parameters.
5929 Update call accordingly.
5930 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5931 and ops with a single res_op parameter. Update calls to
5932 maybe_build_generic_op and maybe_push_res_to_seq.
5933 (fold_stmt_1): Update calls to gimple_simplify and
5934 replace_stmt_with_simplification.
5935 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5936 and gimple_simplified_result_is_gimple_val.
5937 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5939 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5940 with a gimple_match_op *.
5941 (vn_nary_build_or_lookup): Likewise. Update call to
5942 vn_nary_build_or_lookup_1.
5943 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5944 gimple_match_op *. Update calls to the gimple_resimplify routines
5945 and to gimple_simplified_result_is_gimple_val.
5946 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5947 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5948 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5949 (visit_nary_op): Likewise.
5950 (visit_reference_op_load): Likewise.
5952 2018-05-23 Luis Machado <luis.machado@linaro.org>
5954 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5955 modifier for printing the step amount.
5957 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5960 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5963 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5965 * doc/sourcebuild.texi (Endianness): New subsubsection.
5967 2018-05-23 Luis Machado <luis.machado@linaro.org>
5969 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5970 <prefetch_dynamic_strides>: New const bool field.
5971 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5972 prefetch_dynamic_strides.
5973 (exynosm1_prefetch_tune): Likewise.
5974 (thunderxt88_prefetch_tune): Likewise.
5975 (thunderx_prefetch_tune): Likewise.
5976 (thunderx2t99_prefetch_tune): Likewise.
5977 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5979 (aarch64_override_options_internal): Update to set
5980 PARAM_PREFETCH_DYNAMIC_STRIDES.
5981 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5982 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5983 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5984 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5985 prefetch-dynamic-strides setting.
5987 2018-05-23 Luis Machado <luis.machado@linaro.org>
5989 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5990 <minimum_stride>: New const int field.
5991 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5992 minimum_stride field defaulting to -1.
5993 (exynosm1_prefetch_tune): Likewise.
5994 (thunderxt88_prefetch_tune): Likewise.
5995 (thunderx_prefetch_tune): Likewise.
5996 (thunderx2t99_prefetch_tune): Likewise.
5997 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5998 <default_opt_level>: Set to 3.
5999 (aarch64_override_options_internal): Update to set
6000 PARAM_PREFETCH_MINIMUM_STRIDE.
6001 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6002 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6003 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6004 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6005 stride is constant and is below the minimum stride threshold.
6007 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6009 * config/arm/arm-cpus.in (mode26): Delete.
6010 (armv4): Delete mode26 reference.
6011 * config/arm/arm.c (arm_configure_build_target): Delete use of
6014 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6016 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6018 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6019 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6021 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6022 (floatunsdidf2): Ditto.
6024 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6026 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6027 (fixuns_trunc<mode>si2_avx512f): Ditto.
6028 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6029 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6030 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6032 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6034 PR rtl-optimization/79985
6035 * df-scan.c (df_insn_refs_collect): Remove special case for
6036 global registers and asm statements.
6038 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6040 * extend.texi (Global Register Variables): Rewrite the bullet list.
6041 Note that the register is available for allocation. Note that access
6042 via inline asm must use constraints. Add note about async-signal
6043 handlers. Remove paragraph about automagic register selection.
6045 2018-05-23 Richard Biener <rguenther@suse.de>
6047 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6048 of fixed offset from memset VN.
6050 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6052 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6054 (alloc_cand_and_find_basis): Initialize first_interp field.
6055 (slsr_process_mul): Modify first_interp field.
6056 (slsr_process_add): Likewise.
6057 (slsr_process_cast): Modify first_interp field for each new
6059 (slsr_process_copy): Likewise.
6060 (dump_candidate): Dump first_interp field.
6061 (replace_mult_candidate): Process all interpretations, not just
6063 (replace_rhs_if_not_dup): Likewise.
6064 (replace_one_candidate): Likewise.
6066 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6068 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6070 (aarch64_needs_frame_chain): New function.
6071 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6073 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6076 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6077 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6078 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6079 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6080 as true for strict-align.
6081 (aarch64_can_inline_p): Perform checks even when callee has no
6082 attributes to check for strict alignment.
6083 * doc/extend.texi (AArch64 Function Attributes): Document
6085 * doc/invoke.texi: (AArch64 Options): Likewise.
6087 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6089 PR tree-optimization/85853
6090 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6091 the handling of the root of the node to...
6092 (vect_slp_analyze_node_operations_1): ...this new function,
6093 and run the whole thing with the child nodes' def types
6094 set according to their SLP node's def type.
6096 2018-05-23 Richard Biener <rguenther@suse.de>
6099 * tree-data-ref.c (create_runtime_alias_checks): Defer
6100 and ignore overflow warnings.
6102 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6104 PR tree-optimization/85822
6105 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6108 2018-05-23 Richard Biener <rguenther@suse.de>
6110 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6111 memset constants via native_interpret_expr.
6113 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6116 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6118 (cgraph_node::create_alias): Likewise.
6119 (cgraph_node::get_availability): Check ifunc_resolver instead
6120 of looking up ifunc attribute.
6121 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6122 * varasm.c (do_assemble_alias): Likewise.
6123 (assemble_alias): Likewise.
6124 (default_binds_local_p_3): Likewise.
6125 * cgraph.h (cgraph_node): Add ifunc_resolver.
6126 (cgraph_node::only_called_directly_or_aliased_p): Return false
6128 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6130 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6131 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6132 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6133 instead of looking up ifunc attribute.
6135 2018-05-22 Luis Machado <luis.machado@linaro.org>
6137 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6139 2018-05-22 Martin Sebor <msebor@redhat.com>
6142 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6143 only when expasion succeeds.
6144 (expand_builtin_strcmp): Same.
6145 (expand_builtin_strncmp): Same.
6147 2018-05-22 Martin Sebor <msebor@redhat.com>
6149 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6151 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6152 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6154 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6155 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6157 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6158 Add check for legitimate_address.
6159 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6160 (aarch64_swap_ldrstr_operands): New.
6161 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6164 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6165 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6167 * config/aarch64/aarch64.md: New patterns to generate stp
6169 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6170 single words and double words.
6171 (load_pair_sw, load_pair_dw): Likewise.
6172 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6174 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6176 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6177 for different mode ldpstp and add peephole for merged zero stores.
6179 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6181 (aarch64_gen_store_pair): Rename calls to match new patterns.
6182 (aarch64_gen_load_pair): Rename calls to match new patterns.
6183 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6184 (load_pair<DREG:mode><DREG2:mode>): ... This.
6185 (store_pair<mode>): Rename to...
6186 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6187 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6189 (V_INT_EQUIV): Handle SImode.
6190 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6193 2018-05-22 Martin Sebor <msebor@redhat.com>
6196 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6197 or ajust the presumed bound on an operation to avoid unnecessary
6200 2018-05-22 Martin Sebor <msebor@redhat.com>
6202 PR tree-optimization/85826
6203 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6204 assuming that a DECL necesarily has a constant size.
6206 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6209 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6211 2018-05-22 Richard Biener <rguenther@suse.de>
6213 PR tree-optimization/85834
6214 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6215 non-constant and non-zero memset arguments.
6217 2018-05-22 Martin Liska <mliska@suse.cz>
6220 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6222 2018-05-22 Richard Biener <rguenther@suse.de>
6224 PR tree-optimization/85863
6225 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6226 comparisons when vectype is specified.
6227 (vectorizable_condition): Do not specify vectype for
6228 vect_is_simple_cond when SLP vectorizing.
6230 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6233 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6234 define __ibm128 as long double.
6235 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6236 as a distinct type when IEEE 128-bit support is enabled.
6237 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6238 128-bit types to use the correct functions.
6239 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6240 convert between 128-bit floating point types that have different
6241 modes but the same representation, instead of using gen_lowpart to
6243 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6245 (IFKF_reg): New attributes to give the register constraints for
6247 (extend<mode>tf2_internal): New insns to mark an explicit
6248 conversion between 128-bit floating point types that have a
6249 different mode but share the same representation.
6251 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6253 PR tree-optimization/85814
6254 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6255 a null return from get_strinfo when unsharing the next
6256 strinfo in the chain.
6258 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6261 * varasm.c (weak_finish): Clean up weak_decls.
6263 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6265 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6266 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6267 UNSPEC_UADALP values.
6268 * config/aarch64/iterators.md (ABAL): New int iterator.
6271 (sur): Add mappings for the above.
6272 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6274 (aarch64_<sur>abal<mode>_4): Likewise.
6275 (aarch64_<sur>adalp<mode>_3): Likewise.
6276 (<sur>sadv16qi): New define_expand.
6278 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6280 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6281 (*movdf_internal): Ditto.
6282 (*rcpsf2_sse): Ditto.
6283 (*rsqrtsf2_sse): Ditto.
6284 (*sqrt<mode>2_sse): Ditto.
6286 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6288 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6290 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6291 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6292 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6293 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6295 * config/aarch64/arm_neon.h: Likewise.
6296 * config/aarch64/iterators.md (VQ_I): New.
6298 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6300 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6302 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6305 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6307 * config/nds32/constraints.md (S): New constraint.
6308 * config/nds32/nds32.md (call_internal): Use constraint S.
6309 (call_value_internal): Likewise.
6310 (sibcall_internal): Likewise.
6311 (sibcall_value_internal): Likewise.
6313 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6314 Chung-Ju Wu <jasonwucj@gmail.com>
6316 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6319 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6320 Chung-Ju Wu <jasonwucj@gmail.com>
6322 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6323 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6324 (nds32_rtx_costs_impl): Simplify.
6325 (nds32_address_cost_impl): Simplify.
6326 (nds32_init_rtx_costs): New function.
6327 (nds32_rtx_costs_speed_prefer): Likewise.
6328 (nds32_rtx_costs_size_prefer): Likewise.
6329 (nds32_address_cost_speed_prefer): Likewise.
6330 (nds32_address_cost_speed_fwprop): Likewise.
6331 (nds32_address_cost_size_prefer): Likewise.
6332 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6333 * config/nds32/nds32.c (nds32_option_override): Use
6334 nds32_init_rtx_costs function.
6336 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6338 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6339 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6340 (TARGET_PIPELINE_N8): Likewise.
6341 (TARGET_PIPELINE_N10): Likewise.
6342 (TARGET_PIPELINE_N13): Likewise.
6343 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6345 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6347 * config/nds32/nds32-fpu.md: Update copyright year.
6349 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6351 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6353 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6355 * config/nds32/nds32.c
6356 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6357 * config/nds32/nds32.opt (minline-asm-r15): New option.
6359 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6361 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6363 * config/nds32/nds32.md (abssi2): New pattern.
6365 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6367 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6368 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6369 Merge insn pattern from sse_cvtsi2ss<round_name> and
6370 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6371 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6372 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6373 using SWI48 mode iterator.
6374 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6375 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6376 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6377 pattern from sse_cvttss2si<round_saeonly_name>
6378 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6379 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6380 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6381 using SWI48 mode iterator.
6382 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6383 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6384 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6385 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6386 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6387 using SWI48 mode iterator.
6388 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6389 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6390 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6391 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6392 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6393 SWI48 mode iterator.
6394 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6395 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6396 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6397 pattern from sse_cvttsd2si<round_saeonly_name>
6398 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6400 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6402 * config/nds32/nds32-md-auxiliary.c
6403 (nds32_valid_smw_lwm_base_p): Refine.
6404 (nds32_output_smw_single_word): Refine.
6405 (nds32_output_smw_double_word): New.
6406 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6408 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6410 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6411 (nds32_output_stack_pop): Refine.
6412 (nds32_expand_unaligned_load): Refine.
6413 (nds32_expand_unaligned_store): Refine.
6415 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6416 Chung-Ju Wu <jasonwucj@gmail.com>
6418 * config/nds32/constants.md: Add TP_REGNUM constant.
6419 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6420 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6422 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6423 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6424 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6425 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6426 and PIC code generation.
6427 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6429 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6431 * config/nds32/nds32.md: Support TLS and PIC.
6432 * config/nds32/nds32.c: Support TLS and PIC.
6433 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6434 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6437 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6439 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6440 mode with E_ prefix.
6442 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6443 Chung-Ju Wu <jasonwucj@gmail.com>
6445 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6446 * config/nds32/nds32-md-auxiliary.c
6447 (symbolic_reference_mentioned_p): New.
6448 (nds32_legitimize_ict_address): New.
6449 (nds32_expand_ict_move): New.
6450 (nds32_indirect_call_referenced_p): New.
6451 (nds32_symbol_binds_local_p): Delete.
6452 (nds32_long_call_p): Modify.
6453 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6454 * config/nds32/nds32-protos.h
6455 (symbolic_reference_mentioned_p): Declare.
6456 (nds32_legitimize_ict_address): Declare.
6457 (nds32_expand_ict_move): Declare.
6458 (nds32_indirect_call_referenced_p): Declare.
6459 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6460 (nds32_relax_group): Use nds32_ict_const_p as condition.
6461 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6462 (nds32_asm_file_start): Output ict_model directive in asm code.
6463 (nds32_legitimate_address_p): Consider indirect call.
6464 (nds32_print_operand): Consider indirect call.
6465 (nds32_print_operand_address): Consider indirect call.
6466 (nds32_insert_attributes): Handle "indirect_call" attribute.
6467 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6468 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6469 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6470 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6471 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6472 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6473 (TARGET_ICT_MODEL_SMALL): Define.
6474 (TARGET_ICT_MODEL_LARGE): Define.
6475 * config/nds32/nds32.md (movsi): Consider ict model.
6476 (call, call_value): Consider ict model.
6477 (sibcall, sibcall_value): Consider ict model.
6478 * config/nds32/nds32.opt (mict-model): New option.
6479 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6482 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6483 Monk Chiang <sh.chiang04@gmail.com>
6484 Jim Wilson <jimw@sifive.com>
6486 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6487 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6488 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6489 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6490 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6491 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6492 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6493 compute save_libcall_adjustment properly.
6494 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6495 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6496 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6497 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6498 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6499 (ABI_SPEC): Handle mabi=ilp32e.
6500 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6501 (RVE): Add RVE mask.
6502 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6503 <-march>: Add rv32e as an example.
6505 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6508 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6509 (intra_create_variable_infos): Handle C++ constructors.
6511 2018-05-18 Martin Liska <mliska@suse.cz>
6513 * passes.def: Remove a redundant pass.
6515 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6518 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6520 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6522 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6524 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6526 (armv2, armv2a, armv3, armv3m): Delete architectures.
6527 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6528 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6529 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6531 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6532 (*mulsidi3adddi): Likewise.
6533 (mulsidi3): Likewise.
6534 (*mulsidi3_nov6): Likewise.
6535 (umulsidi3): Likewise.
6536 (umulsidi3_nov6): Likewise.
6537 (umaddsidi4): Likewise.
6538 (*umulsidi3adddi): Likewise.
6539 (smulsi3_highpart): Likewise.
6540 (*smulsi3_highpart_nov6): Likewise.
6541 (umulsi3_highpart): Likewise.
6542 (*umulsi3_highpart_nov6): Likewise.
6543 * config/arm/arm.h (arm_arch3m): Delete.
6544 * config/arm/arm.c (arm_arch3m): Delete.
6545 (arm_option_override_internal): Update armv3-related comment.
6546 (arm_configure_build_target): Delete use of isa_bit_mode32.
6547 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6548 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6549 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6551 (mulusa3): Likewise.
6552 * config/arm/arm-protos.h (arm_arch3m): Delete.
6553 * config/arm/arm-tables.opt: Regenerate.
6554 * config/arm/arm-tune.md: Likewise.
6555 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6556 deleted architectures.
6558 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6560 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6561 (armv5t, armv5te): New features.
6562 (ARMv5, ARMv5e): Delete fgroups.
6563 (ARMv5t, ARMv5te): Adjust for above changes.
6565 (armv5, armv5e): Delete arches.
6566 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6568 (*call_reg_arm): Likewise.
6569 (*call_value_reg_armv5): Likewise.
6570 (*call_value_reg_arm): Likewise.
6571 (*call_symbol): Likewise.
6572 (*call_value_symbol): Likewise.
6573 (*sibcall_insn): Likewise.
6574 (*sibcall_value_insn): Likewise.
6576 (prefetch): Likewise.
6577 (define_split and define_peephole2 dependent on arm_arch5):
6579 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6581 (TARGET_ARM_QBIT): Likewise.
6582 (TARGET_DSP_MULTIPLY): Likewise.
6583 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6584 (arm_arch5, arm_arch5e): Delete.
6585 (arm_arch5t, arm_arch5te): Declare.
6586 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6587 (arm_arch5t): Declare.
6588 (arm_option_reconfigure_globals): Update for the above.
6589 (arm_options_perform_arch_sanity_checks): Update comment, replace
6590 use of arm_arch5 with arm_arch5t.
6591 (use_return_insn): Likewise.
6592 (arm_emit_call_insn): Likewise.
6593 (output_return_instruction): Likewise.
6594 (arm_final_prescan_insn): Likewise.
6595 (arm_coproc_builtin_available): Likewise.
6596 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6597 arm_arch5e with arm_arch5t and arm_arch5te.
6598 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6599 (arm_arch5t, arm_arch5te): Declare.
6600 * config/arm/arm-tables.opt: Regenerate.
6601 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6602 * config/arm/t-multilib: Likewise.
6603 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6604 instead of arm_arch5.
6605 (*call_reg_thumb1): Likewise.
6606 (*call_value_reg_thumb1_v5): Likewise.
6607 (*call_value_reg_thumb1): Likewise.
6608 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6610 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6612 2018-05-18 Martin Liska <mliska@suse.cz>
6614 PR gcov-profile/84846
6615 * doc/gcov.texi: Document -t option of gcov tool.
6617 2018-05-18 Martin Liska <mliska@suse.cz>
6619 PR gcov-profile/84846
6620 * gcov.c (print_usage): Add new -t option.
6621 (process_args): Handle the option.
6622 (generate_results): Use stdout as output when requested by
6625 2018-05-18 Martin Liska <mliska@suse.cz>
6627 PR gcov-profile/84846
6628 * coverage.c (coverage_init): Write PWD to .gcno file.
6629 * doc/gcov.texi: Document how working directory is printed.
6630 * gcov-dump.c (dump_gcov_file): Print PWD.
6631 * gcov.c (output_intermediate_file): Likewise.
6632 (read_graph_file): Read PWD string.
6633 (output_lines): Print PWD.
6635 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6638 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6639 for retval and return false if all args to phi are zero.
6641 2018-05-18 Richard Biener <rguenther@suse.de>
6643 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6645 (evrp_dom_walker::before_dom_children): Call it.
6647 2018-05-18 Richard Biener <rguenther@suse.de>
6649 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6650 results when processing array refs with variable index.
6652 2018-05-18 Toon Moene <toon@moene.org>
6654 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6655 directly after that of -floop-interchange. Indicate that both
6656 options are enabled by default when specifying -O3.
6658 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6660 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6661 iterator. Delete separate integer-mode vec_set<mode> expander.
6662 (aarch64_simd_vec_setv2di): Delete.
6663 (vec_setv2di): Delete.
6664 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6665 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6666 the "w, r" alternative.
6668 2018-05-18 Martin Liska <mliska@suse.cz>
6670 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6671 * tree-pass.h (make_pass_lower_switch_O0): New function.
6672 * tree-switch-conversion.c (node_has_low_bound): Remove.
6673 (node_has_high_bound): Likewise.
6674 (node_is_bounded): Likewise.
6675 (class pass_lower_switch): Make it a template type and create
6677 (pass_lower_switch::execute): Add template argument.
6678 (make_pass_lower_switch): New function.
6679 (make_pass_lower_switch_O0): New function.
6680 (do_jump_if_equal): Remove.
6681 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6682 all the hard work to tree optimization passes.
6684 2018-05-18 Martin Liska <mliska@suse.cz>
6686 * dbgcnt.c (limit_low): Renamed from limit.
6687 (limit_high): New variable.
6688 (dbg_cnt_is_enabled): Check for upper limit.
6689 (dbg_cnt): Adjust dumping.
6690 (dbg_cnt_set_limit_by_index): Add new argument for high
6692 (dbg_cnt_set_limit_by_name): Likewise.
6693 (dbg_cnt_process_single_pair): Parse new format.
6694 (dbg_cnt_process_opt): Use strtok.
6695 (dbg_cnt_list_all_counters): Remove 'value' and add
6697 * doc/invoke.texi: Document changes.
6699 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6701 * doc/sourcebuild.texi (scalar_all_fma): Document.
6702 * tree.def (FMA_EXPR): Delete.
6703 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6704 * internal-fn.c (ternary_direct): New macro.
6705 (expand_ternary_optab_fn): Likewise.
6706 (direct_ternary_optab_supported_p): Likewise.
6707 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6708 * builtins.c (fold_builtin_fma): Delete.
6709 (fold_builtin_3): Don't call it.
6710 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6711 * expr.c (expand_expr_real_2): Likewise.
6712 * fold-const.c (operand_equal_p): Likewise.
6713 (fold_ternary_loc): Likewise.
6714 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6715 * gimple.c (DEFTREECODE): Likewise.
6716 * gimplify.c (gimplify_expr): Likewise.
6717 * optabs-tree.c (optab_for_tree_code): Likewise.
6718 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6719 * tree-eh.c (operation_could_trap_p): Likewise.
6720 (stmt_could_throw_1_p): Likewise.
6721 * tree-inline.c (estimate_operator_cost): Likewise.
6722 * tree-pretty-print.c (dump_generic_node): Likewise.
6723 (op_code_prio): Likewise.
6724 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6725 * tree-ssa-operands.c (get_expr_operands): Likewise.
6726 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6727 * fold-const-call.h (fold_fma): Delete.
6728 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6729 CFN_FNMA and CFN_FNMS.
6731 * genmatch.c (combined_fn): New enum.
6732 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6733 (commutative_op): New function.
6734 (commutate): Use it. Handle more than 2 operands.
6735 (dt_operand::gen_gimple_expr): Use commutative_op.
6736 (parser::parse_expr): Allow :c to be used with non-binary
6737 operators if the commutative operand is known.
6738 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6739 CFN_FMS, CFN_FNMA and CFN_FNMS.
6740 (backprop::process_assign_use): Remove FMA_EXPR handling.
6741 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6742 (gen_hsa_fma): New function.
6743 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6744 IFN_FNMA and IFN_FNMS.
6745 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6746 * gimple-fold.h (follow_all_ssa_edges): Declare.
6747 * gimple-fold.c (follow_all_ssa_edges): New function.
6748 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6749 gimple_build interface and use follow_all_ssa_edges to fold the result.
6750 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6751 instead of checking for optabs directly.
6752 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6753 rather than FMA_EXPRs.
6754 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6755 call to IFN_FMA instead of an FMA_EXPR.
6757 2018-05-17 Jim Wilson <jimw@sifive.com>
6759 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6760 sign extended promoted subreg, and the range does not have the sign bit
6761 set, then do a sign extend.
6763 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6764 test, check for sign extended subreg and/or constant operands, and
6765 do a sign extend in that case.
6767 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6769 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6770 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6772 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6773 Change logics_shift_reg to logics_shift_imm.
6774 (thunderx2t99_fp_loadpair_basic): Delete.
6775 (thunderx2t99_fp_storepair_basic): Delete.
6776 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6777 (thunderx2t99_asimd_polynomial): Delete.
6778 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6779 and neon_fp_mul_d_scalar_q.
6780 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6781 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6782 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6783 (thunderx2t99_asimd_lut): Add missing tbl types.
6784 (thunderx2t99_asimd_ext): Delete.
6785 (thunderx2t99_asimd_load1_1_mult): Delete.
6786 (thunderx2t99_asimd_load1_2_mult): Delete.
6787 (thunderx2t99_asimd_load1_ldp): New.
6788 (thunderx2t99_asimd_load1): New.
6789 (thunderx2t99_asimd_load2): Add missing *load2* types.
6790 (thunderx2t99_asimd_load3): New.
6791 (thunderx2t99_asimd_load4): New.
6792 (thunderx2t99_asimd_store1_1_mult): Delete.
6793 (thunderx2t99_asimd_store1_2_mult): Delete.
6794 (thunderx2t99_asimd_store2_mult): Delete.
6795 (thunderx2t99_asimd_store2_onelane): Delete.
6796 (thunderx2t99_asimd_store_stp): New.
6797 (thunderx2t99_asimd_store1): New.
6798 (thunderx2t99_asimd_store2): New.
6799 (thunderx2t99_asimd_store3): New.
6800 (thunderx2t99_asimd_store4): New.
6802 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6804 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6805 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6807 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6808 Segher Boessenkool <segher@kernel.crashing.org>
6811 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6814 2018-05-17 Richard Biener <rguenther@suse.de>
6816 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6817 for pruning loop and prune defs feeding only already visited PHIs.
6819 2018-05-17 Richard Biener <rguenther@suse.de>
6821 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6823 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6824 Richard Biener <rguenther@suse.de>
6826 PR tree-optimization/85793
6827 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6828 for VMAT_ELEMENTWISE.
6830 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6832 * internal-fn.h (lookup_internal_fn): Declare
6833 * internal-fn.c (lookup_internal_fn): New function.
6834 * gimple.c (gimple_build_call_from_tree): Handle calls to
6836 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6837 internal function names.
6838 * tree-pretty-print.c (dump_generic_node): Likewise.
6839 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6841 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6843 * gimple-fold.h (gimple_build): Make the function forms take
6844 combined_fn rather than built_in_function.
6845 (gimple_simplify): Likewise.
6846 * gimple-match-head.c (gimple_simplify): Likewise.
6847 * gimple-fold.c (gimple_build): Likewise.
6848 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6849 rather than gimple_build_call_internal.
6850 (get_initial_defs_for_reduction): Likewise.
6851 (vect_create_epilog_for_reduction): Likewise.
6852 (vectorizable_live_operation): Likewise.
6854 2018-05-17 Martin Liska <mliska@suse.cz>
6856 * gimple-ssa-sprintf.c (format_directive): Do not use
6857 space in between 'G_' and '('.
6859 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6862 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6863 even if the mask is not all ones.
6866 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6868 (ix86_gimple_fold_builtin): Likewise.
6871 * config/i386/i386.c: Include tree-vector-builder.h.
6872 (ix86_vector_shift_count): New function.
6873 (ix86_fold_builtin): Fold shift builtins by scalar count.
6874 (ix86_gimple_fold_builtin): Likewise.
6876 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6877 _mm512_setzero): New intrinsics.
6879 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6880 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6882 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6883 code generation for cases where splatting a value is not useful.
6884 * simplify-rtx.c (simplify_ternary_operation): Simplify
6885 vec_merge across a vec_duplicate and a paradoxical subreg forming
6886 a vector mode to a vec_concat.
6888 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6890 * config.gcc: Support "goldmont-plus".
6891 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6893 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6894 PROCESSOR_GOLDMONT_PLUS.
6895 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6896 (processor_target_table): Add "goldmont-plus".
6897 (PTA_GOLDMONT_PLUS): Define.
6898 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6899 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6900 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6901 (fold_builtin_cpu): Add "goldmont-plus".
6902 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6903 (ix86_option_override_internal): Add "goldmont-plus".
6904 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6905 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6906 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6907 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6909 2018-05-17 Richard Biener <rguenther@suse.de>
6911 PR tree-optimization/85757
6912 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6913 remove defs that only feed that PHI from further processing.
6915 2018-05-16 Jim Wilson <jimw@sifive.com>
6917 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6919 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6921 2018-05-16 Mark Wielaard <mark@klomp.org>
6923 * dwarf2out.c (count_index_strings): New function.
6924 (output_indirect_strings): Call count_index_strings and generate
6925 header for dwarf_version >= 5.
6927 2018-05-16 Mark Wielaard <mark@klomp.org>
6929 * dwarf2out.c (dwarf_FORM): New function.
6930 (set_indirect_string): Use dwarf_FORM.
6931 (reset_indirect_string): Likewise.
6932 (size_of_die): Likewise.
6933 (value_format): Likewise.
6934 (output_die): Likewise.
6935 (add_skeleton_AT_string): Likewise.
6936 (output_macinfo_op): Likewise.
6937 (index_string): Likewise.
6938 (output_index_string_offset): Likewise.
6939 (output_index_string): Likewise.
6940 (count_index_strings): Likewise.
6942 2018-05-16 Carl Love <cel@us.ibm.com>
6944 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6945 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6947 2018-05-16 Martin Jambor <mjambor@suse.cz>
6949 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6950 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6952 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6954 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6955 (fnma<mode>4): Likewise.
6956 (fms<mode>4): Likewise.
6957 (fnms<mode>4): Likewise.
6958 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6959 (aarch64_fnma<mode>4): Likewise.
6960 (aarch64_fms<mode>4): Likewise.
6961 (aarch64_fnms<mode>4): Likewise.
6962 (aarch64_fnmadd<mode>4): Likewise.
6964 2018-05-16 Jason Merrill <jason@redhat.com>
6966 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6968 2018-05-16 Richard Biener <rguenther@suse.de>
6970 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6971 (dump_stmt_cost): Declare.
6972 (add_stmt_cost): Dump cost we add.
6973 (add_stmt_costs): New function.
6974 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6976 (vect_analyze_stmt): Adjust prototype.
6977 (vectorizable_condition): Likewise.
6978 (vectorizable_live_operation): Likewise.
6979 (vectorizable_reduction): Likewise.
6980 (vectorizable_induction): Likewise.
6981 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6982 cost vector to pass to vectorizable_ and record afterwards.
6983 (vect_model_reduction_cost): Take cost vector argument and adjust.
6984 (vect_model_induction_cost): Likewise.
6985 (vectorizable_reduction): Likewise.
6986 (vectorizable_induction): Likewise.
6987 (vectorizable_live_operation): Likewise.
6988 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6989 SLP_TREE_NUMBER_OF_VEC_STMTS.
6990 (vect_analyze_slp_cost_1): Remove.
6991 (vect_analyze_slp_cost): Likewise.
6992 (vect_slp_analyze_node_operations): Take visited args and
6993 a target cost vector. Avoid processing already visited stmt sets.
6994 (vect_slp_analyze_operations): Use a local cost vector to gather
6995 costs and register those of non-discarded instances.
6996 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6997 (vect_schedule_slp_instance): Remove copying of
6998 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7000 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7001 adding cost. Record cost entry location.
7002 (vect_prologue_cost_for_slp_op): Function to compute cost of
7003 a constant or invariant generated for SLP vect in the prologue,
7004 split out from vect_analyze_slp_cost_1.
7005 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7006 (vect_model_promotion_demotion_cost): Likewise.
7007 (vect_model_store_cost): Likewise, make static.
7008 (vect_model_load_cost): Likewise.
7009 (vectorizable_bswap): Add cost vector arg and adjust.
7010 (vectorizable_call): Likewise.
7011 (vectorizable_simd_clone_call): Likewise.
7012 (vectorizable_conversion): Likewise.
7013 (vectorizable_assignment): Likewise.
7014 (vectorizable_shift): Likewise.
7015 (vectorizable_operation): Likewise.
7016 (vectorizable_store): Likewise.
7017 (vectorizable_load): Likewise.
7018 (vectorizable_condition): Likewise.
7019 (vectorizable_comparison): Likewise.
7020 (can_vectorize_live_stmts): Likewise.
7021 (vect_analyze_stmt): Likewise.
7022 (vect_transform_stmt): Adjust calls to vectorizable_*.
7023 * tree-vectorizer.c: Include gimple-pretty-print.h.
7024 (dump_stmt_cost): New function.
7026 2018-05-16 Richard Biener <rguenther@suse.de>
7028 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7029 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7030 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7031 (check_name): New callback.
7032 (dse_classify_store): Track cycles via a visited bitmap of PHI
7033 defs and simplify handling of in-loop and across loop dead stores
7034 and properly fail for loop-variant refs. Handle byte-tracking with
7035 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7038 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7040 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7041 (vect_get_mask_type_for_stmt): Likewise.
7042 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7044 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7045 to determine the statement's vector type and the vector type that
7046 should be used for calculating nunits. Deal with cases in which
7047 the type has to be deferred.
7048 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7049 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7050 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7051 (vect_determine_vf_for_stmt): New functions, split out from...
7052 (vect_determine_vectorization_factor): ...here.
7053 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7054 (vect_get_mask_type_for_stmt): New functions, split out from
7055 vect_determine_vectorization_factor.
7057 2018-05-16 Richard Biener <rguenther@suse.de>
7059 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7060 verify the [VEC_]COND_EXPR embedded comparison.
7062 2018-05-15 Martin Sebor <msebor@redhat.com>
7064 PR tree-optimization/85753
7065 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7066 RECORD_TYPE in addition to ARRAY_TYPE.
7068 2018-05-15 Martin Sebor <msebor@redhat.com>
7071 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7073 2018-05-15 Richard Biener <rguenther@suse.de>
7075 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7076 add by_clobber_p one. Change algorithm to collect all defs
7077 representing uses we need to walk and try reducing them to
7078 a single one before failing.
7079 (dse_dom_walker::dse_optimize_stmt): Adjust.
7081 2018-05-13 Mark Wielaard <mark@klomp.org>
7083 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7084 (size_of_loc_descr): Likewise.
7085 (output_loc_operands): Likewise.
7086 (output_loc_operands_raw): Likewise.
7087 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7088 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7089 (hash_loc_operands): Likewise.
7090 (compare_loc_operands): Likewise.
7092 2018-05-14 Mark Wielaard <mark@klomp.org>
7094 * dwarf2out.c (count_index_addrs): New function.
7095 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7097 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7099 PR tree-optimization/83648
7100 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7101 return value as malloc candidate.
7103 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7106 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7107 param as true in call to suggest_attribute.
7109 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7111 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7112 -mreadonly-in-sdata.
7114 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7116 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7118 (aarch64_crypto_aesd_fused): Likewise.
7120 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7122 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7123 (movsi_aarch64): Likewise.
7124 (load_pairsi): Likewise.
7125 (load_pairdi): Likewise.
7126 (store_pairsi): Likewise.
7127 (store_pairdi): Likewise.
7128 (load_pairsf): Likewise.
7129 (load_pairdf): Likewise.
7130 (store_pairsf): Likewise.
7131 (store_pairdf): Likewise.
7132 (zero_extend): Likewise.
7133 (trunc): Swap alternatives.
7134 (fcvt_target): Add '?' to prefer w over r.
7136 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7139 * config/i386/i386.md: Disallow non-commutative arithmetics in
7140 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7141 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7142 in the peephole2 before it.
7144 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7146 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7147 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7148 (ix86_handle_option): Handle -mcldemote.
7149 * config.gcc: New header.
7150 * config/i386/cldemoteintrin.h: New file.
7151 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7152 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7154 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7155 OPTION_MASK_ISA_CLDEMOTE.
7156 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7157 (ix86_valid_target_attribute_inner_p): Ditto.
7158 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7159 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7160 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7161 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7162 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7164 * config/i386/i386.opt: Add -mcldemote.
7165 * config/i386/x86intrin.h: New header.
7166 * doc/invoke.texi: Add -mcldemote.
7168 2018-05-14 Richard Biener <rguenther@suse.de>
7170 * doc/match-and-simplify.texi: Adjust :s documentation.
7172 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7174 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7175 intended memcpy size.
7176 (REORDER_45): Likewise.
7178 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7180 * sort.cc: New file.
7181 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7182 * vec.c (qsort_chk): Use gcc_qsort.
7183 * Makefile.in (OBJS-libcommon): Add sort.o.
7184 (build/sort.o): New target. Use it...
7185 (BUILD_RTL): ... here, and...
7186 (build/gencfn-macros): ... here, and...
7187 (build/genmatch): ... here.
7189 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7190 Chung-Ju Wu <jasonwucj@gmail.com>
7192 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7193 * config/nds32/nds32-graywolf.md: New file.
7194 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7195 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7197 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7198 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7199 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7200 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7201 * config/nds32/pipelines.md: Include n15 settings.
7203 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7204 Chung-Ju Wu <jasonwucj@gmail.com>
7206 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7207 * config/nds32/nds32-n13.md: New file.
7208 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7209 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7211 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7212 * config/nds32/nds32.md (pipeline_model): Add n13.
7213 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7214 * config/nds32/pipelines.md: Include n13 settings.
7216 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7217 Chung-Ju Wu <jasonwucj@gmail.com>
7219 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7220 * config/nds32/nds32-n10.md: New file.
7221 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7222 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7224 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7225 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7226 * config/nds32/nds32.md (pipeline_model): Add n10.
7227 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7228 * config/nds32/pipelines.md: Include n10 settings.
7230 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7231 Kito Cheng <kito.cheng@gmail.com>
7232 Chung-Ju Wu <jasonwucj@gmail.com>
7234 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7235 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7236 Add enum values for DSP extension instructions.
7237 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7239 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7240 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7242 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7243 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7244 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7245 * config/nds32/nds32-intrinsic.md: Likewise.
7246 * config/nds32/nds32_intrinsic.h: Likewise.
7247 * config/nds32/nds32-md-auxiliary.c: Likewise.
7248 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7249 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7250 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7251 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7252 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7253 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7254 TYPE_DMAC in switch statement.
7255 * config/nds32/nds32.c: New checking and implementation for DSP
7256 extension instructions.
7257 * config/nds32/nds32.h: Likewise.
7258 * config/nds32/nds32.md: Likewise.
7259 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7260 * config/nds32/predicates.md: Implement new predicates for DSP
7263 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7265 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7266 Reformat alternatives and attributes so it is easier to identify
7267 which constraints/attributes go with which instruction.
7268 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7269 (mov<mode>_softfloat32, FMOVE64): Likewise.
7270 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7271 (mov<mode>_softfloat64, FMOVE64): Likewise.
7273 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7275 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7277 (Basic PowerPC Built-in Functions): The new name of the
7278 subsection previously known as "PowerPC Built-in Functions".
7279 (Basic PowerPC Built-in Functions Available on all Configurations):
7281 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7282 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7283 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7284 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7286 2018-05-11 Martin Jambor <mjambor@suse.cz>
7289 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7292 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7295 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7297 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7299 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7300 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7301 (ix86_handle_option): Handle -mwaitpkg.
7302 * config.gcc: New header.
7303 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7304 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7305 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7307 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7308 OPTION_MASK_ISA_WAITPKG.
7309 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7310 (ix86_option_override_internal): Add PTA_WAITPKG.
7311 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7312 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7313 IX86_BUILTIN_TPAUSE.
7314 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7315 __builtin_ia32_umwait and __builtin_ia32_tpause.
7316 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7317 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7318 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7319 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7320 UNSPECV_TPAUSE): New.
7321 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7322 * config/i386/i386.opt: Add -mwaitpkg.
7323 * config/i386/waitpkgintrin.h: New file.
7324 * config/i386/x86intrin.h: New header.
7325 * doc/invoke.texi: Add -mwaitpkg.
7327 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7330 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7332 (cortex-m0): Use armv6s-m isa.
7333 (cortex-m0plus): Likewise.
7334 (cortex-m1): Likewise.
7335 (cortex-m0.small-multiply): Likewise.
7336 (cortex-m0plus.small-multiply): Likewise.
7337 (cortex-m1.small-multiply): Likewise.
7339 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7340 Jakub Jelinek <jakub@redhat.com>
7342 PR tree-optimization/85692
7343 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7344 source permute as well.
7346 2018-05-11 Martin Liska <mliska@suse.cz>
7349 * doc/extend.texi: Document LLVM style format for no_sanitize
7352 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7354 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7355 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7356 (mode_supports_vsx_dform_quad): Likewise.
7357 (mode_supports_vmx_dform): Move these functions to be next to the
7358 other mode_supports functions.
7359 (mode_supports_dq_form): Likewise.
7360 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7361 mode_supports_dq_form.
7362 (reg_offset_addressing_ok_p): Likewise.
7363 (offsettable_ok_by_alignment): Likewise.
7364 (rs6000_legitimate_offset_address_p): Likewise.
7365 (legitimate_lo_sum_address_p): Likewise.
7366 (rs6000_legitimize_address): Likewise.
7367 (rs6000_legitimize_reload_address): Likewise.
7368 (rs6000_secondary_reload_inner): Likewise.
7369 (rs6000_preferred_reload_class): Likewise.
7370 (rs6000_output_move_128bit): Likewise.
7372 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7374 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7375 Generate SImode target register for null target.
7376 <case IX86_BUILTIN_XGETBV>: Ditto.
7377 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7378 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7380 2018-05-10 Carl Love <cel@us.ibm.com>
7382 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7383 dcbtt and dcbtstt if operands[2] is 0.
7385 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7388 * config/i386/sse.md (usadv64qi): New expander.
7390 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7392 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7393 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7394 -maltivec=be support.
7395 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7396 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7397 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7398 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7399 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7400 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7401 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7402 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7403 altivec_vsumsws): Adjust.
7404 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7405 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7406 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7408 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7409 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7410 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7411 (altivec_lve<VI_char>x): Delete expand.
7412 (*altivec_lve<VI_char>x_internal): Rename to...
7413 (altivec_lve<VI_char>x): ... this.
7414 (altivec_lvxl_<mode>): Delete expand.
7415 (*altivec_lvxl_<mode>_internal): Rename to ...
7416 (altivec_lvxl_<mode>): ... this.
7417 (altivec_stvxl_<mode>): Delete expand.
7418 (*altivec_stvxl_<mode>_internal): Rename to ...
7419 (altivec_stvxl_<mode>): ... this.
7420 (altivec_stve<VI_char>x): Delete expand.
7421 (*altivec_stve<VI_char>x_internal): Rename to ...
7422 (altivec_stve<VI_char>x): ... this.
7423 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7424 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7425 reduc_plus_scal_<mode>): Adjust.
7426 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7428 (rs6000_cpu_cpp_builtins): Adjust.
7429 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7430 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7431 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7432 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7433 -maltivec=be support.
7434 (rs6000_split_vec_extract_var): Adjust.
7435 (rs6000_split_v4si_init): Adjust.
7436 (swap_selector_for_mode): Delete.
7437 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7438 altivec_expand_stvex_be): Delete.
7439 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7440 -maltivec=be support.
7441 (rs6000_gimple_fold_builtin): Ditto.
7442 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7444 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7445 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7446 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7447 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7448 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7449 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7450 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7451 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7452 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7453 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7454 anonymous split): Adjust.
7455 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7456 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7458 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7460 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7461 when --with-gxx-include-dir is also specified.
7462 * configure: Regenerate.
7464 2018-05-09 Jim Wilson <jimw@sifive.com>
7467 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7468 * config/riscv/t-withmultilib: New.
7469 * config/riscv/withmultilib.h: New.
7470 * doc/install.texi: Document RISC-V --with-multilib-list support.
7472 2018-05-09 Richard Biener <rguenther@suse.de>
7474 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7476 (vect_bb_vectorization_profitable_p): Adjust. Compute
7477 actual scalar cost using the cost vector and the add_stmt_cost
7480 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7482 PR rtl-optimization/85645
7483 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7484 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7486 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7488 PR rtl-optimization/85645
7489 * shrink-wrap.c (spread_components): Return a boolean saying if
7490 anything was changed.
7491 (try_shrink_wrapping_separate): Iterate spread_components until
7492 nothing changes anymore.
7494 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7496 PR rtl-optimization/85645
7497 * regrename.c (build_def_use): Also kill the chains that include the
7498 destination of a REG_CFA_REGISTER note.
7500 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7502 PR rtl-optimization/85645
7503 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7504 insn that has a REG_CFA_REGISTER note.
7506 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7508 * cfgexpand.c (expand_clobber): New function.
7509 (expand_gimple_stmt_1): Use it.
7510 * tree-vect-stmts.c (vect_clobber_variable): New function,
7512 (vectorizable_simd_clone_call): ...here.
7513 (vectorizable_store): Emit a clobber either side of an
7514 IFN_STORE_LANES sequence.
7515 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7517 2018-05-09 Tom de Vries <tom@codesourcery.com>
7520 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7521 (define_insn "trap_if_false"): Add exit after trap.
7523 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7525 PR rtl-optimization/85638
7526 * bb-reorder.c: Include common/common-target.h.
7527 (create_forwarder_block): New function extracted from...
7528 (fix_up_crossing_landing_pad): ...here. Rename into...
7529 (dw2_fix_up_crossing_landing_pad): ...this.
7530 (sjlj_fix_up_crossing_landing_pad): New function.
7531 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7532 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7533 from both partitions and exit the loop after one iteration.
7535 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7538 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7540 (Basic PowerPC Built-in Functions): The new name of the
7541 subsection previously known as "PowerPC Built-in Functions".
7542 (Basic PowerPC Built-in Functions Available on all Configurations):
7544 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7546 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7548 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7550 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7553 2018-05-08 Jim Wilson <jimw@sifive.com>
7555 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7556 (LD_EMUL_SUFFIX): New.
7557 (LINK_SPEC): Use it.
7559 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7561 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7563 (Basic PowerPC Built-in Functions): The new name of the
7564 subsection previously known as "PowerPC Built-in Functions".
7565 (Basic PowerPC Built-in Functions Available on all Configurations):
7567 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7569 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7571 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7573 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7576 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7579 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7580 after cmpelim optimization.
7582 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7584 * config.gcc: Support "goldmont".
7585 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7586 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7588 * config/i386/i386.c (m_GOLDMONT): Define.
7589 (processor_target_table): Add "goldmont".
7590 (PTA_GOLDMONT): Define.
7591 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7592 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7593 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7594 (fold_builtin_cpu): Add "goldmont".
7595 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7596 (ix86_option_override_internal): Add "goldmont".
7597 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7598 (processor_type): Add PROCESSOR_GOLDMONT.
7599 * config/i386/i386.md: Add CPU "glm".
7600 * config/i386/glm.md: New file.
7601 * config/i386/x86-tune.def: Add m_GOLDMONT.
7602 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7604 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7607 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7609 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7610 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7611 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7614 * config/i386/i386.c (ix86_fold_builtin): Handle
7615 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7618 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7619 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7621 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7624 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7625 (check_arch): Likewise.
7626 (check_fpu): Return the result rather than printing it.
7627 (end arch): Fix operator precedence.
7628 (end cpu): Likewise.
7629 (END): Print the result from check_fpu.
7631 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7632 Alan Hayward <alan.hayward@arm.com>
7633 David Sherwood <david.sherwood@arm.com>
7635 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7636 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7637 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7638 (*fcmuo<mode>_and): New patterns.
7640 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7642 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7643 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7644 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7645 (cmp_op, sve_imm_con): New code attributes.
7646 (SVE_COND_INT_CMP, imm_con): Delete.
7647 (cmp_op): Remove above unspecs from int attribute.
7648 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7650 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7651 comparison-specific unspecs.
7652 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7653 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7654 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7655 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7656 (*vec_fcm<cmp_op><mode>): Rename to...
7657 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7658 (*vec_fcmuo<mode>): Rename to...
7659 (*fcmuo<mode>): ...this and adjust likewise.
7660 (*pred_fcm<cmp_op><mode>): New pattern.
7661 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7662 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7664 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7666 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7667 (aarch64_emit_sve_predicated_cond): New function.
7668 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7669 (aarch64_emit_unspec_cond_or): Replace with...
7670 (aarch64_emit_sve_or_conds): ...this new function. Use
7671 aarch64_emit_sve_ptrue_op for the individual comparisons and
7672 aarch64_emit_binop to OR them together.
7673 (aarch64_emit_inverted_unspec_cond): Replace with...
7674 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7675 aarch64_emit_sve_ptrue_op for the comparison and
7676 aarch64_emit_unop to invert the result.
7677 (aarch64_expand_sve_vec_cmp_float): Update after the above
7678 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7680 2018-05-07 Nathan Sidwell <nathan@acm.org>
7682 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7683 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7684 (Backwards Compatibility): Likewise.
7686 2018-05-07 Luis Machado <luis.machado@linaro.org>
7690 2018-05-07 Luis Machado <luis.machado@linaro.org>
7692 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7693 <prefetch_dynamic_strides>: New const bool field.
7694 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7695 prefetch_dynamic_strides.
7696 (exynosm1_prefetch_tune): Likewise.
7697 (thunderxt88_prefetch_tune): Likewise.
7698 (thunderx_prefetch_tune): Likewise.
7699 (thunderx2t99_prefetch_tune): Likewise.
7700 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7702 (aarch64_override_options_internal): Update to set
7703 PARAM_PREFETCH_DYNAMIC_STRIDES.
7704 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7705 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7706 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7707 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7708 prefetch-dynamic-strides setting.
7710 2018-05-07 Luis Machado <luis.machado@linaro.org>
7712 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7713 <minimum_stride>: New const int field.
7714 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7715 minimum_stride field.
7716 (exynosm1_prefetch_tune): Likewise.
7717 (thunderxt88_prefetch_tune): Likewise.
7718 (thunderx_prefetch_tune): Likewise.
7719 (thunderx2t99_prefetch_tune): Likewise.
7720 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7721 (aarch64_override_options_internal): Update to set
7722 PARAM_PREFETCH_MINIMUM_STRIDE.
7723 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7724 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7725 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7726 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7727 stride is constant and is below the minimum stride threshold.
7729 2018-05-07 Luis Machado <luis.machado@linaro.org>
7731 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7734 2018-05-07 Luis Machado <luis.machado@linaro.org>
7736 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7737 <prefetch_dynamic_strides>: New const bool field.
7738 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7739 prefetch_dynamic_strides.
7740 (exynosm1_prefetch_tune): Likewise.
7741 (thunderxt88_prefetch_tune): Likewise.
7742 (thunderx_prefetch_tune): Likewise.
7743 (thunderx2t99_prefetch_tune): Likewise.
7744 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7746 (aarch64_override_options_internal): Update to set
7747 PARAM_PREFETCH_DYNAMIC_STRIDES.
7748 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7749 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7750 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7751 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7752 prefetch-dynamic-strides setting.
7754 2018-05-07 Luis Machado <luis.machado@linaro.org>
7756 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7757 <minimum_stride>: New const int field.
7758 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7759 minimum_stride field.
7760 (exynosm1_prefetch_tune): Likewise.
7761 (thunderxt88_prefetch_tune): Likewise.
7762 (thunderx_prefetch_tune): Likewise.
7763 (thunderx2t99_prefetch_tune): Likewise.
7764 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7765 (aarch64_override_options_internal): Update to set
7766 PARAM_PREFETCH_MINIMUM_STRIDE.
7767 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7768 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7769 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7770 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7771 stride is constant and is below the minimum stride threshold.
7773 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7776 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7777 the type is addressable. Don't force op into register if it has
7780 2018-05-05 Roland McGrath <mcgrathr@google.com>
7783 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7784 any section for which we don't know a specific type it should have,
7785 regardless of name. Previously this was done only for the exact
7786 names ".init_array", ".fini_array", and ".preinit_array".
7787 (default_elf_asm_named_section): Add comment about
7788 relationship with default_section_type_flags and SECTION_NOTYPE.
7789 (get_section): Don't consider it a type conflict if one side has
7790 SECTION_NOTYPE and the other doesn't, as long as neither has the
7791 SECTION_BSS et al used in the default_section_type_flags logic.
7793 2018-05-05 Tom de Vries <tom@codesourcery.com>
7796 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7797 (workaround_barsyncs): New function.
7798 (nvptx_reorg): Use workaround_barsyncs.
7799 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7800 (define_expand "nvptx_membar_cta"): New define_expand.
7801 (define_insn "*nvptx_membar_cta"): New insn.
7803 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7805 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7806 To improve optimization opportunities.
7807 * builtin-types.def: The new needed builtin types for the above.
7809 2018-05-04 Richard Biener <rguenther@suse.de>
7811 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7812 * gimple-ssa-store-merging.c
7813 (imm_store_chain_info::output_merged_store): Remove redundant create,
7814 release split_store vector contents on failure.
7815 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7816 scalar stmt vector on cache hit.
7818 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7820 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7822 * config.gcc (powerpc-xilinx-eabi*): Remove.
7823 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7825 (fusion_addis_mem_combo_load): Ditto.
7826 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7828 (rs6000_cpu_cpp_builtins): Ditto.
7829 * config/rs6000/rs6000-linux.c
7830 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7831 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7832 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7834 (rs6000_setup_reg_addr_masks): Ditto.
7835 (rs6000_init_hard_regno_mode_ok): Ditto.
7836 (rs6000_option_override_internal): Ditto.
7837 (legitimate_lo_sum_address_p): Ditto.
7838 (rs6000_legitimize_address): Ditto.
7839 (rs6000_legitimize_reload_address): Ditto.
7840 (rs6000_legitimate_address_p): Ditto.
7841 (abi_v4_pass_in_fpr): Ditto.
7842 (setup_incoming_varargs): Ditto.
7843 (rs6000_gimplify_va_arg): Ditto.
7844 (rs6000_split_multireg_move): Ditto.
7845 (rs6000_savres_strategy): Ditto.
7846 (rs6000_emit_prologue_components): Ditto.
7847 (rs6000_emit_epilogue_components): Ditto.
7848 (rs6000_emit_prologue): Ditto.
7849 (rs6000_emit_epilogue): Ditto.
7850 (rs6000_elf_file_end): Ditto.
7851 (rs6000_function_value): Ditto.
7852 (rs6000_libcall_value): Ditto.
7853 * config/rs6000/rs6000.h: Ditto.
7854 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7855 (TARGET_MINMAX): ... this. New.
7856 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7857 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7858 (*movsi_internal1_single): Delete.
7859 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7860 mfpu=, mxilinx-fpu): Delete.
7861 * config/rs6000/singlefp.h: Delete.
7862 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7863 * config/rs6000/t-rs6000: Ditto.
7864 * config/rs6000/t-xilinx: Delete.
7865 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7866 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7867 (VStype_simple): Delete.
7868 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7869 * config/rs6000/xfpu.h: Delete.
7870 * config/rs6000/xfpu.md: Delete.
7871 * config/rs6000/xilinx.h: Delete.
7872 * config/rs6000/xilinx.opt: Delete.
7873 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7874 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7876 2018-05-04 Tom de Vries <tom@codesourcery.com>
7879 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7882 2018-05-04 Richard Biener <rguenther@suse.de>
7885 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7886 (expand_complex_div_wide): Likewise.
7887 (expand_complex_operations_1): Likewise.
7888 (expand_complex_libcall): Preserve EH info of the original stmt.
7889 (tree_lower_complex): Handle removed blocks.
7890 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7891 on complex multiplication and division libcall builtins.
7893 2018-05-04 Richard Biener <rguenther@suse.de>
7896 * fold-const.c (negate_expr_p): Restrict negation of operand
7897 zero of a division to when we know that can happen without
7899 (fold_negate_expr_1): Likewise.
7901 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7904 * real.h (real_nextafter): Declare.
7905 * real.c (real_nextafter): New function.
7906 * fold-const-call.c (fold_const_nextafter): New function.
7907 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7908 CASE_CFN_NEXTTOWARD.
7909 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7910 even when arg1_mode is different from arg0_mode.
7912 2018-05-03 Nathan Sidwell <nathan@acm.org>
7914 * doc/extend.texi (Deprecated Features): Remove
7916 (Backwards Compatibility): Likewise.
7917 * doc/invoke.texi (C++ Language Options): Likewise.
7918 (C++ Dialect Options): Likewise.
7920 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7923 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7924 _mm512_mask_mullox_epi64): New intrinsics.
7926 2018-05-03 Tom de Vries <tom@codesourcery.com>
7929 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7930 dump files): Add offload-tree.
7932 2018-05-03 Richard Biener <rguenther@suse.de>
7934 PR tree-optimization/85615
7935 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7936 to loops not nested in BBs loop father to avoid creating multi-entry
7939 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7941 PR tree-optimization/70291
7942 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7943 arguments. Change return type to tree. Emit libcall as a new
7944 statement rather than replacing existing one when inplace_p is true.
7945 (expand_complex_multiplication_components): New function.
7946 (expand_complex_multiplication): Expand floating-point complex
7947 multiplication using the above.
7948 (expand_complex_division): Rename inner_type parameter to type.
7949 Update expand_complex_libcall call-site.
7950 (expand_complex_operations_1): Update expand_complex_multiplication
7951 and expand_complex_division call-sites.
7953 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7956 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7957 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7958 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7959 the highest significant bit of the shift count mask is clear. In
7960 check whether and[sq]i3 is needed verify that all significant bits
7961 of the shift count other than the highest are set.
7963 2018-05-02 Tom de Vries <tom@codesourcery.com>
7966 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7967 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7968 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7969 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7970 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7971 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7972 __builtin_goacc_parlevel_size.
7974 2018-05-02 Richard Biener <rguenther@suse.de>
7976 PR tree-optimization/85597
7977 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7978 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7981 2018-05-02 Tom de Vries <tom@codesourcery.com>
7984 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7985 dump files): Add ltrans-tree.
7987 2018-05-02 Tom de Vries <tom@codesourcery.com>
7990 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7991 dump files): Add wpa-ipa.
7993 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7995 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7996 powerpc*-*-linux*paired* target.
7997 * config/rs6000/750cl.h: Delete.
7998 * config/rs6000/paired.h: Delete.
7999 * config/rs6000/paired.md: Delete.
8000 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8002 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8003 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8004 comment. Remove paired float support.
8005 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8006 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8008 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8009 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8011 * config/rs6000/rs6000.c: Remove paired float support.
8012 (paired_expand_vector_init, paired_expand_vector_move,
8013 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8014 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8015 paired_expand_builtin, paired_expand_predicate_builtin,
8016 paired_init_builtins): Delete.
8017 * config/rs6000/rs6000.h: Remove paired float support.
8018 * config/rs6000/rs6000.md: Remove paired float support.
8019 (move_from_CR_ov_bit): Delete.
8020 * config/rs6000/rs6000.opt (mpaired): Delete.
8021 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8022 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8024 2018-05-02 Richard Biener <rguenther@suse.de>
8027 * gimplify.c (gimplify_save_expr): When in SSA form allow
8028 SAVE_EXPRs to compute to SSA vars.
8030 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8033 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8034 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8035 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8036 clobber operands[2], instead use a new pseudo. Formatting fixes.
8038 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8040 PR tree-optimization/85586
8041 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8042 exit early for statements in the same group if the accesses are
8045 2018-05-02 Tom de Vries <tom@codesourcery.com>
8048 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8051 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8053 PR tree-optimization/85143
8054 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8056 2018-05-01 Tom de Vries <tom@codesourcery.com>
8059 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8060 not found" error message.
8062 2018-05-01 Tom de Vries <tom@codesourcery.com>
8065 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8066 * vec.c (test_ordered_remove_if): New function.
8067 (vec_c_tests): Call test_ordered_remove_if.
8068 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8069 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8070 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8071 VEC_ORDERED_REMOVE_IF.
8073 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8075 PR tree-optimization/82665
8076 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8077 pointer subtraction where arguments come from a memchr call.
8079 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8081 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8082 --push-state --as-needed and --pop-state instead of --as-needed and
8083 --no-as-needed if ld supports it.
8084 * configure: Regenerated.
8087 * doc/install.texi2html: Replace _002d with - and _002a with * in
8088 generated html files using sed.
8090 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8093 * gcc-rich-location.c (blank_line_before_p): New function.
8094 (use_new_line): New function.
8095 (gcc_rich_location::add_fixit_insert_formatted): New function.
8096 * gcc-rich-location.h
8097 (gcc_rich_location::add_fixit_insert_formatted): New function.
8099 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8101 * selftest.c (assert_streq): Rename "expected" and "actual" to
8102 "val1" and "val2". Extend NULL-handling to cover both inputs
8103 symmetrically, while still requiring both to be non-NULL for a pass.
8104 * selftest.h (assert_streq): Rename "expected" and "actual" to
8106 (ASSERT_EQ): Likewise.
8107 (ASSERT_EQ_AT): Likewise.
8108 (ASSERT_KNOWN_EQ): Likewise.
8109 (ASSERT_KNOWN_EQ_AT): Likewise.
8110 (ASSERT_NE): Likewise.
8111 (ASSERT_MAYBE_NE): Likewise.
8112 (ASSERT_MAYBE_NE_AT): Likewise.
8113 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8114 the assertion to pass.
8115 (ASSERT_STREQ_AT): Likewise.
8117 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8119 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8120 interaction with -pie.
8122 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8124 * selftest.h: Fix alphabetization of per-source-file selftest
8127 2018-04-30 Jason Merrill <jason@redhat.com>
8129 PR c++/61982 - dead stores to destroyed objects.
8130 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8133 2018-04-30 Jason Merrill <jason@redhat.com>
8135 * tree.c (build_clobber): New.
8136 * tree.h: Declare it.
8137 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8139 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8141 * diagnostic-show-locus.c (layout::layout): Update for
8142 location_get_source_line returning a char_span.
8143 (struct char_span): Move to input.h.
8144 (struct correction): Update for fields in char_span becoming
8146 (struct source_line): Update for location_get_source_line
8147 returning a char_span.
8148 (layout::print_line): Likewise.
8149 * edit-context.c (edited_file::print_content): Likewise.
8150 (edited_file::print_diff_hunk): Likewise.
8151 (edited_file::print_run_of_changed_lines): Likewise.
8152 (edited_file::get_num_lines): Likewise.
8153 (edited_line::edited_line): Likewise.
8154 * final.c (asm_show_source): Likewise.
8155 * input.c (location_get_source_line): Convert return type
8156 from const char * to char_span, losing the final "line_len"
8158 (dump_location_info): Update for the above.
8159 (get_substring_ranges_for_loc): Likewise. Use a char_span
8160 when handling the literal within the line.
8161 (test_reading_source_line): Update for location_get_source_line
8162 returning a char_span.
8163 * input.h (class char_span): Move here from
8164 diagnostic-show-locus.c, converting from a struct to a class.
8165 Make data members private.
8166 (char_span::operator bool): New.
8167 (char_span::length): New.
8168 (char_span::get_buffer): New.
8169 (char_span::operator[]): New.
8170 (char_span::subspan): Make const.
8171 (char_span::xstrdup): New.
8172 (location_get_source_line): Convert return type from const char *
8173 to char_span, losing the final "line_size" param.
8175 2018-04-30 Jan Hubicka <jh@suse.cz>
8177 * lto-wrapper.c (ltrans_priorities): New static var.
8178 (cmp_priority): New.
8179 (run_gcc): Read priorities and if doing parallel build order
8180 the Makefile by them.
8182 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8184 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8186 2018-04-30 Richard Biener <rguenther@suse.de>
8188 * tree-cfg.c (verify_address): Remove base argument, add
8189 flag whether to check TREE_ADDRESSABLE and do that.
8190 (verify_expr): Remove.
8191 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8192 (verify_gimple_assign_single): Likewise.
8193 (verify_gimple_switch): Likewise.
8194 (verify_expr_location_1): Dereference tp once. Add (disabled)
8195 piece from verify_expr.
8196 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8198 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8200 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8202 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8204 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8205 (small_data_pattern): Likewise.
8206 (arc_rewrite_small_data): Likewise.
8207 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8208 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8209 (get_symbol_alignment): New function.
8210 (legitimate_small_data_address_p): Likewise.
8211 (legitimate_scaled_address): Update, call
8212 legitimate_small_data_address_p.
8213 (output_sdata): New static variable.
8214 (arc_print_operand): Update how we handle small data operands.
8215 (arc_print_operand_address): Likewise.
8216 (arc_legitimate_address_p): Update, use
8217 legitimate_small_data_address_p.
8218 (arc_rewrite_small_data_p): Remove.
8219 (arc_rewrite_small_data_1): Likewise.
8220 (arc_rewrite_small_data): Likewise.
8221 (small_data_pattern): Likewise.
8222 (compact_sda_memory_operand): Update to use
8223 legitimate_small_data_address_p and get_symbol_alignment.
8224 (prepare_move_operands): Don't rewite sdata pattern.
8225 (prepare_extend_operands): Remove.
8226 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8228 (zero_extendqisi2): Likewise.
8229 (zero_extendhisi2): Likewise.
8230 (extendqihi2): Likewise.
8231 (extendqisi2): Likewise.
8232 (extendhisi2): Likewise.
8236 * config/arc/constraints.md (Usd): Change it to memory constraint.
8238 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8240 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8241 as source of std instructions.
8242 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8243 allow 6-bit constants as source for store instructions.
8244 (movdi_insn): Update instruction pattern to allow 6-bit constants
8245 as source for store instructions.
8247 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8249 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8251 2018-04-30 Nathan Sidwell <nathan@acm.org>
8252 Sandra Loosemore <sandra@codesourcery.com>
8254 * dumpfile.c (dump_open): Allow '-' for stdout.
8255 * doc/invoke.texi (Developer Options): Document dump filename
8256 determination early. Document stdin/stdout selection.
8258 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8260 Microblaze Target: PIC data text relative
8262 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8263 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8265 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8266 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8267 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8268 New addressing mode for data-text relative position indepenedent code.
8269 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8270 'ADDRESS_SYMBOLIC_TXT_REL'.
8271 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8272 (microblaze_legitimate_pic_operand): Exclude function calls from
8273 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8274 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8276 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8277 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8278 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8279 for 'address + offset'.
8280 (microblaze_expand_prologue): Add new function prologue call for
8282 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8283 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8284 table in case of TARGET_PIC_DATA_TEXT_REL.
8285 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8286 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8287 Add new macros 'UNSPEC_TEXT',
8288 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8289 + exclude function calls from 'UNSPEC_PLT' in case of data text
8291 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8292 new target hook for generating address diff vector tables in case of
8294 * doc/tm.texi : Regenerate.
8295 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8296 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8297 of addr diff vector generation.
8298 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8299 target hook definition.
8300 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8301 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8302 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8303 option for data text relative.
8305 2018-04-30 Richard Biener <rguenther@suse.de>
8307 * tree-chrec.h (evolution_function_is_constant_p): Remove
8309 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8311 2018-04-30 Richard Biener <rguenther@suse.de>
8314 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8316 2018-04-30 Richard Biener <rguenther@suse.de>
8318 PR tree-optimization/28364
8319 PR tree-optimization/85275
8320 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8321 copying first exit test.
8323 2018-04-28 Mark Wielaard <mark@klomp.org>
8325 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8327 (dwarf_AT): Handle DW_AT_addr_base.
8328 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8330 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8333 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8334 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8335 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8336 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8338 2018-04-28 Richard Biener <rguenther@suse.de>
8340 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8341 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8342 to reflect use. Only add interesting stmts.
8344 2018-04-27 Martin Jambor <mjambor@suse.cz>
8347 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8348 the jump function allows for passing through aggregate values.
8350 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8352 * input.h (in_system_header_at): Convert from macro to inline
8354 (from_macro_expansion_at): Likewise.
8355 (from_macro_definition_at): Likewise.
8357 2018-04-27 Jeff Law <law@redhat.com>
8359 * config.gcc: Mark tile* targets as deprecated/obsolete.
8361 2018-04-27 Richard Biener <rguenther@suse.de>
8363 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8366 2018-04-27 Richard Biener <rguenther@suse.de>
8368 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8370 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8372 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8373 with Yd constraint. Set "preferred_for_speed" attribute from
8374 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8376 (*movdi_internal): Ditto.
8377 (movti_interunit splitters): Remove
8378 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8379 (movdi_interunit splitters): Ditto.
8380 * config/i386/constraints.md (Ye): Remove.
8381 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8383 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8386 * config/aarch64/constraints.md (Usg): Limit to 31.
8389 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8391 PR tree-optimization/85529
8392 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8393 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8394 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8395 zero extension or masking of the MSB bit.
8396 (optimize_range_tests): Add FIRST_BB argument, pass it through
8397 to optimize_range_tests_var_bound.
8398 (maybe_optimize_range_tests, reassociate_bb): Adjust
8399 optimize_range_tests callers.
8401 2018-04-26 Richard Biener <rguenther@suse.de>
8402 Jakub Jelinek <jakub@redhat.com>
8404 * cgraph.h (symbol_table): Just declare debug method here.
8405 * symtab.c (symbol_table::debug): Define.
8407 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8409 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8411 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8413 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8414 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8415 (*movdi_internal): Substitute Yi and Yj constraint with x
8416 and Ym and Yn constraint with y constraint. Update "isa"
8417 attribute and set "preferred_for_speed" attribute from
8418 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8419 (*movsi_internal): Ditto.
8420 (*movdf_internal): Ditto.
8421 (*movsf_internal): Ditto.
8422 (*zero_extendsidi2): Ditto.
8423 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8424 (sse2_loadld): Ditto.
8425 (*vec_extract<ssevecmodelower>_0): Ditto.
8426 (*vec_extractv4si_0_zext_sse4): Ditto.
8427 (vec_concatv2di): Ditto.
8428 (*vec_dup<mode>): Ditto.
8429 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8430 * config/i386/constraints.md (Yi): Remove.
8435 2018-04-26 Nathan Sidwell <nathan@acm.org>
8437 * dumpfile.c (dump_open): New.
8438 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8439 (dump_finish): Detect stdio/stderr by value not name.
8441 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8443 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8445 2018-04-26 Tom de Vries <tom@codesourcery.com>
8448 * config/nvptx/nvptx.c (verify_neutering_jumps)
8449 (verify_neutering_labels): New function
8450 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8452 2018-04-26 Tom de Vries <tom@codesourcery.com>
8455 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8456 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8459 2018-04-26 Richard Biener <rguenther@suse.de>
8460 Tom de Vries <tom@codesourcery.com>
8463 * lto-streamer-out.c (output_function): Fixup loops if required to match
8464 discovery done in the reader.
8466 2018-04-26 Richard Biener <rguenther@suse.de>
8468 PR tree-optimization/85116
8469 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8470 have a loop exit from the single latch predecessor. Remove
8471 case of header with just condition.
8472 (ch_base::copy_headers): Exclude infinite loops from any
8474 (pass_ch::execute): Record exits.
8476 2018-04-26 Richard Biener <rguenther@suse.de>
8478 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8479 prologue cost vector and pass it to vect_get_load_cost.
8480 (vect_get_peeling_costs_all_drs): Likewise.
8481 (vect_peeling_hash_get_lowest_cost): Likewise.
8482 (vect_enhance_data_refs_alignment): Likewise.
8484 2018-04-26 Richard Biener <rguenther@suse.de>
8487 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8488 checking of integer<->pointer conversions.
8489 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8490 sign-/zero-extending pointer types.
8491 (expand_omp_for_static_chunk): Likewise.
8493 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8494 Jean Lee <xiaoyur347@gmail.com>
8496 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8497 (TARGET_ASAN_SHADOW_OFFSET): Define.
8498 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8499 true for -fsanitize=address.
8501 2018-04-25 Mark Wielaard <mark@klomp.org>
8503 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8506 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8508 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8509 than "alu", remove explicit "memory" and "imm_disp" attributes.
8510 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8513 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8514 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8515 gen_lowpart_no_emit.
8517 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8520 * config/i386/i386.c (ix86_expand_builtin): Change memory
8521 operand to XI, extend p0 to Pmode.
8522 * config/i386/i386.md: Change unspec volatile and operand
8523 1 mode to XI, change operand 0 mode to P.
8525 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8527 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8528 GET_MODE_MASK before any checking.
8529 (nds32_can_use_bset_p): Likewise.
8530 (nds32_can_use_btgl_p): Likewise.
8532 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8534 * config/nds32/nds32-doubleword.md: New define_split pattern for
8535 illegal register number.
8537 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8539 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8541 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8543 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8545 2018-04-25 Richard Biener <rguenther@suse.de>
8547 * lto-streamer.h (LTO_major_version): Bump to 8.
8549 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8551 * BASE-VER: Set to 9.0.0.
8553 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8555 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8556 in __abskf2 and __powikf2.
8558 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8561 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8562 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8563 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8564 Use the above on operand 2. Reindent.
8565 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8567 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8570 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8572 * config/i386/i386.opt (mcet): Removed.
8573 * doc/install.texi: Remove -mcet documentation.
8574 * doc/invoke.texi: Likewise.
8576 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8579 * doc/install.texi: Remove -mcet from bootstrap-cet.
8581 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8584 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8585 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8589 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8590 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8591 containing a CONST_VECTOR.
8593 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8595 * doc/install.texi: Update newlib dependency for nvptx.
8597 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8600 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8601 instead of INTVAL when shifting x left.
8603 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8605 PR tree-optimization/85478
8606 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8607 vect_grouped_store_supported for single element vectors.
8609 2018-04-24 Richard Biener <rguenther@suse.de>
8612 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8613 load cost increase to the case of non-constant step.
8615 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8618 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8619 destination if any_malformed_asm.
8621 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8624 * expr.c (store_field): In the bitfield case, if the value comes from
8625 a function call and is returned in registers by means of a PARALLEL,
8626 do not change the mode of the temporary unless BLKmode and VOIDmode.
8628 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8630 PR rtl-optimization/85423
8631 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8632 dependencies to debug insns when the previous insn is non-debug.
8634 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8636 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8637 enums into a single definition.
8638 (fls): Fix predicates and printing.
8641 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8643 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8644 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8645 and short u6 immediate.
8646 (check_if_valid_sleep_operand): Remove.
8647 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8649 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8651 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8652 flag_always_save_lp condition.
8653 * config/nds32/nds32.opt (malways-save-lp): New option.
8655 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8657 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8658 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8659 * config/nds32/nds32.h
8660 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8661 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8663 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8665 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8666 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8668 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8669 Chung-Ju Wu <jasonwucj@gmail.com>
8671 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8673 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8674 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8676 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8678 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8680 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8682 * config/nds32/nds32-protos.h (nds32_data_alignment,
8683 nds32_local_alignment): Declare.
8684 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8685 nds32_local_alignment): New functions.
8686 (TARGET_CONSTANT_ALIGNMENT): Define.
8687 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8689 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8691 * config/nds32/nds32.c
8692 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8693 (TARGET_MODES_TIEABLE_P): Likewise.
8695 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8697 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8700 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8701 Chung-Ju Wu <jasonwucj@gmail.com>
8703 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8704 for unaligned access.
8705 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8707 * config/nds32/nds32-intrinsic.md: Likewise.
8708 * config/nds32/nds32_intrinsic.h: Likewise.
8709 * config/nds32/nds32.h (nds32_builtins): Likewise.
8710 * config/nds32/nds32.opt (munaligned-access): New option.
8711 * config/nds32/nds32.c (nds32_asm_file_start): Display
8712 flag_unaligned_access status.
8714 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8716 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8717 -mno-relax is present.
8718 * config/riscv/linux.h (LINK_SPEC): Ditto.
8720 2018-04-20 Martin Sebor <msebor@redhat.com>
8723 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8725 (gimple_fold_builtin_stxcpy_chk): Same.
8726 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8728 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8731 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8732 __powikf2 when long double is IEEE 128-bit.
8734 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8736 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8737 step to make sure stack always aligned.
8739 2018-04-20 Carl Love <cel@us.ibm.com>
8742 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8743 size check for arg0.
8745 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8746 Tom de Vries <tom@codesourcery.com>
8749 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8750 Emit insns for calls too.
8751 (nvptx_find_par): Always look for worker-level predecessor insn.
8752 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8754 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8755 (nvptx_process_pars): Propagate frames for calls.
8757 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8760 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8762 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8763 (ix86_handle_option): Don't handle OPT_mibt.
8764 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8766 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8768 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8769 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8770 (ix86_target_macros): Define __CET__ with flag_cf_protection
8771 for -fcf-protection.
8772 * config/i386/i386.c (isa2_opts): Remove -mibt.
8773 * config/i386/i386.h (TARGET_IBT): Removed.
8774 (TARGET_IBT_P): Likewise.
8775 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8776 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8777 * config/i386/i386.opt (mcet): Update help message.
8780 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8781 -mcet as an alias for -mshstk.
8783 2018-04-20 Richard Biener <rguenther@suse.de>
8786 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8787 complexity by forcing a single use of the multiply operand.
8789 2018-04-20 Martin Jambor <mjambor@suse.cz>
8792 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8793 recursion dependency to only apply to non-clones.
8795 2018-04-20 Martin Jambor <mjambor@suse.cz>
8798 * ipa-cp.c (create_specialized_node): Check that clones of
8799 self-recursive edges exist during IPA-CP.
8801 2018-04-19 Toon Moene <toon@moene.org>
8803 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8806 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8808 PR tree-optimization/85467
8809 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8810 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8811 VECTOR_CST element to type.
8813 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8816 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8817 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8818 (builtin_longjmp): Likewise.
8819 (save_stack_nonlocal): New pattern.
8820 (restore_stack_nonlocal): Likewise.
8822 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8825 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8826 Replace ASM_OUTPUT_LABEL with fprintf.
8828 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8831 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8832 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8833 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8834 define __IBT__ and __SHSTK__ for -fcf-protection.
8835 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8837 (ix86_trampoline_init): Likewise.
8838 (x86_output_mi_thunk): Likewise.
8839 (ix86_notrack_prefixed_insn_p): Likewise.
8840 (ix86_option_override_internal): Don't disallow -fcf-protection.
8841 * config/i386/i386.md (rdssp<mode>): Also enable for
8843 (incssp<mode>): Likewise.
8844 (nop_endbr): Likewise.
8845 * config/i386/i386.opt (mcet): Change help message to built-in
8849 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8850 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8851 enable CET built-in functions.
8853 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8855 * common/config/i386/i386-common.c
8856 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8857 OPTION_MASK_ISA_MOVDIRI_UNSET,
8858 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8859 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8860 * config.gcc (movdirintrin.h): New header.
8861 * config/i386/cpuid.h (bit_MOVDIRI,
8862 bit_MOVDIR64B): New bits.
8863 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8865 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8866 (VOID, PVOID, PCVOID)): New function types.
8867 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8868 __builtin_ia32_directstoreu_u64,
8869 __builtin_ia32_movdir64b): New builtins.
8870 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8871 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8873 (ix86_valid_target_attribute_inner_p): Ditto.
8874 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8875 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8876 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8877 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8878 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8879 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8880 (movdiri<mode>, movdir64b_<mode>): New.
8881 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8882 * config/i386/immintrin.h: Include movdirintrin.h.
8883 * config/i386/movdirintrin.h: New file.
8884 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8886 2018-04-19 Richard Biener <rguenther@suse.de>
8889 * cfg.c (clear_bb_flags): When loop state says we have
8890 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8892 2018-04-19 Richard Biener <rguenther@suse.de>
8894 PR tree-optimization/84737
8895 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8896 copying restrict info.
8897 (vect_setup_realignment): Use it.
8898 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8899 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8900 the first DR to all generated stores.
8901 (vectorizable_load): Likewise for loads.
8903 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8905 PR tree-optimization/85446
8906 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8907 the integral and pointer types to have the same precision.
8909 * doc/install.texi: Document --disable-cet being the default and
8912 2018-04-18 Martin Liska <mliska@suse.cz>
8914 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8917 2018-04-18 Martin Liska <mliska@suse.cz>
8920 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8923 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8924 arguments if they are comparable.
8926 2018-04-18 Martin Liska <mliska@suse.cz>
8929 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8932 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8935 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8938 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8939 ENDBR after calling __morestack.
8941 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8944 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8945 by using gcc_base_ver to generate a gcc_driver_version, and use
8946 it when generating GCC_DRIVER_NAME.
8947 * configure: Regenerate.
8949 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8952 * config.gcc: Obsolete powerpc*-*-*spe*.
8954 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8957 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8958 (stabstr_D): Change type of unum from unsigned int to
8959 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8962 2018-04-17 Jim Wilson <jimw@sifive.com>
8965 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8966 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8967 Set arg_pointer_offset after using pretend_args_size.
8969 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8971 PR rtl-optimization/85431
8972 * dse.c (record_store): Ignore zero width stores.
8975 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8976 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8977 __builtin_stack_restore rather than after it.
8978 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8979 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8980 argument instead of virtual_dynamic_stack_rtx.
8982 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8984 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8986 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8987 Add note to error message to explain internal mapping of overloaded
8988 built-in function name to non-overloaded built-in function name.
8989 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8992 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8995 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8996 where the inputs overlap with the output.
8998 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9000 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9001 (=v, v) alternative and explicit "memory" attribute.
9002 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9003 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9005 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9006 "sselog1" type instead of "sselog".
9007 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9008 "sselog". Remove explicit "memory" attribute.
9009 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9010 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9012 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9013 "sselog1" type instead of "sselog". Remove explicit "memory"
9015 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9016 use "sselog1" type instead of "sselog". Remove explicit "memory"
9018 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9019 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9021 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9022 "sselog1" type instead of "sselog". Remove explicit "memory"
9024 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9025 use "sselog1" type instead of "sselog". Remove explicit "memory"
9029 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9032 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9035 2018-04-17 Martin Jambor <mjambor@suse.cz>
9038 * ipa-cp.c (create_specialized_node): Call
9039 expand_all_artificial_thunks if necessary.
9041 2018-04-17 Martin Liska <mliska@suse.cz>
9044 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9045 in message, remote space in between '_G' and '('.
9047 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9050 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9051 avx512f_vmcmp<mode>3<round_saeonly_name>,
9052 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9053 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9054 avx512f_rndscale<mode><round_saeonly_name>,
9055 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9056 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9057 Use %<iptr>2 instead of %2 for -masm=intel.
9058 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9059 avx512f_vcvttss2usi<round_saeonly_name>,
9060 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9062 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9063 avx512f_vcvttsd2usi<round_saeonly_name>,
9064 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9065 Use %q1 instead of %1 for -masm=intel.
9066 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9067 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9068 of %3 for -masm=intel.
9069 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9071 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9073 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9075 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9076 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9078 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9080 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9081 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9082 %g1 and one with %0 and %1.
9083 (avx512er_vmrcp28<mode><round_saeonly_name>,
9084 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9086 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9087 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9088 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9089 of %0 and %{%4%} for -masm=intel.
9090 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9091 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9092 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9093 order of %0 and %{%5%}%{z%} for -masm=intel.
9095 2018-04-17 Jan Hubicka <jh@suse.cz>
9098 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9100 2018-04-17 Martin Liska <mliska@suse.cz>
9103 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9104 for target_clone error message. Make default implementation
9105 clone to be a local declaration.
9106 (separate_attrs): Add new argument and check for an empty
9108 (expand_target_clones): Handle it.
9109 (ipa_target_clone): Make redirection just for target_clones
9112 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9113 Tom de Vries <tom@codesourcery.com>
9116 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9117 tiled basic blocks without omp continue statements.
9119 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9122 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9123 vec_extract expression as having side effects to make sure it gets
9126 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9129 * config/i386/i386.c (get_builtin_code_for_version): Check
9132 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9135 * gcc/config.gcc: Support "skylake".
9136 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9138 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9139 (processor_target_table): Add "skylake".
9140 (ix86_option_override_internal): Add "skylake".
9141 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9142 PROCESSOR_CANNONLAKE.
9143 (get_builtin_code_for_version): Fix priority for
9144 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9145 PROCESSOR_SKYLAKE-AVX512.
9146 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9147 (processor_type): Add PROCESSOR_SKYLAKE.
9149 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9150 Jason Merrill <jason@redhat.com>
9153 * convert.c (convert_to_integer_1): Use direct recursion for
9154 enumeral types and types with a precision less than the number
9155 of bits in their mode.
9157 2018-04-16 Julia Koval <julia.koval@intel.com>
9160 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9161 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9163 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9166 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9167 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9168 and -mno-direct-move.
9170 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9173 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9174 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9175 Ensure negative shifts result in {0}.
9177 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9179 PR rtl-optimization/79916
9180 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9181 regs (if any) to define how to gnerate SD moves when LRA is in
9184 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9186 PR rtl-optimization/85393
9187 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9188 * except.c (expand_dw2_landing_pad_for_region): Make static.
9189 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9190 a label and unconditional jump to old_bb, rather than
9191 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9194 PR rtl-optimization/85376
9195 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9196 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9197 instead of a specific value.
9199 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9200 Bin Cheng <bin.cheng@arm.com>
9202 PR tree-optimization/82965
9203 PR tree-optimization/83991
9204 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9205 by_profile_only parameter.
9206 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9207 information if the loop was predicted to iterate too many times.
9208 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9210 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9213 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9216 2018-04-13 Martin Liska <mliska@suse.cz>
9217 Jakub Jelinek <jakub@redhat.com>
9220 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9221 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9222 * builtins.c (expand_builtin_memory_copy_args): Use
9223 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9224 handle dest_addr == pc_rtx.
9226 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9229 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9230 asked to not generate direct moves.
9231 (fix_trunc<mode>si2_stfiwx): Similar.
9232 (fix_trunc<mode>si2_internal): Similar.
9234 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9237 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9238 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9239 lookup if dest in some wider mode is known to be const0_rtx and
9240 if so, record permanent equivalence for it to be ZERO_EXTEND of
9241 the narrower mode destination.
9243 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9245 * lto-streamer-out.c (output_function): Revert 259346.
9246 * omp-expand.c (expand_oacc_for): Likewise.
9248 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9250 PR rtl-optimization/85354
9251 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9252 * sel-sched.c (sel_global_init): ... here.
9254 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9257 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9258 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9259 mode for PE-COFF targets.
9260 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9261 (i386_pe_asm_lto_end): Likewise.
9262 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9263 (TARGET_ASM_LTO_END): Likewise.
9264 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9265 (i386_pe_asm_lto_start): New function.
9266 (i386_pe_asm_lto_end): Likewise.
9268 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9269 Richard Biener <rguenther@suse.de>
9272 * lto-streamer-out.c (output_function): Fix CFG loop state before
9274 * omp-expand.c (expand_oacc_for): Handle calls to internal
9275 functions like regular functions.
9277 2018-04-12 Richard Biener <rguenther@suse.de>
9280 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9281 for the early LTO debug to properly generate references to it
9282 during DIE emission. Do not re-use that for the skeleton for
9284 (dwarf2out_early_finish): Likewise.
9286 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9289 * config/i386/sse.md
9290 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9291 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9292 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9293 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9294 and output is a reg, avoid creating invalid lowpart subreg, but
9295 instead split into a 512-bit move. Don't split if not AVX512VL,
9296 input is xmm16+ reg and output is a mem.
9297 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9298 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9299 xmm16+ reg and output is a mem.
9301 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9303 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9304 also for flag_dwarf2_cfi_asm.
9306 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9308 PR rtl-optimization/85342
9309 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9310 a bool scalar var inside of the loop instead. Don't try to update
9311 recog_data.operand after failed apply_change_group.
9313 2018-04-12 Tom de Vries <tom@codesourcery.com>
9316 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9317 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9318 array with flexible array member as array without given dimension.
9319 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9320 argument for undefined param to true.
9322 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9325 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9326 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9327 from PowerPC section.
9328 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9329 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9330 help text that is too long.
9331 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9332 help text that is too long.
9333 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9334 help text that is too long.
9336 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9338 * config/alpha/alpha.md (stack_probe_internal): Rename
9339 from "probe_stack". Update all callers.
9341 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9343 PR rtl-optimization/84566
9344 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9345 sched_macro_fuse_insns.
9347 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9350 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9351 (compute_block_dependences): ... from here.
9353 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9355 PR tree-optimization/85331
9356 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9357 from int to HOST_WIDE_INT.
9359 2018-04-11 Martin Jambor <mjambor@suse.cz>
9362 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9363 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9364 not the same as the source val.
9365 (cgraph_edge_brings_value_p): New parameter.
9366 (gather_edges_for_value): Pass destination value to
9367 cgraph_edge_brings_value_p.
9368 (perhaps_add_new_callers): Likewise.
9369 (get_info_about_necessary_edges): Likewise and exclude values brought
9370 only by self-recursive edges.
9371 (create_specialized_node): Redirect only clones of self-calling edges.
9372 (+self_recursive_pass_through_p): New function.
9373 (find_more_scalar_values_for_callers_subset): Use it.
9374 (find_aggregate_values_for_callers_subset): Likewise.
9375 (known_aggs_to_agg_replacement_list): Removed.
9376 (decide_whether_version_node): Re-calculate known constants for all
9377 remaining context clones.
9379 2018-04-11 Richard Biener <rguenther@suse.de>
9382 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9383 from early DWARF output.
9384 (dwarf2out_early_finish): Output line info unconditionally into
9385 early DWARF and add reference to it.
9387 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9390 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9391 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9392 other than V2DFmode using iptr mode attribute.
9393 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9395 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9397 PR rtl-optimization/84659
9398 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9400 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9403 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9405 (output_loc_list): Pass address of a dummy size variable even in the
9406 locview handling loop.
9407 (index_location_lists): Add comment on why skip_loc_list_entry can't
9410 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9413 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9416 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9419 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9420 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9421 and -mstring-compare-inline-limit.
9423 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9426 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9427 for stack clash protection in a register whenever we need it to be in
9430 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9432 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9433 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9435 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9438 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9440 (mlong-double-): Ditto.
9441 * config/rs6000/sysv4.opt (msdata=): Ditto.
9442 (mtls-size=): Ditto.
9444 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9446 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9447 erroneous entries for
9448 "vector int vec_ldl (int, long int *)", and
9449 "vector unsigned int vec_ldl (int, unsigned long int *)".
9450 Add comments and entries for
9451 "vector bool char vec_ldl (int, bool char *)",
9452 "vector bool short vec_ldl (int, bool short *)",
9453 "vector bool int vec_ldl (int, bool int *)",
9454 "vector bool long long vec_ldl (int, bool long long *)",
9455 "vector pixel vec_ldl (int, pixel *)",
9456 "vector long long vec_ldl (int, long long *)",
9457 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9458 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9459 type tree bool_long_long_type_node and correct definition of
9460 bool_V2DI_type_node to make reference to this new type tree.
9461 (rs6000_mangle_type): Replace erroneous reference to
9462 bool_long_type_node with bool_long_long_type_node.
9463 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9464 comments to emphasize sign distinctions for char and int types and
9465 replace RS6000_BTI_bool_long constant with
9466 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9467 use of RS6000_BTI_pixel.
9468 (bool_long_type_node): Remove this macro definition.
9469 (bool_long_long_type_node): New macro definition
9471 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9473 PR rtl-optimization/85300
9474 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9475 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9476 simplify_unary_operation fails.
9478 2018-04-10 Martin Liska <mliska@suse.cz>
9480 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9481 cgraph_edge and ipa_ref.
9483 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9487 * config/i386/sse.md
9488 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9489 computation of the VEC_MERGE selector from mask.
9490 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9491 Fix decoding of the VEC_MERGE selector into mask.
9493 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9495 PR tree-optimization/85286
9496 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9498 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9500 * final.c (final_1): Set insn_last_address as well as
9501 insn_current_address.
9503 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9506 * explow.c (emit_stack_probe): Call validize_mem on memory location
9507 before passing it to gen_probe_stack. Create address operand and
9508 legitimize it for the probe_stack_address case.
9510 2018-04-09 Jan Hubicka <jh@suse.cz>
9513 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9514 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9515 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9517 (free_lang_data): Rebuild type inheritance graph.
9519 2018-04-09 Martin Sebor <msebor@redhat.com>
9521 * invoke.texi (-finline-small-functions): Mention other optimization
9523 (-findirect-inlining, -fpartial-inlining): Same.
9524 (-finline-functions-called-once): Same.
9525 (-freorder-blocks-and-partition): Same.
9527 2018-04-09 Jan Hubicka <jh@suse.cz>
9530 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9531 jumps; choose last target that matches the criteria (i.e.
9532 no partition changes for non-crossing jumps).
9533 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9534 support for redirecting crossing jumps to non-crossing.
9536 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9538 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9539 also for naked functions.
9541 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9543 * config/arc/arc.md (add_shift): New pattern.
9544 (add_shift2): Likewise.
9545 (sub_shift): Likewise.
9546 (sub_shift_cmp0_noout): Likewise.
9547 (compare_si_ashiftsi): Likewise.
9548 (xbfu_cmp0_noout): New combine pattern.
9549 (xbfu_cmp0"): Likewise.
9550 (movsi_set_cc_insn): Place the predicable variant first.
9551 (commutative_binary_cmp0_noout): Remove clobber.
9552 (commutative_binary_cmp0): New pattern.
9553 (noncommutative_binary_cmp0): Likewise.
9554 (noncommutative_binary_cmp0_noout): Likewise.
9555 (noncommutative_binary_comparison_result_used): Removed.
9556 (rsub_cmp0): New pattern.
9557 (rsub_cmp0_noout): Likewise.
9558 (extzvsi): Changed, keep only meaningful variants.
9559 (SQH, SEZ): New iterators.
9560 (SQH_postfix): New mode attribute.
9561 (SEZ_prefix): New code attribute.
9562 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9563 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9564 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9566 (noncommutative_operator): Check the availability of barrel
9569 2018-04-09 Richard Biener <rguenther@suse.de>
9571 PR tree-optimization/85284
9572 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9573 Only use the niter constraining form of simple_iv when the exit
9576 2018-04-09 Tom de Vries <tom@codesourcery.com>
9579 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9580 (define_expand "*memory_barrier"): New define_expand.
9581 (define_insn "memory_barrier"): New insn.
9583 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9585 PR rtl-optimization/80463
9586 PR rtl-optimization/83972
9587 PR rtl-optimization/83480
9589 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9590 correct producer for the insn.
9591 (tidy_control_flow): Fixup seqnos in case of debug insns.
9593 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9595 PR rtl-optimization/83913
9597 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9598 different sched-times when merging exprs.
9600 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9602 PR rtl-optimization/83962
9604 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9605 tidy_fallthru_edge and tidy_control_flow.
9607 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9609 PR rtl-optimization/83530
9611 * sel-sched.c (force_next_insn): New global variable.
9612 (remove_insn_for_debug): When force_next_insn is true, also leave only
9613 next insn in the ready list.
9614 (sel_sched_region): When the region wasn't scheduled, make another pass
9615 over it with force_next_insn set to 1.
9617 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9619 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9621 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9622 for interrupt control.
9623 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9624 functions for interrupt control.
9625 * config/nds32/nds32-intrinsic.md: Likewise.
9626 * config/nds32/nds32_intrinsic.h: Likewise.
9627 * config/nds32/nds32.h (nds32_builtins): Likewise.
9629 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9631 * config/nds32/nds32.c (nds32_init_machine_status,
9632 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9633 strict_aligned_p field.
9634 (nds32_expand_to_rtl_hook): New function.
9635 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9636 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9638 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9639 Chung-Ju Wu <jasonwucj@gmail.com>
9641 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9642 * config/nds32/nds32-n7.md: New file.
9643 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9644 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9646 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9647 * config/nds32/nds32.md (pipeline_model): Add n7.
9648 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9649 * config/nds32/pipelines.md: Include n7 settings.
9651 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9652 Chung-Ju Wu <jasonwucj@gmail.com>
9654 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9655 * config/nds32/nds32-e8.md: New file.
9656 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9657 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9659 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9660 * config/nds32/nds32.md (pipeline_model): Add e8.
9661 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9662 * config/nds32/pipelines.md: Include e8 settings.
9664 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9665 Chung-Ju Wu <jasonwucj@gmail.com>
9667 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9668 * config/nds32/nds32-n8.md: New file.
9669 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9670 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9672 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9673 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9674 * config/nds32/nds32.md (pipeline_model): Add n8.
9675 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9676 * config/nds32/pipelines.md: Include n8 settings.
9678 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9679 Chung-Ju Wu <jasonwucj@gmail.com>
9681 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9682 * config/nds32/nds32-n9-2r1w.md: New file.
9683 * config/nds32/nds32-n9-3r2w.md: New file.
9684 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9685 nds32_register_ports): New or modify for cpu n9.
9686 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9688 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9689 * config/nds32/nds32-utils.c: New file.
9690 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9691 TARGET_MUL_SLOW): Define.
9692 * config/nds32/nds32.md (pipeline_model): New attribute.
9693 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9694 New options that support cpu n9.
9695 * config/nds32/pipelines.md: Include n9 settings.
9696 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9698 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9700 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9701 information if necessary.
9702 (output_cond_branch_compare_zero): Likewise.
9703 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9704 (nds32_target_alignment): Refine for alignment.
9705 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9706 (FUNCTION_BOUNDARY): Modify.
9707 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9709 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9711 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9713 * config/nds32/constants.md (unspec_volatile_element): Add values for
9714 TLB operation and data prefetch.
9715 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9716 functions for TLB operation and data prefetch.
9717 * config/nds32/nds32-intrinsic.md: Likewise.
9718 * config/nds32/nds32_intrinsic.h: Likewise.
9719 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9720 (nds32_print_operand): Likewise.
9721 * config/nds32/nds32.h (nds32_builtins): Likewise.
9723 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9724 Andrew Pinski <pinsika@gcc.gnu.org>
9727 * match.pd: Use constant_boolean_node of correct type instead of
9728 boolean_true_node or boolean_false_node for simplifying
9729 pointer comparisons to zero.
9731 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9733 PR tree-optimization/80021
9734 * tree.c (verify_type_variant): Make error call in verify_variant_match
9735 translatable and remove final full stop.
9737 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9739 * config/nds32/constants.md (unspec_volatile_element): Add
9740 UNSPEC_VOLATILE_EH_RETURN.
9741 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9742 nds32_output_stack_pop): Support dwarf exception handling process.
9743 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9744 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9745 exception handling process.
9746 (nds32_compute_stack_frame): Likewise.
9747 (nds32_return_addr_rtx): Likewise.
9748 (nds32_initial_elimination_offset): Likewise.
9749 (nds32_expand_prologue): Likewise.
9750 (nds32_expand_epilogue): Likewise.
9751 (nds32_dynamic_chain_address): New function.
9752 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9754 (DYNAMIC_CHAIN_ADDRESS): Define.
9755 (EH_RETURN_DATA_REGNO): Define.
9756 (EH_RETURN_STACKADJ_RTX): Define.
9757 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9758 patterns for dwarf exception handling.
9760 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9762 * config/nds32/nds32.h: Clean up obsolete macros.
9764 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9766 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9767 Add enum values for particular instructions.
9768 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9769 particular intrinsic functions.
9770 * config/nds32/nds32-intrinsic.md: Likewise.
9771 * config/nds32/nds32_intrinsic.h: Likewise.
9772 * config/nds32/nds32.h (nds32_builtins): Likewise.
9773 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9774 (btst, ave): New patterns for particular instructions.
9776 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9778 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9779 Add enum values for atomic load/store and memory sync.
9780 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9782 * config/nds32/nds32-intrinsic.md: Likewise.
9783 * config/nds32/nds32_intrinsic.h: Likewise.
9784 * config/nds32/nds32.h (nds32_builtins): Likewise.
9786 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9788 PR tree-optimization/85257
9789 * fold-const.c (native_encode_vector): If not all elts could fit
9790 and off is -1, return 0 rather than offset.
9791 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9792 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9793 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9794 adjust buffer in native_interpret_expr call.
9796 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9798 * config/nds32/constants.md (unspec_volatile_element): Add cache
9799 control enum values.
9800 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9801 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9802 * config/nds32/nds32.c (nds32_cctl_names): New.
9803 (nds32_print_operand): Handle cache control register names.
9804 * config/nds32/nds32.h (nds32_builtins): New enum values.
9805 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9807 * config/nds32/nds32.md (type): Add mmu.
9808 * config/nds32/pipelines.md (simple_insn): Add mmu.
9810 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9812 * config/nds32/nds32.md (type): Remove call.
9813 * config/nds32/pipelines.md (simple_insn): Likewise.
9815 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9817 * config/nds32/constants.md (unspec_volatile_element): Add
9818 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9819 UNSPEC_VOLATILE_FMFCFG.
9820 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9821 description for fmfcfg and fmfcsr.
9822 (bdesc_1arg): Add fmtcsr.
9823 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9824 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9825 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9826 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9827 unspec_fmfcfg): New patterns.
9828 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9829 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9830 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9831 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9832 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9833 __nds32__fmfcfg): Define.
9835 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9837 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9838 intrinsic register names.
9839 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9840 intrinsic register enum values and macros.
9842 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9844 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9845 for load/store addressing form.
9846 (nds32_print_operand_address): Likewise.
9848 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9851 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9852 based on LABEL_REF. Remove useless assertion.
9853 (pic_address_needs_scratch): Fix formatting.
9854 (sparc_legitimize_pic_address): Minor tweaks.
9855 (sparc_delegitimize_address): Adjust assertion accordingly.
9856 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9857 into symbolic_operand.
9858 (movsi_high_pic_label_ref): Likewise.
9859 (movsi_lo_sum_pic_label_ref): Likewise.
9860 (movdi_pic_label_ref): Likewise.
9861 (movdi_high_pic_label_ref): Likewise.
9862 (movdi_lo_sum_pic_label_ref): Likewise.
9864 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9866 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9867 custom LIB_SPEC setup.
9869 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9870 Kito Cheng <kito.cheng@gmail.com>
9872 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9873 * config/riscv/freebsd.h: New.
9875 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9877 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9878 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9881 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9882 Kito Cheng <kito.cheng@gmail.com>
9884 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9885 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9886 * config/nds32/nds32-protos.h (nds32_output_call,
9887 nds32_output_return): Declare.
9888 * config/nds32/nds32.md: Refine all the call and return patterns.
9890 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9893 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9894 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9896 PR rtl-optimization/84872
9897 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9898 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9901 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9903 * expr.c (copy_blkmode_to_reg): Revert 254862.
9904 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9906 2018-04-06 Richard Biener <rguenther@suse.de>
9909 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9910 after seeing a component reference with an adjacent field. Treat
9911 refs to arrays at struct end of external decls similar to
9912 refs to unconstrained commons.
9914 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9917 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9918 look through SAVE_EXPRs with non-side-effects argument. Adjust
9920 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9923 2018-04-06 Richard Biener <rguenther@suse.de>
9926 * alias.c (find_base_term): New wrapper around find_base_term
9927 unwinding CSELIB_VAL_PTR changes.
9928 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9931 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9933 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9935 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9936 constant definitions.
9937 ("nop"): lr 0,0 -> nopr r0
9938 ("nop_lr0", "nop_lr1"): New insn definitions.
9940 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9942 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9943 NDS32_V3PUSH_AVAILABLE_P macro.
9945 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9946 Chung-Ju Wu <jasonwucj@gmail.com>
9948 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9949 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9950 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9951 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9952 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9953 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9954 * config/nds32/constraints.md: New constraints and checking for hard
9955 float configuration.
9956 * config/nds32/iterators.md: New mode iterator and attribute for hard
9957 float configuration.
9958 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9960 * config/nds32/nds32-fpu.md: New file.
9961 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9962 deal with hard float code generation.
9963 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9965 (abi_type, float_reg_number): New enum type.
9966 * config/nds32/nds32-predicates.c: New predicates for hard float.
9967 * config/nds32/nds32-protos.h: Declare functions for hard float.
9968 * config/nds32/nds32.c: Implementation for hard float configuration.
9969 * config/nds32/nds32.h: Definitions for hard float configuration.
9970 * config/nds32/nds32.md: Include hard float machine description and
9971 modify patterns for hard float configuration.
9972 * config/nds32/nds32.opt: New options for hard float configuration.
9973 * config/nds32/predicates.md: New predicates for hard float
9976 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9978 * common/config/nds32/nds32-common.c
9979 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9981 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9984 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9985 CONSTRUCTOR_ELT (ctor, ...)->value.
9987 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9990 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9992 2018-04-05 Tom de Vries <tom@codesourcery.com>
9995 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9998 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9999 Kito Cheng <kito.cheng@gmail.com>
10001 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10002 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10003 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10004 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10006 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10007 Kito Cheng <kito.cheng@gmail.com>
10009 * config/nds32/constraints.md (Ufe): New memory constraint.
10010 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10011 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10012 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10014 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10015 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10017 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10019 * config/nds32/nds32.md: Use optimize_size in the condition for
10020 alu-shift instructions.
10022 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10024 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10026 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10028 * config/nds32/nds32.md (negsi2): Refine pattern.
10030 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10031 Chung-Ju Wu <jasonwucj@gmail.com>
10033 * config/nds32/iterators.md (shift_rotate): New code iterator.
10034 (shift): New code attribute.
10035 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10036 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10037 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10038 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10039 bit-wise operations.
10040 (andsi3, *andsi3): Ditto.
10041 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10042 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10043 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10044 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10045 nds32_ior_operand, nds32_xor_operand): New predicates.
10047 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10049 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10050 (addsi3, subsi3): ... this.
10052 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10054 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10056 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10058 * config/nds32/nds32.md: Adjust indention.
10060 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10062 * config/nds32/nds32.md (feature): New attribute.
10064 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10066 * config/nds32/nds32.md (subtype): New attribute.
10068 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10071 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10072 expansion to perform a bitwise AND of the argument followed by a
10073 boolean negation of the result.
10075 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10077 PR rtl-optimization/84878
10078 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10079 the basic block. Assert the use reference is not artificial and that
10080 it has an associated insn.
10082 2018-04-04 Michael Matz <matz@suse.de>
10084 * builtins.c (compute_objsize): Pass correct operand
10085 to array_at_struct_end_p.
10087 2018-04-04 Richard Biener <rguenther@suse.de>
10090 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10091 from contexts for DINFO_LEVEL_TERSE and below.
10093 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10095 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10096 resiter_operand condition.
10097 * config/nds32/nds32.md (*move<mode>): Ditto.
10099 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10100 Monk Chiang <sh.chiang04@gmail.com>
10102 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10104 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10106 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10108 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10109 Kito Cheng <kito.cheng@gmail.com>
10111 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10112 nds32_cond_code_str, output_cond_branch,
10113 output_cond_branch_compare_zero, nds32_expand_cbranch,
10114 nds32_expand_cstore, nds32_expand_movcc,
10115 nds32_output_cbranchsi4_equality_zero,
10116 nds32_output_cbranchsi4_equality_reg,
10117 nds32_output_cbranchsi4_equality_reg_or_const_int,
10118 nds32_output_cbranchsi4_greater_less_zero: New functions.
10119 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10120 nds32_expand_cstore, nds32_expand_movcc,
10121 nds32_output_cbranchsi4_equality_zero,
10122 nds32_output_cbranchsi4_equality_reg,
10123 nds32_output_cbranchsi4_equality_reg_or_const_int,
10124 nds32_output_cbranchsi4_greater_less_zero): Declare.
10125 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10126 nds32_rimm11s_operand): New predicates.
10127 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10128 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10131 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10133 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10134 * config/nds32/nds32.md: Ditto.
10135 * config/nds32/pipelines.md: Ditto.
10137 2018-04-04 Richard Biener <rguenther@suse.de>
10139 PR tree-optimization/85168
10140 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10141 propagating abnormals.
10143 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10145 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10147 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10148 Kito Cheng <kito.cheng@gmail.com>
10150 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10151 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10152 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10153 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10154 * config/nds32/nds32.md (sibcall_internal): New.
10155 (sibcall_register): Remove.
10156 (sibcall_immediate): Remove.
10157 (sibcall_value_internal): New.
10158 (sibcall_value_register): Remove.
10159 (sibcall_value_immediate): Remove.
10160 * config/nds32/predicates.md (nds32_general_register_operand): New.
10161 (nds32_call_address_operand): New.
10163 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10165 PR rtl-optimization/85167
10166 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10167 bb_defs if *split_p, instead preinitialize it to NULL.
10169 PR tree-optimization/85156
10170 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10171 evaluating the argument multiple times.
10173 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10175 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10177 (_mm_cvtpd_ps): Likewise.
10178 (_mm_cvttpd_epi32): Likewise.
10179 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10180 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10181 vector, pixel, and bool following altivec.h include.
10183 2018-04-03 Martin Sebor <msebor@redhat.com>
10185 * doc/extend.texi (Common Function Attributes): Clarify.
10186 (const attribute): Likewise.
10187 (pure attribute): Likewise.
10189 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10192 * config/i386/i386.c (ix86_expand_vector_set): Use
10193 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10195 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10197 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10198 instructions when changing rounding bits to preserve precision bits
10199 in the x87 control word.
10201 2018-04-03 Martin Liska <mliska@suse.cz>
10203 PR tree-optimization/82491
10204 * rtl.h (strip_offset_and_add): Replace += suboffset with
10205 poly_uint64 () + suboffset.
10207 2018-03-29 Martin Liska <mliska@suse.cz>
10208 Martin Jambor <mjambor@suse.cz>
10211 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10212 param_type is not an integral or pointer type.
10214 2018-04-03 Richard Biener <rguenther@suse.de>
10216 * sese.h (recompute_all_dominators): Remove.
10218 2018-04-02 Martin Sebor <msebor@redhat.com>
10220 * doc/invoke.texi (-Wrestrict): Fix typos.
10222 2018-04-02 Jim Wilson <jimw@sifive.com>
10224 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10225 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10226 (<optab>di3, <optab>si3_extend): Likewise.
10227 (<optab>si3_mask, <optab>si3_mask_1): New.
10228 (<optab>di3_mask, <optab>di3_mask_1): New.
10229 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10230 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10231 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10233 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10235 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10238 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10240 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10241 (nds32_canonicalize_comparison): New function.
10243 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10244 Kito Cheng <kito.cheng@gmail.com>
10245 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10247 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10248 * config/nds32/constants.md (unspec_volatile_element): Add
10249 UNSPEC_VOLATILE_RELAX_GROUP.
10250 * config/nds32/nds32-relax-opt.c: New file.
10251 * config/nds32/nds32-predicates.c
10252 (nds32_symbol_load_store_p): New function.
10253 * config/nds32/nds32-protos.h
10254 (nds32_symbol_load_store_p): Declare function.
10255 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10256 * config/nds32/nds32.c
10257 (nds32_register_pass): New function to register pass.
10258 (nds32_register_passes): New function to register passes.
10259 * config/nds32/nds32.md (relax_group): New pattern.
10260 * config/nds32/nds32.opt (mrelax-hint): New option.
10261 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10263 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10265 * config/nds32/t-nds32: Modify files dependency.
10267 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10269 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10270 (PROFILE_HOOK): Define its implementation.
10272 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10274 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10275 type and 32-bit size.
10277 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10279 PR middle-end/85090
10280 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10281 (V_128_256): New mode iterator.
10282 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10283 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10284 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10286 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10287 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10289 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10292 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10293 NaN inputs correctly.
10295 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10298 * config/rs6000/vsx.md (??r): New mode attribute.
10299 (*vsx_mov<mode>_64bit): Use it.
10300 (*vsx_mov<mode>_32bit): Likewise.
10302 2018-03-30 Martin Sebor <msebor@redhat.com>
10304 PR tree-optimization/84818
10305 * builtins.c (check_access): Use warning_n.
10307 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10310 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10312 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10315 2018-03-30 Julia Koval <julia.koval@intel.com>
10318 * x86-tune.def (movx, partial_reg_dependency): Enable for
10321 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10323 PR inline-asm/84985
10324 * lra-constraints.c (process_alt_operands): Move setting
10325 this_alternative_matches below.
10327 2018-03-29 Martin Liska <mliska@suse.cz>
10330 * doc/invoke.texi: Document how LTO works with debug info.
10331 Describe auto-load support of binutils. Mention 'x86-64'
10332 as valid option value of -march option.
10334 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10336 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10339 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10340 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10341 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10344 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10347 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10348 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10349 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10350 for RS6000_BTM_POWERPC64.
10351 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10352 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10353 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10356 (DIVDEU): Likewise.
10358 2018-03-28 Carl Love <cel@us.ibm.com>
10361 2017-09-27 Carl Love <cel@us.ibm.com>
10363 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10364 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10365 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10368 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10370 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10371 instead of __vector bool.
10372 (_mm_max_pu8): Likewise.
10373 (_mm_min_pi16): Likewise.
10375 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10378 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10379 (DIVWEUO): Likewise.
10380 (DIVDEO): Likewise.
10381 (DIVDEUO): Likewise.
10382 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10383 DIVWEUO and DIVDEUO.
10384 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10385 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10386 (div_extend): Likewise.
10387 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10389 (__builtin_divweuo): Likewise.
10390 (__builtin_divdeo): Likewise.
10391 (__builtin_divdeuo): Likewise.
10393 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10396 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10397 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10399 PR tree-optimization/82004
10400 * gimple-match-head.c (optimize_pow_to_exp): New function.
10401 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10402 Don't fold to exp if optimize_pow_to_exp is false.
10404 2018-03-28 Martin Liska <mliska@suse.cz>
10407 * calls.c (initialize_argument_information): Fix trailing space.
10408 * common.opt: Fix typo and provide better explanation for
10409 -fsanitize-coverage option.
10410 * config/i386/i386.opt: Fix typo.
10412 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10413 Martin Liska <mliska@suse.cz>
10416 * gimplify.c (asan_poison_variable): Don't do the check for
10417 gimplify_omp_ctxp here.
10418 (gimplify_decl_expr): Do it here.
10419 (gimplify_target_expr): Likewise.
10421 2018-03-28 Martin Liska <mliska@suse.cz>
10424 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10425 chkp_type_bounds_count if MPX is not enabled.
10427 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10429 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10431 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10434 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10435 function to create the function decl for complex long double
10436 multiply and divide for -mabi=ieeelongdouble.
10437 (init_float128_ieee): Call it.
10439 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10442 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10443 -fcf-protection=branch -mibt.
10444 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10446 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10449 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10450 arm_word_relocations.
10452 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10455 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10456 extern array declarations.
10458 2018-03-27 Richard Biener <rguenther@suse.de>
10460 PR middle-end/84067
10461 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10462 explicit single_use checks.
10464 2018-03-27 Richard Biener <rguenther@suse.de>
10466 PR tree-optimization/85082
10467 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10470 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10472 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10473 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10474 Turn on fasynchronous-unwind-tables and funwind-tables.
10476 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10479 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10480 (*bmi_blsr_<mode>_ccz): Ditto.
10482 2018-03-26 Tom de Vries <tom@codesourcery.com>
10484 PR tree-optimization/85063
10485 * omp-general.c (offloading_function_p): New function. Factor out
10487 * omp-offload.c (pass_omp_target_link::gate): ... here.
10488 * omp-general.h (offloading_function_p): Declare.
10489 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10490 with attribute omp declare target for offloading functions.
10492 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10494 PR tree-optimization/84005
10495 * tree-data-ref.h (get_base_for_alignment): Declare.
10496 * tree-data-ref.c (get_base_for_alignment_1): New function.
10497 (get_base_for_alignment): Likewise.
10498 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10499 get_base_for_alignment to find a suitable base object, instead
10500 of always using drb->base_address.
10502 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10504 PR inline-asm/85022
10505 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10506 known size by default.
10508 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10510 PR inline-asm/85030
10511 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10512 and non BLKmode operands.
10514 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10517 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10518 Clean up attributes.
10520 2018-03-23 Richard Biener <rguenther@suse.de>
10523 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10524 we are going to emit early debug for LTO.
10526 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10528 PR inline-asm/85034
10529 * function.c (match_asm_constraints_1): Don't optimize if input
10530 doesn't satisfy general_operand predicate for output's mode.
10532 PR inline-asm/85022
10533 * alias.c (write_dependence_p): Don't require for x_canonicalized
10534 non-VOIDmode if x has VOIDmode.
10537 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10538 just don't try to optimize it rather than assert it never happens.
10540 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10542 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10543 macro expansions for definition of ST_INTERNAL_<mode> and
10544 LD_INTERNAL_<mode> builtins.
10545 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10547 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10549 (altivec_expand_st_builtin): Likewise.
10550 (altivec_expand_builtin): Remove calls to deleted functions.
10551 (rs6000_address_for_altivec): Delete this function.
10552 * config/rs6000/vector.md: Remove expands for
10553 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10555 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10558 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10559 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10560 re-computing once computed.
10561 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10562 (arm_init_machine_status): Initialize
10563 machine->static_chain_stack_bytes.
10565 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10568 * doc/extend.texi: Add four new prototypes for vec_ld.
10569 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10570 definitions for more logical presentation.
10571 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10572 entries for V1TI variants of __builtin_altivec_ld builtin.
10573 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10574 handling of V1TI variant of LVX icode pattern.
10575 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10576 (rs6000_gimple_fold_builtin): Likewise.
10577 (altivec_init_builtins): Add code to define
10578 __builtin_altivec_lvx_v1ti function.
10580 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10582 PR inline-asm/84941
10583 * function.c (match_asm_constraints_1): Don't do the optimization
10584 if input isn't a REG, SUBREG, MEM or constant.
10586 2018-03-22 Tom de Vries <tom@codesourcery.com>
10588 PR tree-optimization/84956
10589 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10590 bb_has_abnormal_pred.
10592 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10595 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10596 DECL_INITIAL (decl) to decl at the end.
10597 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10598 adjust the comment.
10600 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10602 * doc/extend.texi (__builtin_tgmath): Document when complex
10603 integer types are treated as _Complex _Float64.
10605 2018-03-21 Tom de Vries <tom@codesourcery.com>
10607 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10609 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10611 PR tree-optimization/84960
10612 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10613 if it is ENTRY block, move them into single succ of ENTRY in that case.
10615 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10617 PR tree-optimization/84811
10618 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10619 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10620 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10621 (known_subrange_p): Update accordingly. Cast each value involved
10622 in the size comparison, rather than casting the result of the
10625 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10627 PR tree-optimization/84982
10628 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10629 by flipping the least significant bit rather than all bits from
10630 bitpos to bitpos + bitsize - 1.
10632 2018-03-21 Nathan Sidwell <nathan@acm.org>
10634 * doc/extend.texi (Deprecated Features): Remove mention of
10635 long-deleted deprecations.
10637 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10640 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10641 * configure: Regenerate.
10643 2018-03-21 Tom de Vries <tom@codesourcery.com>
10645 PR tree-optimization/83126
10646 * tree-parloops.c (num_phis): New function.
10647 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10649 2018-03-21 Nathan Sidwell <nathan@acm.org>
10651 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10652 mention anon-struct/union members and trailing attributes.
10654 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10656 PR tree-optimization/84969
10657 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10658 builtin memset partitions if they set different rhs values.
10660 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10662 PR rtl-optimization/84989
10663 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10664 VEC_DUPLICATE with scalar result mode.
10666 2018-03-21 Martin Liska <mliska@suse.cz>
10669 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10670 not intended return statement.
10672 2018-03-21 Martin Liska <mliska@suse.cz>
10675 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10676 (chkp_find_bound_slots_1): Limit number of iterations.
10678 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10681 * Minor grammar fixes for x86 options.
10683 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10686 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10687 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10689 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10692 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10693 (altivec_lvx_<mode>_1op): Likewise.
10694 (altivec_stvx_<mode>_2op): Likewise.
10695 (altivec_stvx_<mode>_1op): Likewise.
10696 (altivec_lvx_<VM2:mode>): New define_expand.
10697 (altivec_stvx_<VM2:mode>): Likewise.
10698 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10699 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10700 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10701 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10702 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10703 (rs6000_gen_lvx): Likewise.
10704 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10705 (altivec_expand_stv_builtin): Likewise.
10706 (altivec_expand_builtin): Likewise.
10707 * config/rs6000/vector.md: Likewise.
10709 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10712 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10715 2018-03-20 Richard Biener <rguenther@suse.de>
10718 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10719 sign-conversions as zero, fall back to standard scalar_stmt
10722 2018-03-20 Martin Liska <mliska@suse.cz>
10725 * predict.c (rebuild_frequencies): Handle case when we have
10726 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10728 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10731 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10732 flag_section_anchors.
10733 * varasm.c (use_blocks_for_decl_p): Remove hack for
10734 dw2_force_const_mem.
10737 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10739 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10740 be created, use lowpart_subreg of operands[0] rather than operands[0]
10742 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10743 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10744 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10745 and n constraint instead of aarch64_shift_imm_di and Usd.
10746 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10747 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10749 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10752 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10753 to favor GPR over NEON registers.
10754 (<shift>di3_neon): Likewise.
10756 2018-03-20 Tom de Vries <tom@codesourcery.com>
10759 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10760 (nvptx_process_pars): Emit bar.sync asap and alap.
10762 2018-03-20 Tom de Vries <tom@codesourcery.com>
10765 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10766 seen_label if seen_label is already set.
10768 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10771 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10772 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10773 Use 1U instead of 1. Formatting fixes.
10776 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10777 instead of TREE_TYPE (s1) for the return value.
10779 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10781 PR tree-optimization/84946
10782 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10783 bitsize + bitsize in poly_uint64 rather than poly_int64.
10786 * dwarf2asm.c: Include fold-const.c.
10787 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10788 of decl rather than decl itself.
10790 PR rtl-optimization/84643
10791 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10793 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10796 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10797 calling assemble_variable.
10799 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10802 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10803 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10805 2018-03-19 Jim Wilson <jimw@sifive.com>
10808 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10809 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10810 (riscv_first_stack_step): Likewise.
10811 (riscv_option_override): Use STACK_BOUNDARY instead of
10812 MIN_STACK_BOUNDARY.
10813 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10814 MIN_STACK_BOUNDARY.
10815 (BIGGEST_ALIGNMENT): Set to 128.
10816 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10817 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10820 2018-03-19 Richard Biener <rguenther@suse.de>
10822 PR tree-optimization/84933
10823 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10824 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10826 2018-03-19 Richard Biener <rguenther@suse.de>
10828 PR tree-optimization/84859
10829 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10830 (cond_if_else_store_replacement): Perform sinking operation on
10831 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10832 Generalize what a BB with a single eligible store is.
10834 2018-03-19 Richard Biener <rguenther@suse.de>
10836 PR tree-optimization/84929
10837 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10838 chrec_is_positive against non-chrec arg.
10840 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10843 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10845 2018-03-18 Martin Liska <mliska@suse.cz>
10847 PR rtl-optimization/84635
10848 * regrename.c (build_def_use): Use matches_mode only when
10851 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10853 PR tree-optimization/84913
10854 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10855 vectorize chains of COND_EXPRs.
10857 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10859 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10861 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10863 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10865 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10867 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10869 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10870 Kito Cheng <kito.cheng@gmail.com>
10872 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10873 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10874 (nds32_adjust_reg_alloc_order): New function.
10875 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10877 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10879 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10880 nds32_print_operand, nds32_print_operand_address): Use
10881 HOST_WIDE_INT_PRINT_DEC instead.
10883 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10885 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10887 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10890 * config/i386/i386.c (initial_ix86_tune_features,
10891 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10892 unsigned long long.
10893 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10894 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10895 rather than 1u << ix86_tune. Formatting fix.
10896 (ix86_option_override_internal): Change ix86_arch_mask from
10897 unsigned int to unsigned HOST_WIDE_INT, initialize to
10898 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10899 (ix86_function_specific_restore): Likewise.
10901 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10904 * postreload.c (reload_combine_recognize_pattern): Perform
10905 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10906 truncate_int_for_mode the result for the destination's mode.
10909 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10910 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10912 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10915 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10918 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10919 regno_allocno_class_array and sorted_pseudos.
10920 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10921 insns where regno is used.
10923 2018-03-16 Martin Liska <mliska@suse.cz>
10926 * multiple_target.c (create_dispatcher_calls): Redirect
10927 reference in the symbol table.
10929 2018-03-16 Martin Liska <mliska@suse.cz>
10932 * multiple_target.c (create_dispatcher_calls): Redirect also
10935 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10939 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10940 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10941 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10943 2018-03-16 Julia Koval <julia.koval@intel.com>
10945 * doc/invoke.texi (Skylake Server): Add CLWB.
10946 Cannonlake): Remove CLWB.
10948 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10950 PR tree-optimization/84841
10951 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10953 (FLOAT_ONE_CONST_TYPE): Define.
10954 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10955 (sort_by_operand_rank): Put entries with higher constant_type last
10956 rather than first to match comments.
10958 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10960 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10963 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10966 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10967 check and use address of target always.
10969 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10972 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10973 (indirect_thunk_bnd_needed): Likewise.
10974 (indirect_thunks_used): Likewise.
10975 (indirect_thunks_bnd_used): Likewise.
10976 (indirect_return_needed): New.
10977 (indirect_return_bnd_needed): Likewise.
10978 (output_indirect_thunk_function): Add a bool argument for
10980 (output_indirect_thunk_function): Don't generate alias for
10981 function return thunk.
10982 (ix86_code_end): Call output_indirect_thunk_function to generate
10983 function return thunks.
10984 (ix86_output_function_return): Set indirect_return_bnd_needed
10985 and indirect_return_needed instead of indirect_thunk_bnd_needed
10986 and indirect_thunk_needed.
10988 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10990 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10991 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10992 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10994 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10995 Paul Hua <paul.hua.gm@gmail.com>
10998 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11000 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11002 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11003 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11004 resp. SFmode cases.
11006 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11009 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11010 instead of GET_MODE_SIZE when comparing Units.
11012 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11015 * varasm.c (hash_section): Return an unchangeble hash value
11016 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11017 Return !aarch64_can_use_per_function_literal_pools_p ().
11019 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11022 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11023 rather than address of cmode as last argument to prepare_cmp_insn.
11025 2018-03-15 Julia Koval <julia.koval@intel.com>
11027 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11028 F_AVX512VNNI, F_AVX512BITALG): New.
11030 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11033 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11034 insn for floating-point loads and stores.
11036 2018-03-14 Carl Love <cel@us.ibm.com>
11038 * config/rs6000/rs6000-c.c: Add macro definitions for
11039 ALTIVEC_BUILTIN_VEC_PERMXOR.
11040 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11041 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11042 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11043 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11045 * config/doc/extend.texi: Add prototypes for vec_permxor.
11047 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11050 * diagnostic-show-locus.c (class layout_point): Convert m_line
11051 from int to linenum_type.
11052 (line_span::comparator): Use linenum "compare" function when
11053 comparing line numbers.
11054 (test_line_span): New function.
11055 (layout_range::contains_point): Convert param "row" from int to
11057 (layout_range::intersects_line_p): Likewise.
11058 (layout::will_show_line_p): Likewise.
11059 (layout::print_source_line): Likewise.
11060 (layout::should_print_annotation_line_p): Likewise.
11061 (layout::print_annotation_line): Likewise.
11062 (layout::print_leading_fixits): Likewise.
11063 (layout::annotation_line_showed_range_p): Likewise.
11064 (struct line_corrections): Likewise for field m_row.
11065 (line_corrections::line_corrections): Likewise for param "row".
11066 (layout::print_trailing_fixits): Likewise.
11067 (layout::get_state_at_point): Likewise.
11068 (layout::get_x_bound_for_row): Likewise.
11069 (layout::print_line): Likewise.
11070 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11071 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11072 * input.c (selftest::test_linenum_comparisons): New function.
11073 (selftest::input_c_tests): Call it.
11074 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11075 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11076 * selftest.h (ASSERT_GT): New macro.
11077 (ASSERT_GT_AT): New macro.
11078 (ASSERT_LT): New macro.
11079 (ASSERT_LT_AT): New macro.
11081 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11083 PR rtl-optimization/84780
11084 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11086 2018-03-14 Martin Liska <mliska@suse.cz>
11088 * tree.c (record_node_allocation_statistics): Use
11089 get_stats_node_kind.
11090 (get_stats_node_kind): New function extracted from
11091 record_node_allocation_statistics.
11092 (free_node): Use get_stats_node_kind.
11094 2018-03-14 Richard Biener <rguenther@suse.de>
11096 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11097 that the value-set of ANTIC_IN doesn't grow.
11100 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11102 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11103 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11105 2018-03-14 Julia Koval <julia.koval@intel.com>
11107 * config.gcc (icelake-client, icelake-server): New.
11109 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11110 (initial_ix86_arch_features): Ditto.
11111 (PTA_SKYLAKE): Add SGX.
11112 (PTA_ICELAKE): Remove.
11113 (PTA_ICELAKE_CLIENT): New.
11114 (PTA_ICELAKE_SERVER): New.
11115 (ix86_option_override_internal): Split up icelake on icelake client and
11117 (get_builtin_code_for_version): Ditto.
11118 (fold_builtin_cpu): Ditto.
11119 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11120 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11121 * config/i386/i386.h (processor_type): Ditto.
11122 * doc/invoke.texi: Ditto.
11124 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11127 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11128 INTEGER_CST offset, add it together with bitpos / 8 and
11129 sign extend based on POINTER_SIZE.
11133 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11136 * config/i386/constraints.md (Yc): New register constraint.
11137 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11138 Use Yc constraint for alternative 2 of operand 0. Remove
11139 preferred_for_speed attribute.
11141 2018-03-14 Richard Biener <rguenther@suse.de>
11143 PR tree-optimization/84830
11144 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11145 with the old one to avoid oscillations.
11147 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11150 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11152 (assign_by_spills): Return a flag of reload assignment failure.
11153 Do not process the reload assignment failures. Do not spill other
11154 reload pseudos if they has the same reg class. Update n if
11156 (lra_assign): Add a return arg. Set up from the result of
11157 assign_by_spills call.
11158 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11159 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11160 usage_insns if it is not NULL.
11161 (spill_hard_reg_in_range): New function.
11162 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11163 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11164 function prototypes.
11165 (lra_assign): Change prototype.
11166 * lra.c (lra): Add code to deal with fails by splitting hard reg
11169 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11171 * config/riscv/riscv.opt (mrelax): New option.
11172 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11173 "norelax" when riscv_mrelax is disabled.
11174 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11176 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11179 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11180 reassociation for int modes.
11182 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11184 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11185 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11187 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11188 * config/aarch64/aarch64-sve.md
11189 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11190 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11191 (*extend<mode><Vwide>2): Rename to...
11192 (aarch64_sve_extend<mode><Vwide>2): ...this.
11193 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11194 renaming the old pattern to...
11195 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11197 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11198 define_expand, renaming the old pattern to...
11199 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11200 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11201 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11202 account when deciding which SVE instruction the optab should use.
11203 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11205 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11207 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11208 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11209 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11210 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11211 (tlsdesc_small_<mode>): Turn a define_expand and use
11212 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11213 (tlsdesc_small_advsimd_<mode>): ...this.
11214 (tlsdesc_small_sve_<mode>): New pattern.
11216 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11218 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11219 (UNSPEC_UMUL_HIGHPART): New constants.
11220 (MUL_HIGHPART): New int iteraor.
11221 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11222 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11224 (*<su>mul<mode>3_highpart): New define_insn.
11226 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11229 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11232 2018-03-13 Martin Liska <mliska@suse.cz>
11235 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11237 (sem_item_optimizer::~sem_item_optimizer): Release it.
11238 (sem_item_optimizer::merge_classes): Register variable aliases.
11239 (sem_item_optimizer::fixup_pt_set): New function.
11240 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11241 * ipa-icf.h: Declare new variables and functions.
11243 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11245 PR middle-end/84834
11246 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11247 integer_pow2p@2 and test integer_pow2p in condition.
11248 (A < 0 ? C : 0): Similarly for @1.
11250 PR middle-end/84831
11251 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11252 characters starting at p contain '\0' character, don't look beyond
11256 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11257 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11260 * reg-stack.c (change_stack): Change update_end var from int to
11261 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11262 also call set_block_for_insn on the newly added insns and rescan.
11265 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11266 on the last operand.
11269 * tree.c (stabilize_reference_1): Return save_expr (e) for
11270 STATEMENT_LIST even if it doesn't have side-effects.
11272 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11274 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11276 2018-03-12 Renlin Li <renlin.li@arm.com>
11278 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11279 aarch64_output_scalar_simd_mov_immediate.
11281 2018-03-12 Martin Sebor <msebor@redhat.com>
11283 PR tree-optimization/83456
11284 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11285 for perfectly overlapping calls to memcpy.
11286 (gimple_fold_builtin_memory_chk): Same.
11287 (gimple_fold_builtin_strcpy): Handle no-warning.
11288 (gimple_fold_builtin_stxcpy_chk): Same.
11289 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11291 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11293 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11294 parameter. Use it for SFmode.
11295 (rs6000_function_arg_advance_1): Adjust.
11296 (rs6000_function_arg): Adjust.
11297 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11299 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11301 PR rtl-optimization/84169
11302 PR rtl-optimization/84780
11303 * combine.c (can_combine_p): Check for a 2-insn combination whether
11304 the destination register is used between the two insns, too.
11306 2018-03-12 Richard Biener <rguenther@suse.de>
11308 PR tree-optimization/84803
11309 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11310 for refs DR analysis didn't process.
11312 2018-03-12 Richard Biener <rguenther@suse.de>
11314 PR tree-optimization/84777
11315 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11316 force-vectorize loops ignore whether we are optimizing for size.
11318 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11320 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11321 (TARGET_MD_ASM_ADJUST): Define.
11323 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11324 Kito Cheng <kito.cheng@gmail.com>
11325 Chung-Ju Wu <jasonwucj@gmail.com>
11327 * config/nds32/nds32.c (nds32_compute_stack_frame,
11328 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11329 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11330 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11331 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11332 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11333 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11334 * config/nds32/nds32.md (prologue, epilogue): Use macro
11335 NDS32_V3PUSH_AVAILABLE_P to do checking.
11337 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11340 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11341 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11342 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11343 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11344 addition of most attributes on !orig_type_die or the attribute not
11345 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11347 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11348 Chung-Ju Wu <jasonwucj@gmail.com>
11350 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11351 __NDS32_VH__ macro.
11352 * config/nds32/nds32.opt (mvh): New option.
11354 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11355 Chung-Ju Wu <jasonwucj@gmail.com>
11357 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11359 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11360 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11363 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11364 Chung-Ju Wu <jasonwucj@gmail.com>
11366 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11368 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11369 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11371 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11372 Kito Cheng <kito.cheng@gmail.com>
11373 Chung-Ju Wu <jasonwucj@gmail.com>
11375 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11376 UNSPEC_FFMISM and UNSPEC_FLMISM.
11377 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11378 for ffb, ffmism and flmism.
11379 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11380 (unspec_ffmism): Ditto.
11381 (unspec_flmism): Ditto.
11382 (nds32_expand_builtin_impl): Check if string extension is available.
11383 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11384 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11386 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11389 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11392 * lra-assigns.c (assign_by_spills): Return a flag of reload
11393 assignment failure. Do not process the reload assignment
11394 failures. Do not spill other reload pseudos if they has the same
11396 (lra_assign): Add a return arg. Set up from the result of
11397 assign_by_spills call.
11398 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11399 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11400 usage_insns if it is not NULL.
11401 (spill_hard_reg_in_range): New function.
11402 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11403 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11404 function prototypes.
11405 (lra_assign): Change prototype.
11406 * lra.c (lra): Add code to deal with fails by splitting hard reg
11409 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11412 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11414 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11417 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11418 (dw_val_node): Add val_symbolic_view.
11419 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11420 (symview_upper_bound): New.
11421 (new_line_info_table): Initialize symviews_since_reset.
11422 (dwarf2out_source_line): Count symviews_since_reset and set
11423 symview_upper_bound.
11424 (dw_val_equal_p): Handle symview.
11425 (add_AT_symview): New.
11426 (print_dw_val): Handle symview.
11427 (attr_checksum, attr_checksum_ordered): Likewise.
11428 (same_dw_val_p, size_of_die): Likewise.
11429 (value_format, output_die): Likewise.
11430 (add_high_low_attributes): Use add_AT_symview for entry_view.
11431 (dwarf2out_finish): Reset symview_upper_bound, clear
11434 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11437 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11438 Add strict argument and use it.
11439 (rs6000_split_multireg_move): Update for new strict argument.
11440 (mem_operand_gpr): Disallow all non-offsettable addresses.
11441 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11443 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11446 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11447 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11448 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11451 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11452 decl, use remap_type if we want to use the type.
11454 2018-03-09 Martin Sebor <msebor@redhat.com>
11456 PR tree-optimization/84526
11457 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11459 (builtin_access::generic_overlap): Be prepared to handle non-array
11462 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11464 PR rtl-optimization/84682
11465 * lra-constraints.c (process_address_1): Check is_address flag
11466 for address constraints.
11467 (process_alt_operands): Likewise.
11468 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11469 preprocess_constraints.
11470 * recog.h (preprocess_constraints): Add oploc parameter.
11472 * recog.c (preprocess_constraints): Test address_operand for
11473 CT_ADDRESS constraints.
11475 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11478 * lra-assigns.c (assign_by_spills): Return a flag of reload
11479 assignment failure. Do not process the reload assignment
11480 failures. Do not spill other reload pseudos if they has the same
11482 (lra_assign): Add a return arg. Set up from the result of
11483 assign_by_spills call.
11484 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11485 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11486 usage_insns if it is not NULL.
11487 (spill_hard_reg_in_range): New function.
11488 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11489 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11490 function prototypes.
11491 (lra_assign): Change prototype.
11492 * lra.c (lra): Add code to deal with fails by splitting hard reg
11495 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11498 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11499 Accept complain bool parameter. Only emit errors if it is true.
11500 (arm_parse_cpu_option_name): Likewise.
11501 (arm_target_thumb_only): Adjust callers of the above.
11502 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11503 prototype to take a default true bool parameter.
11504 (arm_parse_arch_option_name): Likewise.
11506 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11507 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11511 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11512 * configure: Regenerate.
11513 * configure.ac ("linker --version-script option"): New.
11514 ("linker soname option"): New.
11516 2018-03-09 Richard Biener <rguenther@suse.de>
11518 PR tree-optimization/84775
11519 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11520 immediate uses of predicate stmts and mark them modified.
11523 PR tree-optimization/84178
11524 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11526 (version_loop_for_if_conversion): Delay update_ssa call.
11527 (tree_if_conversion): Delay update_ssa until after predicate
11530 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11533 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11534 when the function accesses prior frames.
11536 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11539 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11540 gen_llsym, otherwise call maybe_gen_llsym.
11542 PR inline-asm/84742
11543 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11544 has ',' character inside of it.
11546 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11549 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11550 as clobbering CC_REGNUM.
11552 2018-03-08 Richard Biener <rguenther@suse.de>
11554 PR middle-end/84552
11555 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11556 (follow_copies_to_constant): Do not follow SSA names registered
11559 2018-03-08 Richard Biener <rguenther@suse.de>
11561 PR tree-optimization/84178
11562 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11564 (version_loop_for_if_conversion): Delay update_ssa call.
11565 (tree_if_conversion): Delay update_ssa until after predicate
11568 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11570 PR tree-optimization/84178
11571 * tree-if-conv.c (release_bb_predicate): Remove the
11572 the assertion that the stmts have NULL use_ops.
11573 Discard the statements, asserting that they haven't
11574 yet been added to a BB.
11576 2018-03-08 Richard Biener <rguenther@suse.de>
11578 PR tree-optimization/84746
11579 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11580 (phi_translate): Pass in destination ANTIC_OUT set.
11581 (phi_translate_1): Likewise. For a simplified result lookup
11582 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11583 (phi_translate_set): Adjust.
11584 (do_pre_regular_insertion): Likewise.
11585 (do_pre_partial_partial_insertion): Likewise.
11587 2018-03-08 Martin Liska <mliska@suse.cz>
11589 PR gcov-profile/84735
11590 * doc/gcov.texi: Document usage of profile files.
11591 * gcov-io.h: Document changes in the format.
11593 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11597 * dwarf2out.c (struct dw_line_info_table): Update comments for
11599 (FORCE_RESET_NEXT_VIEW): New.
11600 (FORCE_RESETTING_VIEW_P): New.
11601 (RESETTING_VIEW_P): Check for -1 too.
11602 (ZERO_VIEW_P): Likewise.
11603 (new_line_info_table): Force-reset next view.
11604 (dwarf2out_begin_function): Likewise.
11605 (dwarf2out_source_line): Simplify zero_view_p initialization.
11606 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11607 view directly. Omit view when omitting .loc at line 0.
11609 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11611 PR tree-optimization/84740
11612 * tree-switch-conversion.c (process_switch): Call build_constructors
11613 only if info.phi_count is non-zero.
11615 PR tree-optimization/84739
11616 * tree-tailcall.c (find_tail_calls): Check call arguments against
11617 DECL_ARGUMENTS (current_function_decl) rather than
11618 DECL_ARGUMENTS (func) when checking for tail recursion.
11620 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11622 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11623 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11624 Volker Reichelt's entry and add entries for people that perform
11625 GCC fuzzy testing and report numerous bugs.
11627 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11630 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11631 readonly data in sdata, if that is disabled.
11632 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11633 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11634 -mreadonly-in-sdata option.
11636 2018-03-07 Martin Sebor <msebor@redhat.com>
11638 PR tree-optimization/84468
11639 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11640 basic block when looking for nul assignment.
11642 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11645 * except.h (output_function_exception_table): Adjust prototype.
11646 * except.c (output_function_exception_table): Remove FNNAME parameter
11647 and add SECTION parameter. Ouput one part of the table at a time.
11648 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11649 the first part of the exception table and emit unwind directives.
11650 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11651 (i386_pe_seh_cold_init): Likewise.
11652 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11653 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11654 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11655 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11656 * config/i386/winnt.c: Include except.h.
11657 (struct seh_frame_state): Add reg_offset, after_prologue and
11658 in_cold_section fields.
11659 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11660 (i386_pe_seh_cold_init): New function.
11661 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11662 to seh->in_cold_section.
11663 (seh_emit_push): Record the offset of the push.
11664 (seh_emit_save): Record the offet of the save.
11665 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11666 Test seh->after_prologue to disregard the epilogue.
11667 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11668 (i386_pe_end_cold_function): New function.
11670 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11673 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11674 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11677 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11679 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11680 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11682 PR middle-end/84723
11683 * multiple_target.c: Include tree-inline.h and intl.h.
11684 (expand_target_clones): Diagnose and fail if node->definition and
11685 !tree_versionable_function_p (node->decl).
11687 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11689 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11691 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11692 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11693 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11695 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11698 * combine.c (try_combine): Use reg_or_subregno instead of handling
11699 just paradoxical SUBREGs and REGs.
11701 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11703 * config/arc/arc.c (arc_finalize_pic): Remove function.
11704 (arc_must_save_register): We use single base PIC register, remove
11705 checks to save/restore the PIC register.
11706 (arc_expand_prologue): Likewise.
11707 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11709 (arc_verify_short): Likewise.
11710 (arc_attr_type): Likewise.
11711 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11712 (walk_stores): Likewise.
11713 (arc_address_cost): Make it static.
11714 (arc_verify_short): Likewise.
11715 (branch_dest): Likewise.
11716 (arc_attr_type): Likewise.
11717 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11718 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11719 (arc_final_prescan_insn): Remove inserting the nops due to
11720 hardware hazards. It is done in reorg step.
11721 (insn_length_variant_t): Remove.
11722 (insn_length_parameters_t): Likewise.
11723 (arc_insn_length_parameters): Likewise.
11724 (arc_get_insn_variants): Likewise.
11725 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11727 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11729 PR inline-asm/84683
11730 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11733 PR tree-optimization/84687
11734 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11736 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11738 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11740 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11741 Rename to ppc_speculation_barrier.
11742 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11743 __builtin_ppc_speculation_barrier.
11745 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11748 * combine.c (combine_simplify_rtx): Don't try to simplify if
11749 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11752 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11754 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11755 to 32 bytes when compiling for POWER9.
11757 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11760 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11761 regparm >= 3 with no arg reg available also for calls with
11762 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11765 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11767 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11769 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11772 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11774 2018-03-05 Richard Biener <rguenther@suse.de>
11776 PR tree-optimization/84486
11777 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11778 When inserting a __builtin_assume_aligned call set the LHS
11779 SSA name alignment info accordingly.
11781 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11783 PR tree-optimization/84114
11784 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11785 Avoid reassociation of FLOAT_MODE addition.
11787 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11789 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11790 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11791 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11792 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11793 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11794 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11795 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11797 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11798 __builtin_ia32_wbinvd): New builtins.
11799 (SPECIAL_ARGS2): New.
11800 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11801 (SPECIAL_ARGS2): New.
11802 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11803 (ix86_valid_target_attribute_inner_p): Ditto.
11804 (ix86_init_mmx_sse_builtins): Add special_args2.
11805 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11806 TARGET_WBNOINVD_P): New.
11807 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11808 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11809 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11810 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11811 * config/i386/pconfigintrin.h: New file.
11812 * config/i386/wbnoinvdintrin.h: Ditto.
11813 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11815 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11817 2018-03-05 Richard Biener <rguenther@suse.de>
11819 PR tree-optimization/84670
11820 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11822 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11823 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11824 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11825 doesn't grow if all successors (recursively) were visited at least
11828 2018-03-05 Richard Biener <rguenther@suse.de>
11830 PR tree-optimization/84650
11831 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11832 if executed in the loop pipeline.
11834 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11836 * doc/configfiles.texi (Configuration Files): Move info about
11837 conditionalizing $target-protos.h to...
11838 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11839 differs from $target-protos.h.
11841 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11842 Chung-Ju Wu <jasonwucj@gmail.com>
11844 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11845 * config/nds32/nds32-multiple.md (setmemsi): Define.
11846 * config/nds32/nds32-memory-manipulation.c
11847 (nds32_gen_dup_4_byte_to_word_value): New.
11848 (emit_setmem_word_loop): New.
11849 (emit_setmem_byte_loop): New.
11850 (nds32_expand_setmem_loop): New.
11851 (nds32_expand_setmem_loop_v3m): New.
11852 (nds32_expand_setmem_unroll): New.
11853 (nds32_expand_setmem): New.
11855 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11856 Chung-Ju Wu <jasonwucj@gmail.com>
11858 * config/nds32/nds32-memory-manipulation.c
11859 (nds32_emit_load_store): New.
11860 (nds32_emit_post_inc_load_store): New.
11861 (nds32_emit_mem_move): New.
11862 (nds32_emit_mem_move_block): New.
11863 (nds32_expand_movmemsi_loop_unknown_size): New.
11864 (nds32_expand_movmemsi_loop_known_size): New.
11865 (nds32_expand_movmemsi_loop): New.
11866 (nds32_expand_movmemsi_unroll): New.
11867 (nds32_expand_movmemqi): Rename ...
11868 (nds32_expand_movmemsi): ... to this.
11869 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11870 (movmemsi): ... to this.
11871 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11872 (nds32_expand_movmemsi): ... to this.
11874 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11875 Monk Chiang <sh.chiang04@gmail.com>
11876 Chung-Ju Wu <jasonwucj@gmail.com>
11878 * config/nds32/nds32-protos.h
11879 (nds32_expand_load_multiple): New arguments.
11880 (nds32_expand_store_multiple): Ditto.
11881 (nds32_valid_multiple_load_store): Rename ...
11882 (nds32_valid_multiple_load_store_p): ... to this.
11883 * config/nds32/nds32-memory-manipulation.c
11884 (nds32_expand_load_multiple): Refine implementation.
11885 (nds32_expand_store_multiple): Ditto.
11886 * config/nds32/nds32-multiple.md
11887 (load_multiple): Update nds32_expand_load_multiple interface.
11888 (store_multiple): Update nds32_expand_store_multiple interface.
11889 * config/nds32/nds32-predicates.c
11890 (nds32_valid_multiple_load_store): Rename ...
11891 (nds32_valid_multiple_load_store_p): ... to this and refine
11893 * config/nds32/predicates.md
11894 (nds32_load_multiple_and_update_address_operation): New predicate.
11895 (nds32_store_multiple_and_update_address_operation): New predicate.
11897 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11898 Chung-Ju Wu <jasonwucj@gmail.com>
11900 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11901 (combo): New attribute.
11902 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11904 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11906 * config/nds32/nds32.opt: Change -mcmodel= default value.
11908 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11909 Monk Chiang <sh.chiang04@gmail.com>
11910 Chung-Ju Wu <jasonwucj@gmail.com>
11912 * config/nds32/constants.md (unspec_element): New enum.
11913 * config/nds32/constraints.md (Umw): New constraint.
11914 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11915 * config/nds32/nds32-intrinsic.md: Likewise.
11916 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11917 (nds32_valid_smw_lwm_base_p): New.
11918 (nds32_output_smw_single_word): New.
11919 (nds32_output_lmw_single_word): New.
11920 (nds32_expand_unaligned_load): New.
11921 (nds32_expand_unaligned_store): New.
11922 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11923 (nds32_output_smw_single_word): Declare.
11924 (nds32_output_lmw_single_word): Declare.
11925 (nds32_expand_unaligned_load): Declare.
11926 (nds32_expand_unaligned_store): Declare.
11927 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11928 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11929 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11930 NDS32_BUILTIN_UASTORE_DW.
11931 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11934 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11935 Kito Cheng <kito.cheng@gmail.com>
11936 Chung-Ju Wu <jasonwucj@gmail.com>
11938 * config/nds32/nds32-intrinsic.c
11939 (nds32_expand_builtin_null_ftype_reg): Delete.
11940 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11941 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11942 (nds32_read_argument): New.
11943 (nds32_legitimize_target): Ditto.
11944 (nds32_legitimize_argument): Ditto.
11945 (nds32_check_constant_argument): Ditto.
11946 (nds32_expand_unop_builtin): Ditto.
11947 (nds32_expand_unopimm_builtin): Ditto.
11948 (nds32_expand_binop_builtin): Ditto.
11949 (nds32_builtin_decl_impl): Ditto.
11950 (builtin_description): Ditto.
11951 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11952 (nds32_init_builtins_impl): Ditto.
11953 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11954 (nds32_builtin_decl): New.
11955 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11956 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11958 2018-03-02 Jeff Law <law@redhat.com>
11960 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11961 (redundant_insn, fill_simple_delay_slots): Likewise.
11962 (fill_slots_from_thread): Likewise.
11963 * resource.c (mark_referenced_resources): Likewise.
11964 (mark_set_resources, find_dead_or_set_registers): Likewise.
11966 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11968 * substring-locations.h (format_warning_va): Formatting fix for
11969 ATTRIBUTE_GCC_DIAG.
11970 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11972 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11973 * substring-locations.c: Include intl.h.
11974 (format_warning_va): Turned into small wrapper around
11975 format_warning_n_va, renamed to ...
11976 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11977 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11979 (format_warning_at_substring_n): New function.
11980 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11981 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11982 format_warning_at_substring with just a shorter name instead of
11983 const function pointer.
11984 (fmtwarn_n): New function.
11985 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11986 appropriate, get rid of all the fmtstr temporaries, move conditionals
11987 with G_() wrapped string literals directly into fmtwarn arguments,
11988 cast dir.len to (int), formatting fixes.
11990 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11992 * doc/invoke.texi: Remove "Cilk Plus" references.
11994 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11995 Richard Biener <rguenther@suse.de>
11998 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11999 for error or warning attributes if CALL_FROM_THUNK_P is set.
12002 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12005 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12006 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12009 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12010 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12012 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12013 instead of -1U in last predictors element's probability member.
12015 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12018 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12019 arguments if they are comparable.
12021 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12023 PR tree-optimization/84634
12024 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12025 masks and masked_loop_p with a single loop_masks, making sure it's
12026 null for bb vectorization.
12028 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12030 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12031 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12032 loop->force_vectorize to check whether there is no alias.
12034 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12037 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12039 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12041 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12042 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12043 instead of a loop around prev_real_insn.
12044 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12047 PR inline-asm/84625
12048 * config/i386/i386.c (ix86_print_operand): Use conditional
12049 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12052 2018-03-02 Richard Biener <rguenther@suse.de>
12054 PR tree-optimization/84427
12055 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12056 (bitmap_set_subtract_values): Rewrite to handle multiple
12059 (prune_clobbered_mems): Likewise.
12060 (phi_translate): Take edge instead of pred/phiblock.
12061 (phi_translate_1): Likewise.
12062 (phi_translate_set): Likewise. Insert all translated
12063 exprs for a value into the set, keeping possibly multiple
12064 expressions per value.
12065 (compute_antic_aux): Adjust for phi_translate changes.
12066 When intersecting union the expressions and prune those
12067 not in the final value set, keeping possibly multiple
12068 expressions per value. Do not use value-insertion
12069 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12070 all expressions. Add verification that the value-sets
12071 only shrink during iteration.
12072 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12073 (do_pre_regular_insertion): Likewise.
12074 (do_pre_partial_partial_insertion): Likewise.
12076 2018-03-02 Richard Biener <rguenther@suse.de>
12079 * config/darwin.c (saved_debug_info_level): New static global.
12080 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12081 (darwin_asm_lto_end): Restore debug info generation settings.
12083 2018-03-01 Martin Liska <mliska@suse.cz>
12086 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12087 volatile arguments.
12089 2018-03-01 Richard Biener <rguenther@suse.de>
12092 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12093 type annotation with LTO when debug was disabled at compile-time.
12095 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12097 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12099 (mips_final_postscan_insn): Likewise.
12101 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12103 PR rtl-optimization/84528
12104 * alias.c (init_alias_target): Add commentary.
12105 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12106 a unique base value if the frame pointer is not eliminated
12107 to the stack pointer.
12109 2018-03-01 Tom de Vries <tom@codesourcery.com>
12111 PR rtl-optimization/83327
12112 * lra-int.h (hard_regs_spilled_into): Declare.
12113 * lra.c (hard_regs_spilled_into): Define.
12114 (init_reg_info): Init hard_regs_spilled_into.
12115 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12116 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12117 (process_bb_lives): Handle hard_regs_spilled_into.
12118 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12119 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12121 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12123 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12124 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12125 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12126 * config/rs6000/aix72.h: New file.
12128 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12130 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12131 instead of warning_at with conditional singular and plural messages
12135 * stor-layout.c (update_alignment_for_field): For
12136 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12137 && !DECL_PACKED (field), do the alignment update, just use
12138 only desired_align instead of MAX (type_align, desired_align)
12140 (place_field): Don't do known_align < desired_align handling
12141 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12142 is non-NULL, instead do it after rli->prev_field handling and
12143 only if not within a bitfield word. For DECL_PACKED (field)
12144 use type_align of BITS_PER_UNIT.
12146 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12148 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12149 superfluous parentheses and trailing spaces.
12151 2018-02-28 Richard Biener <rguenther@suse.de>
12153 PR tree-optimization/84584
12154 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12155 SCoPs with fake exit edge.
12157 2018-02-28 Martin Liska <mliska@suse.cz>
12160 * timevar.c (timer::print): Fix format to properly print 100%
12163 2018-02-28 Richard Biener <rguenther@suse.de>
12165 PR middle-end/84607
12166 * genmatch.c (capture_info::walk_match): Do not mark
12167 captured expressions without operands as expr_p given
12168 they act more like predicates and should be subject to
12169 "lost tail" side-effect preserving.
12171 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12173 PR rtl-optimization/81611
12174 * auto-inc-dec.c (attempt_change): Move dead note from
12175 mem_insn if it's the next use of regno
12176 (find_address): Take address use of reg holding
12177 non-incremented value. Add parm to limit search to the named
12179 (merge_in_block): Attempt to use a mem insn that is the next
12180 use of the original regno.
12182 2018-02-27 Martin Sebor <msebor@redhat.com>
12185 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12186 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12188 2018-02-27 Martin Sebor <msebor@redhat.com>
12190 PR translation/84207
12191 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12192 n argument to unsigned HOST_WIDE_INT.
12193 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12194 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12195 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12196 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12198 2018-02-27 Richard Biener <rguenther@suse.de>
12200 PR tree-optimization/84512
12201 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12202 Do not use the estimate returned from record_stmt_cost for
12203 the scalar iteration cost but sum properly using add_stmt_cost.
12205 2018-02-27 Richard Biener <rguenther@suse.de>
12207 PR tree-optimization/84466
12208 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12209 Adjust last change to less strictly validate use operands.
12211 2018-02-27 Martin Liska <mliska@suse.cz>
12213 PR gcov-profile/84548
12214 * gcov.c (process_file): Allow partial overlap and consider it
12215 also as group functions.
12216 (output_lines): Properly calculate range of lines for a group.
12218 2018-02-27 Martin Liska <mliska@suse.cz>
12220 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12221 'ggc' suffixes. Change first column width.
12222 (timer::print): Fix formatting of the column.
12224 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12226 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12227 preserve inline entry blocks for the sake of debug inline
12228 entry point markers alone.
12229 (remove_unused_locals): Suggest in comments a better place to
12230 force the preservation of inline entry blocks that are
12231 otherwise unused, but do not preserve them.
12233 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12235 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12237 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12240 * config/i386/constraints.md (Bs): Replace
12241 ix86_indirect_branch_register with
12242 TARGET_INDIRECT_BRANCH_REGISTER.
12244 * config/i386/i386.md (indirect_jump): Likewise.
12245 (tablejump): Likewise.
12246 (*sibcall_memory): Likewise.
12247 (*sibcall_value_memory): Likewise.
12248 Peepholes of indirect call and jump via memory: Likewise.
12249 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12250 (*sibcall_value_GOT_32): Likewise.
12251 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12252 (GOT_memory_operand): Likewise.
12253 (call_insn_operand): Likewise.
12254 (sibcall_insn_operand): Likewise.
12255 (GOT32_symbol_operand): Likewise.
12256 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12258 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12260 PR rtl-optimization/83496
12261 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12262 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12263 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12264 redundant insn, if any.
12265 (relax_delay_slots): Likewise.
12266 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12268 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12270 PR tree-optimization/83965
12271 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12272 that grouped statements are part of a reduction chain. Return
12273 true if the statement is not marked as a reduction itself but
12274 is part of a group.
12275 (vect_recog_dot_prod_pattern): Don't check whether the statement
12276 is part of a group here.
12277 (vect_recog_sad_pattern): Likewise.
12278 (vect_recog_widen_sum_pattern): Likewise.
12280 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12283 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12285 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12288 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12290 (ix86_output_indirect_function_return): New prototype.
12291 (ix86_split_simple_return_pop_internal): Likewise.
12292 * config/i386/i386.c (indirect_return_via_cx): New.
12293 (indirect_return_via_cx_bnd): Likewise.
12294 (indirect_thunk_name): Handle return va CX_REG.
12295 (output_indirect_thunk_function): Create alias for
12296 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12297 (ix86_output_indirect_jmp): Remove the bool argument.
12298 (ix86_output_indirect_function_return): New function.
12299 (ix86_split_simple_return_pop_internal): Likewise.
12300 * config/i386/i386.md (*indirect_jump): Don't pass false
12301 to ix86_output_indirect_jmp.
12302 (*tablejump_1): Likewise.
12303 (simple_return_pop_internal): Change it to define_insn_and_split.
12304 Call ix86_split_simple_return_pop_internal to split it for
12305 -mfunction-return=.
12306 (simple_return_indirect_internal): Call
12307 ix86_output_indirect_function_return instead of
12308 ix86_output_indirect_jmp.
12310 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12313 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12314 memset and value initialization afterwards.
12316 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12318 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12320 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12323 * common/config/aarch64/aarch64-common.c
12324 (aarch_option_optimization_table[]): Switch
12325 off fomit-frame-pointer
12327 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12328 Chung-Ju Wu <jasonwucj@gmail.com>
12330 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12332 (store_multiple): Ditto.
12334 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12336 * config.gcc: Add --with-cpu support for nds32 target.
12337 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12338 * config/nds32/nds32.opt: Add -mcpu= option.
12340 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12342 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12343 isel=yes): Warn for these deprecated options.
12345 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12347 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12348 ISA_2_5_MASKS_EMBEDDED.
12350 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12352 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12353 p->max as pointers rather than using iterative_hash_expr.
12355 2018-02-23 Carl Love <cel@us.ibm.com>
12357 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12358 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12360 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12361 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12362 P8V_BUILTIN_VEC_VUNSIGNED2.
12364 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12367 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12368 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12370 * lra-constraints.c (curr_insn_transform): Set up
12371 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12373 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12374 * lra-eliminations.c (spill_pseudos): Ditto.
12375 (process_insn_for_elimination): Ditto.
12376 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12377 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12378 LRA_NON_CLOBBERED_ALT.
12380 2018-02-22 Martin Sebor <msebor@redhat.com>
12382 PR tree-optimization/84480
12383 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12384 to maybe_diag_stxncpy_trunc. Call it.
12385 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12386 from gimple_fold_builtin_strcpy. Print inlining stack.
12387 (handle_builtin_stxncpy): Print inlining stack.
12388 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12390 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12393 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12394 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12395 and -fcheck-pointer-bounds are used together.
12396 (indirect_thunk_prefix): New enum.
12397 (indirect_thunk_need_prefix): New function.
12398 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12399 "_nt" instead of "_bnd" for NOTRACK prefix.
12400 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12401 (output_indirect_thunk_function): Likewise.
12403 (ix86_code_end): Update output_indirect_thunk_function calls.
12404 (ix86_output_indirect_branch_via_reg): Replace
12405 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12406 (ix86_output_indirect_branch_via_push): Likewise.
12407 (ix86_output_function_return): Likewise.
12408 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12409 incompatible with -fcf-protection=branch and
12410 -fcheck-pointer-bounds.
12412 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12415 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12416 Change gcc_assert call to output_operand_lossage.
12418 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12420 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12422 2018-02-22 DJ Delorie <dj@redhat.com>
12423 Sebastian Perta <sebastian.perta@renesas.com>
12424 Oleg Endo <olegendo@gcc.gnu.org>
12426 * config/rx/rx.c (rx_rtx_costs): New function.
12427 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12429 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12431 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12433 2018-02-22 Martin Liska <mliska@suse.cz>
12436 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12437 Add "native" as a possible value.
12439 2018-02-22 Martin Liska <mliska@suse.cz>
12442 * config/i386/i386.c (ix86_option_override_internal):
12443 Add "native" as a possible value for -march and -mtune.
12445 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12448 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12449 to all type variants.
12451 PR tree-optimization/84503
12452 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12453 width as info->bitpos + info->bitsize - start.
12454 (merged_store_group::merge_overlapping): Simplify width computation.
12455 (check_no_overlap): New function.
12456 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12457 start + width and last_order of the group, fail if check_no_overlap
12459 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12460 to group if check_no_overlap fails.
12462 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12464 * config/rs6000/altivec.md: Delete contraint arguments to
12465 define_expand, define_split, and define_peephole2, and in
12466 define_insn_and_split if always unused.
12467 * config/rs6000/darwin.md: Ditto.
12468 * config/rs6000/dfp.md: Ditto.
12469 * config/rs6000/rs6000.md: Ditto.
12470 * config/rs6000/sync.md: Ditto.
12471 * config/rs6000/vector.md: Ditto.
12472 * config/rs6000/vsx.md: Ditto.
12474 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12476 * config/rs6000/altivec.md: Write output control strings as braced
12477 blocks instead of double-quoted strings.
12478 * config/rs6000/darwin.md: Ditto.
12479 * config/rs6000/rs6000.md: Ditto.
12480 * config/rs6000/vector.md: Ditto.
12481 * config/rs6000/vsx.md: Ditto.
12483 2018-02-21 Jason Merrill <jason@redhat.com>
12485 PR c++/84314 - ICE with templates and fastcall attribute.
12486 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12488 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12490 * ipa-cp.c (determine_versionability): Fix comment typos.
12492 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12495 * ipa-cp.c (determine_versionability): Do not version functions caling
12498 2018-02-21 Martin Liska <mliska@suse.cz>
12501 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12502 Add "native" as a possible value.
12503 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12504 the macro when native cpu detection is available.
12506 2018-02-21 Martin Liska <mliska@suse.cz>
12509 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12510 Add "native" as a possible value.
12511 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12512 when native cpu detection is available.
12514 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12515 Martin Sebor <msebor@redhat.com>
12517 PR tree-optimization/84478
12518 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12520 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12521 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12522 support which is conservatively correct, for 2 only stay conservative
12523 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12524 argument to the 2 argument get_range_strlen, adjust 6 arg
12525 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12527 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12528 (gimple_fold_builtin_strlen): Pass true as last argument to
12531 2018-02-20 Martin Sebor <msebor@redhat.com>
12533 PR middle-end/84095
12534 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12535 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12536 (builtin_memref::builtin_memref): Factor out parts into
12537 set_base_and_offset and call it.
12539 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12541 PR middle-end/84406
12542 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12543 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12544 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12545 search at the associated MODE_INT.
12547 2018-02-20 Jeff Law <law@redhat.com>
12549 PR middle-end/82123
12550 PR tree-optimization/81592
12551 PR middle-end/79257
12552 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12553 for range data rather than using global data.
12554 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12555 range data rather than using global data.
12556 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12557 pass it to children as needed.
12558 (struct directive::fmtresult): Similarly.
12559 (struct directive::set_width): Similarly.
12560 (struct directive::set_precision): Similarly.
12561 (format_integer, format_directive, parse_directive): Similarly.
12562 (format_none): Accept unnamed vr_values parameter.
12563 (format_percent, format_floating, format_character): Similarly.
12564 (format_string, format_plain): Similarly.
12565 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12566 the EVRP range analyzer for range data rather than using global data.
12567 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12568 gimple-ssa-evrp-analyze.h
12569 (class sprintf_dom_walker): Add after_dom_children member function.
12570 Add evrp_range_analyzer member.
12571 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12572 range analyzer as needed.
12573 (sprintf_dom_walker::after_dom_children): New member function.
12574 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12576 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12577 (evrp_range_analyzer::pop_to_marker): Likewise.
12579 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12581 PR tree-optimization/84419
12582 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12583 with the required type if its current type is compatible but
12586 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12588 PR middle-end/82004
12589 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12590 after vectorization.
12592 2018-02-20 Martin Liska <mliska@suse.cz>
12595 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12596 possible values if we don't have a hint.
12598 2018-02-20 Martin Liska <mliska@suse.cz>
12602 * final.c (shorten_branches): Build align_tab array with one
12604 * opts.c (finish_options): Add alignment option limit check.
12605 (MAX_CODE_ALIGN): Likewise.
12606 (MAX_CODE_ALIGN_VALUE): Likewise.
12607 * doc/invoke.texi: Document maximum allowed option value for
12608 all -falign-* options.
12610 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12613 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12614 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12615 * var-tracking.c (emit_note_insn_var_location): Remove all references
12616 to NOTE_INSN_CALL_ARG_LOCATION.
12617 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12618 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12619 Use copy_rtx_if_shared.
12620 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12621 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12622 (dwarf2out_var_location): Remove handling of
12623 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12625 * final.c (final_scan_insn): Remove all references to
12626 NOTE_INSN_CALL_ARG_LOCATION.
12627 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12628 before dumping final insns.
12629 * except.c (emit_note_eh_region_end): Remove all references to
12630 NOTE_INSN_CALL_ARG_LOCATION.
12631 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12632 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12633 * config/arc/arc.c (hwloop_optimize): Likewise.
12634 * config/arm/arm.c (create_fix_barrier): Likewise.
12635 * config/s390/s390.c (s390_chunkify_start): Likewise.
12636 * config/sh/sh.c (find_barrier): Likewise.
12637 * config/i386/i386.c (rest_of_insert_endbranch,
12638 ix86_seh_fixup_eh_fallthru): Likewise.
12639 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12640 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12641 * config/frv/frv.c (frv_function_prologue): Likewise.
12642 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12644 (note_outside_basic_block_p): Remove all references to
12645 NOTE_INSN_CALL_ARG_LOCATION.
12646 * gengtype.c (adjust_field_rtx_def): Likewise.
12647 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12649 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12650 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12653 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12656 PR tree-optimization/84452
12657 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12658 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12661 2018-02-19 Martin Liska <mliska@suse.cz>
12664 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12666 2018-02-19 Martin Liska <mliska@suse.cz>
12667 Richard Sandiford <richard.sandiford@linaro.org>
12669 PR tree-optimization/82491
12670 * gimple-fold.c (get_base_constructor): Make earlier bail out
12673 2018-02-19 Carl Love <cel@us.ibm.com>
12675 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12676 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12678 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12679 P8V_BUILTIN_VEC_NEG.
12681 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12683 * config/rl78/rl78.md (movdf): New define expand.
12685 2018-02-19 Martin Liska <mliska@suse.cz>
12688 * doc/invoke.texi: Fix typo.
12689 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12691 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12693 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12694 handle rs6000_single_float and rs6000_double_float specially for
12697 2018-02-16 Jeff Law <law@redhat.com>
12699 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12700 (pop_regs): Use it.
12702 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12705 * ipa-inline.c (inline_small_functions): Fix a typo.
12707 2018-02-16 Nathan Sidwell <nathan@acm.org>
12709 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12711 2018-02-16 Carl Love <cel@us.ibm.com>
12713 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12714 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12715 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12716 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12717 expansion to P8V_BUILTIN_VEC_FLOAT2.
12719 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12721 PR rtl-optimization/70023
12722 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12723 src_regno into account.
12725 2018-02-16 Carl Love <cel@us.ibm.com>
12727 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12728 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12729 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12730 * config/rs6000/rs6000.c: Remove case statements for
12731 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12732 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12733 and P9V_BUILTIN_VEC_VINSERT4B.
12734 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12735 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12736 * config/rs6000/vsx.md:
12737 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12740 2018-02-16 Carl Love <cel@us.ibm.com>
12742 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12744 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12746 * config/rs6000/rs6000-c.c: Add the definitions for
12747 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12748 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12749 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12750 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12751 definition for insert4b and define insn *insert3b_internal.
12752 * doc/extend.texi: Add documentation for vec_extract4b.
12754 2018-02-16 Nathan Sidwell <nathan@acm.org>
12756 * doc/extend.texi (Backwards Compatibility): Mention friend
12757 injection. Note for-scope is deprecated.
12758 * doc/invoke.texi (-ffriend-injection): Deprecate.
12760 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12762 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12763 that moved to I2, also allow destinations that are a paradoxical
12764 subreg (instead of a normal reg).
12766 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12769 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12772 2018-02-16 Richard Biener <rguenther@suse.de>
12774 PR tree-optimization/84037
12775 PR tree-optimization/84016
12777 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12778 Adjust vec_construct for the fact we need additional higher latency
12779 128bit inserts for AVX256 and AVX512 vector builds.
12780 (ix86_add_stmt_cost): Scale vector construction cost for
12783 2018-02-16 Richard Biener <rguenther@suse.de>
12785 PR tree-optimization/84417
12786 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12787 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12788 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12790 2018-02-16 Martin Liska <mliska@suse.cz>
12793 * internal-fn.def (ASAN_CHECK): Set proper flags.
12794 (ASAN_MARK): Likewise.
12796 2018-02-16 Julia Koval <julia.koval@intel.com>
12798 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12799 from PTA_CANNONLAKE.
12801 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12804 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12805 Use ++iter rather than iter++ for std::list iterators.
12806 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12807 defer deleting them until all nodes in the forest are processed. Do
12808 free even leaf nodes. Change to_process into auto_vec.
12811 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12812 * vec.h (vec_default_construct): Use memset instead of placement new
12813 if BROKEN_VALUE_INITIALIZATION is defined.
12814 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12815 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12818 PR rtl-optimization/83723
12819 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12820 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12821 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12823 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12825 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12827 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12829 PR rtl-optimization/81443
12830 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12831 from inner REGs to paradoxical SUBREGs.
12833 2018-02-16 Richard Biener <rguenther@suse.de>
12835 PR tree-optimization/84399
12836 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12837 For operands we can analyze at their definition make sure we can
12838 analyze them at each use as well.
12840 2018-02-16 Richard Biener <rguenther@suse.de>
12842 PR tree-optimization/84190
12843 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12844 volatile accesses if the decl isn't volatile.
12846 2018-02-15 Jason Merrill <jason@redhat.com>
12848 PR c++/84314 - ICE with templates and fastcall attribute.
12849 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12850 TYPE_CANONICAL on an existing type.
12852 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12854 PR tree-optimization/84383
12855 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12856 dstoff nor call operand_equal_p if dstbase is NULL.
12858 PR tree-optimization/84334
12859 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12860 also a CONSTANT_CLASS_P, punt.
12862 2018-02-14 Jim Wilson <jimw@sifive.com>
12864 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12865 first SMALL_OPERAND check. New local min_second_step. Move assert
12866 to where locals are set. Add TARGET_RVC support.
12867 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12869 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12871 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12873 2018-02-14 Martin Sebor <msebor@redhat.com>
12875 PR tree-optimization/83698
12876 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12877 arrays constrain the offset range to their bounds.
12878 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12879 (builtin_access::overlap): Avoid setting the size of overlap if it's
12881 (maybe_diag_overlap): Also consider arrays when deciding what values
12882 of offsets to include in diagnostics.
12884 2018-02-14 Martin Sebor <msebor@redhat.com>
12887 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12888 that correspond to the kind of a declaration.
12890 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12893 * config/pa/pa.md: Load address of PIC label using the linkage table
12894 if the label is nonlocal.
12896 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12898 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12899 warning message if user requests -maltivec=be.
12900 * doc/invoke.texi: Document deprecation of -maltivec=be.
12902 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12905 * config/rs6000/rs6000-c.c: Update definitions for
12906 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12907 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12909 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12912 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12913 add _get_ssp intrinsics. Remove argument from
12914 __builtin_ia32_rdssp[d|q].
12915 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12916 * config/i386/i386-builtin.def: Remove argument from
12917 __builtin_ia32_rdssp[d|q].
12918 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12919 ix86_expand_special_args_builtin for _rdssp[d|q].
12920 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12921 Clear register before usage.
12922 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12923 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12925 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12927 PR tree-optimization/84357
12928 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12929 operand 1 of an ARRAY_REF too.
12931 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12934 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12935 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12937 (set_of_reg): New struct.
12938 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12939 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12940 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12942 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12943 Split into bitclr, bitset, bitinvert patterns if appropriate.
12944 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12945 use rx_fuse_in_memory_bitop.
12946 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12947 to named insn, correct maximum insn length.
12949 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12952 * machmode.def: Define a complex mode for PARTIAL_INT.
12953 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12955 * doc/rtl.texi: Document CSPImode.
12956 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12958 (msp430_hard_regno_nregs_with_padding): Likewise.
12960 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12963 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12965 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12967 PR rtl-optimization/84169
12968 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12969 we generated a parallel as new i3 and we split that to new i2 and i3
12970 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12971 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12972 those to i2, not i1. Partially rewrite this scan code.
12974 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12977 * stor-layout.c (place_field): For variable length fields, adjust
12978 offset_align afterwards not just based on the field's alignment,
12979 but also on the size.
12981 PR middle-end/84309
12982 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12983 of exps and logs in the use_exp2 case.
12985 2018-02-13 Jeff Law <law@redhat.com>
12987 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12988 entry for "vector".
12990 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12993 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12997 * common.opt (gas-loc-support, gas-locview-support): New.
12998 (ginline-points, ginternal-reset-location-views): New.
12999 * doc/invoke.texi: Document them. Use @itemx where intended.
13000 (gvariable-location-views): Adjust.
13001 * target.def (reset_location_view): New.
13002 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13003 (TARGET_RESET_LOCATION_VIEW): New.
13004 * doc/tm.texi: Rebuilt.
13005 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13006 (dwarf2out_default_as_locview_support): New.
13007 (output_asm_line_debug_info): Use option variables.
13008 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13009 (output_loc_list): Likewise.
13010 (add_high_low_attributes): Check option variables.
13011 Don't output entry view attribute in strict mode.
13012 (gen_inlined_subroutine_die): Check option variables.
13013 (dwarf2out_inline_entry): Likewise.
13014 (init_sections_and_labels): Likewise.
13015 (dwarf2out_early_finish): Likewise.
13016 (maybe_reset_location_view): New, from...
13017 (dwarf2out_var_location): ... here. Call it.
13018 * debug.h (dwarf2out_default_as_loc_support): Declare.
13019 (dwarf2out_default_as_locview_support): Declare.
13020 * hooks.c (hook_int_rtx_insn_0): New.
13021 * hooks.h (hook_int_rtx_insn_0): Declare.
13022 * toplev.c (process_options): Take -gas-loc-support and
13023 -gas-locview-support from dwarf2out. Enable
13024 -gvariable-location-views by default only with locview
13025 assembler support. Enable -ginternal-reset-location-views by
13026 default only if the target defines the corresponding hook.
13027 Enable -ginline-points by default if location views are
13028 enabled; force it disabled if statement frontiers are
13030 * tree-inline.c (expand_call_inline): Check option variables.
13031 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13033 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13035 PR tree-optimization/84321
13036 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13037 handling. Also check whether the anti-range contains any values
13038 that satisfy the mask; switch to a VR_RANGE if not.
13040 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13043 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13045 2018-02-13 Martin Jambor <mjambor@suse.cz>
13048 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13049 of call statements, also set location of a load to a temporary.
13051 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13053 * config/rl78/rl78.c (add_vector_labels): New function.
13054 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13055 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13056 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13057 which checks that no arguments are passed.
13058 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13059 * doc/extend.texi: Documentation for the new attribute.
13061 2018-02-13 Andreas Schwab <schwab@suse.de>
13063 * config/riscv/linux.h (CPP_SPEC): Define.
13065 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13068 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13069 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13070 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13071 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13072 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13073 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13074 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13075 temporarily for AES and PCLMUL builtins.
13077 PR tree-optimization/84339
13078 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13079 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13082 PR middle-end/84309
13083 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13084 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13085 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13087 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13089 * omp-simd-clone.h: New file.
13090 * omp-simd-clone.c: Include omp-simd-clone.h.
13091 (expand_simd_clones): No longer static.
13092 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13093 cgraph.h and omp-simd-clone.h.
13094 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13095 (vect_recog_widen_shift_pattern): Formatting fix.
13096 (vect_pattern_recog_1): Don't check optab for calls.
13099 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13100 operands[2] into a REG before using gen_lowpart on it.
13102 2018-02-12 Jeff Law <law@redhat.com>
13105 * config/sh/sh.c (find_barrier): Consider a sibling call
13108 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13109 successfully back substituting a reg.
13111 2018-02-12 Richard Biener <rguenther@suse.de>
13113 PR tree-optimization/84037
13114 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13115 parameter, move visited init to caller.
13116 (vect_slp_analyze_operations): Separate cost from validity
13117 check, initialize visited once for all instances.
13118 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13120 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13122 (vect_model_promotion_demotion_cost): Likewise.
13123 (vectorizable_bswap): Guard cost modeling with !slp_node
13124 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13126 (vectorizable_call): Likewise.
13127 (vectorizable_conversion): Likewise.
13128 (vectorizable_assignment): Likewise.
13129 (vectorizable_shift): Likewise.
13130 (vectorizable_operation): Likewise.
13131 (vectorizable_store): Likewise.
13132 (vectorizable_load): Likewise.
13133 (vectorizable_condition): Likewise.
13134 (vectorizable_comparison): Likewise.
13136 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13139 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13140 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13141 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13143 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13145 PR middle-end/83665
13146 * params.def (inline-min-speedup): Increase from 8 to 15.
13147 (max-inline-insns-auto): Decrease from 40 to 30.
13148 * ipa-split.c (consider_split): Add some buffer for function to
13149 be considered inlining candidate.
13150 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13153 2018-02-12 Richard Biener <rguenther@suse.de>
13155 PR tree-optimization/84037
13156 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13157 matched stmts if we cannot swap the non-matched ones.
13159 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13161 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13162 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13163 _mm_maskz_scalef_round_ss): New intrinsics.
13164 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13165 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13166 __builtin_ia32_scalefss_round): Remove.
13167 (__builtin_ia32_scalefsd_mask_round,
13168 __builtin_ia32_scalefss_mask_round): New intrinsics.
13169 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13170 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13171 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13172 "<round_constraint>")): Changed to ...
13173 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13174 "<round_scalar_constraint>")): ... this.
13175 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13176 %0, %1, %2<round_op3>}"): Changed to ...
13177 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13178 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13179 %2<round_scalar_mask_op3>}"): ... this.
13180 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13182 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13184 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13185 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13186 (_mm_maskz_sqrt_round_ss): New intrinsics.
13187 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13188 (__builtin_ia32_sqrtsd_mask_round)
13189 (__builtin_ia32_sqrtss_mask_round): New builtins.
13190 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13191 (__builtin_ia32_sqrtss_round): Remove.
13192 (__builtin_ia32_sqrtsd_mask_round)
13193 (__builtin_ia32_sqrtss_mask_round): New builtins.
13194 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13195 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13196 ((match_operand:VF_128 1 "vector_operand"
13197 "xBm,<round_constraint>")): Changed to ...
13198 ((match_operand:VF_128 1 "vector_operand"
13199 "xBm,<round_scalar_constraint>")): ... this.
13200 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13201 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13202 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13203 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13204 %<iptr>1<round_scalar_mask_op3>}): ... this.
13205 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13206 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13208 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13211 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13212 Cast vec_cmpeq result to correct type.
13213 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13214 Cast vec_cmpgt result to correct type.
13216 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13218 * final.c (final_scan_insn_1): Renamed from...
13219 (final_scan_insn): ... this. New wrapper, to recover
13220 seen from the outermost call in recursive ones.
13221 * config/sparc/sparc.c (output_return): Drop seen from call.
13222 (output_sibcall): Likewise.
13223 * config/visium/visium.c (output_branch): Likewise.
13225 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13227 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13230 2018-02-10 Alan Modra <amodra@gmail.com>
13233 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13234 Specify LR as an input.
13236 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13239 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13240 remove_member_access_dummy_vars): New functions.
13241 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13242 lower_omp_1, execute_lower_omp): Use them.
13244 PR rtl-optimization/84308
13245 * shrink-wrap.c (spread_components): Release todo vector.
13247 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13249 PR rtl-optimization/57193
13250 * ira-color.c (struct allocno_color_data): Add member
13251 conflict_allocno_hard_prefs.
13252 (update_conflict_allocno_hard_prefs): New.
13253 (bucket_allocno_compare_func): Add a preference based on
13254 conflict_allocno_hard_prefs.
13255 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13256 (color_allocnos): Remove a dead code. Initiate
13257 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13259 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13262 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13263 constraint from =wa to wa. Avoid a subreg on the output operand,
13264 instead use a pseudo and subreg it in a move.
13265 (p9_xxbrd_<mode>): Changed to ...
13266 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13267 (p9_xxbrd_v2df): New expander.
13268 (p9_xxbrw_<mode>): Changed to ...
13269 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13270 (p9_xxbrw_v4sf): New expander.
13272 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13274 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13276 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13279 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13280 multiply in 32-bit mode.
13281 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13282 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13285 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13287 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13288 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13289 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13290 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13292 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13295 * dwarf2out.c (is_trivial_indirect_ref): New function.
13296 (dwarf2out_late_global_decl): Do not generate a location
13297 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13298 and that are not defined in the current unit.
13300 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13302 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13303 instead of a libcall for UNORDERED.
13305 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13308 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13309 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13311 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13314 * config/s390/s390.c (s390_set_current_function): Invoke
13315 s390_indirect_branch_settings also if fndecl didn't change.
13317 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13319 * config/rs6000/rs6000.md (blockage): Set length to zero.
13321 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13323 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13325 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13328 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13329 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13333 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13334 PARALLEL incoming that failed vt_get_decl_and_offset check.
13336 PR middle-end/84237
13337 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13338 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13340 (get_variable_section): For decls in named .bss* sections pass true as
13341 second argument to bss_initializer_p.
13343 2018-02-09 Marek Polacek <polacek@redhat.com>
13344 Jakub Jelinek <jakub@redhat.com>
13347 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13348 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13349 Sync some changes from cxx_fold_indirect_ref.
13351 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13353 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13355 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13356 (BLOCK_INLINE_ENTRY_LABEL): New.
13357 (dwarf2out_var_location): Disregard inline entry markers.
13358 (inline_entry_data): New struct.
13359 (inline_entry_data_hasher): New hashtable type.
13360 (inline_entry_data_hasher::hash): New.
13361 (inline_entry_data_hasher::equal): New.
13362 (inline_entry_data_table): New variable.
13363 (add_high_low_attributes): Add DW_AT_entry_pc and
13364 DW_AT_GNU_entry_view attributes if a pending entry is found
13365 in inline_entry_data_table. Add old entry_pc attribute only
13366 if debug nonbinding markers are disabled.
13367 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13368 markers are enabled.
13369 (block_within_block_p, dwarf2out_inline_entry): New.
13370 (dwarf2out_finish): Check that no entries remained in
13371 inline_entry_data_table.
13372 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13373 (final_scan_insn, notice_source_line): Likewise.
13374 (rest_of_clean_state): Skip inline entry markers.
13375 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13377 * gimple.c (gimple_build_debug_inline_entry): New.
13378 * gimple.h (enum gimple_debug_subcode): Add
13379 GIMPLE_DEBUG_INLINE_ENTRY.
13380 (gimple_build_debug_inline_entry): Declare.
13381 (gimple_debug_inline_entry_p): New.
13382 (gimple_debug_nonbind_marker_p): Adjust.
13383 * insn-notes.def (INLINE_ENTRY): New.
13384 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13385 inline entry marker notes.
13386 (print_insn): Likewise.
13387 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13388 (INSN_DEBUG_MARKER_KIND): Likewise.
13389 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13390 * tree-inline.c (expand_call_inline): Build and insert
13391 debug_inline_entry stmt.
13392 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13393 inline entry blocks early, if nonbind markers are enabled.
13394 (dump_scope_block): Dump fragment info.
13395 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13396 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13397 (gimple_build_debug_inline_entry): New.
13398 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13399 Enable/disable inline entry points too.
13400 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13401 (DEBUG_INSN): Describe inline entry markers.
13403 * common.opt (gvariable-location-views): New.
13404 (gvariable-location-views=incompat5): New.
13405 * config.in: Rebuilt.
13406 * configure: Rebuilt.
13407 * configure.ac: Test assembler for view support.
13408 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13409 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13410 * dwarf2out.c (var_loc_view): New typedef.
13411 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13412 (dwarf2out_locviews_in_attribute): New.
13413 (dwarf2out_locviews_in_loclist): New.
13414 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13415 (enum dw_line_info_opcode): Add LI_adv_address.
13416 (struct dw_line_info_table): Add view.
13417 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13418 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13419 (zero_view_p): New variable.
13420 (ZERO_VIEW_P): New macro.
13421 (output_asm_line_debug_info): New.
13422 (struct var_loc_node): Add view.
13423 (add_AT_view_list, AT_loc_list): New.
13424 (add_var_loc_to_decl): Add view param. Test it against last.
13425 (new_loc_list): Add view params. Record them.
13426 (AT_loc_list_ptr): Handle loc and view lists.
13427 (view_list_to_loc_list_val_node): New.
13428 (print_dw_val): Handle dw_val_class_view_list.
13429 (size_of_die): Likewise.
13430 (value_format): Likewise.
13431 (loc_list_has_views): New.
13432 (gen_llsym): Set vl_symbol too.
13433 (maybe_gen_llsym, skip_loc_list_entry): New.
13434 (dwarf2out_maybe_output_loclist_view_pair): New.
13435 (output_loc_list): Output view list or entries too.
13436 (output_view_list_offset): New.
13437 (output_die): Handle dw_val_class_view_list.
13438 (output_dwarf_version): New.
13439 (output_compilation_unit_header): Use it.
13440 (output_skeleton_debug_sections): Likewise.
13441 (output_rnglists, output_line_info): Likewise.
13442 (output_pubnames, output_aranges): Update version comments.
13443 (output_one_line_info_table): Output view numbers in asm comments.
13444 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13445 Call maybe_gen_llsym.
13446 (loc_list_from_tree_1): Adjust.
13447 (add_AT_location_description): Create view list attribute if
13448 needed, check it's absent otherwise.
13449 (convert_cfa_to_fb_loc_list): Adjust.
13450 (maybe_emit_file): Call output_asm_line_debug_info for test.
13451 (dwarf2out_var_location): Reset views as needed. Precompute
13452 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13453 attribute. Set view.
13454 (new_line_info_table): Reset next view.
13455 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13456 (dwarf2out_source_line): Likewise. Output view resets and labels to
13457 the assembler, or select appropriate line info opcodes.
13458 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13459 (optimize_string_length): Catch it. Adjust.
13460 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13461 dw_val_class_view_list, and remove it if no longer needed.
13462 (hash_loc_list): Hash view numbers.
13463 (loc_list_hasher::equal): Compare them.
13464 (optimize_location_lists): Check whether a view list symbol is
13465 needed, and whether the locview attribute is present, and
13466 whether they match. Remove the locview attribute if no longer
13468 (index_location_lists): Call skip_loc_list_entry for test.
13469 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13470 Use output_dwarf_version.
13471 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13472 (struct dw_val_node): Add val_view_list.
13473 * final.c (SEEN_NEXT_VIEW): New.
13474 (set_next_view_needed): New.
13475 (clear_next_view_needed): New.
13476 (maybe_output_next_view): New.
13477 (final_start_function): Rename to...
13478 (final_start_function_1): ... this. Take pointer to FIRST,
13479 add SEEN parameter. Emit param bindings in the initial view.
13480 (final_start_function): Reintroduce SEEN-less interface.
13481 (final): Rename to...
13482 (final_1): ... this. Take SEEN parameter. Output final pending
13483 next view at the end.
13484 (final): Reintroduce seen-less interface.
13485 (final_scan_insn): Output pending next view before switching
13486 sections or ending a block. Mark the next view as needed when
13487 outputting variable locations. Notify debug backend of section
13488 changes, and of location view changes.
13489 (rest_of_handle_final): Adjust.
13490 * toplev.c (process_options): Autodetect value for debug variable
13491 location views option. Warn on incompat5 without -gdwarf-5.
13492 * doc/invoke.texi (gvariable-location-views): New.
13493 (gvariable-location-views=incompat5): New.
13494 (gno-variable-location-views): New.
13496 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13498 PR tree-optimization/84136
13499 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13500 that the result of find_edge is non-NULL.
13502 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13505 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13506 storing integer register in SImode. Fix cost of 256 and 512
13507 byte aligned SSE register store.
13509 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13511 * config/i386/i386.c (ix86_multiplication_cost): Fix
13512 multiplication cost for TARGET_AVX512DQ.
13514 2018-02-08 Marek Polacek <polacek@redhat.com>
13516 PR tree-optimization/84238
13517 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13520 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13522 PR tree-optimization/84265
13523 * tree-vect-stmts.c (vectorizable_store): Don't treat
13524 VMAT_CONTIGUOUS accesses as grouped.
13525 (vectorizable_load): Likewise.
13527 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13529 PR tree-optimization/81635
13530 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13531 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13532 (test_round_for_mask): New functions.
13533 (wide_int_cc_tests): Call test_round_for_mask.
13534 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13535 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13536 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13537 range returned by get_range_info.
13539 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13542 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13543 * symtab.c: Include builtins.h
13544 (symtab_node::output_to_lto_symbol_table_p): Move here
13545 from lto-streamer-out.c:output_symbol_p.
13546 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13547 (output_symbol_p): Move all logic to symtab.c
13548 (produce_symtab): Update.
13550 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13552 * config/s390/s390-opts.h (enum indirect_branch): Define.
13553 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13554 (s390_indirect_branch_via_thunk)
13555 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13556 (enum s390_indirect_branch_type): Define.
13557 * config/s390/s390.c (struct s390_frame_layout, struct
13558 machine_function): Remove.
13559 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13560 (indirect_branch_table_label_no, indirect_branch_table_name):
13562 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13563 (enum s390_indirect_branch_option): Define.
13564 (s390_return_addr_from_memory): New function.
13565 (s390_handle_string_attribute): New function.
13566 (s390_attribute_table): Add new attribute handler.
13567 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13568 (s390_indirect_branch_via_thunk): New function.
13569 (s390_indirect_branch_via_inline_thunk): New function.
13570 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13571 sibling call optimization for non z10 compiles.
13572 (s390_emit_call): Force indirect branch target to be a single
13573 register. Add r1 clobber for non-z10 compiles.
13574 (s390_emit_epilogue): Emit return jump via return_use expander.
13575 (s390_reorg): Handle JUMP_INSNs as execute targets.
13576 (s390_option_override_internal): Perform validity checks for the
13577 new command line options.
13578 (s390_indirect_branch_attrvalue): New function.
13579 (s390_indirect_branch_settings): New function.
13580 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13581 (s390_output_indirect_thunk_function): New function.
13582 (s390_code_end): Implement target hook.
13583 (s390_case_values_threshold): Implement target hook.
13584 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13586 * config/s390/s390.h (struct s390_frame_layout)
13587 (struct machine_function): Move here from s390.c.
13588 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13589 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13590 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13591 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13592 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13593 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13594 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13595 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13596 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13597 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13598 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13599 (mnemonic attribute): Add values which aren't recognized
13601 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13602 pattern for branch conversion. Fix mnemonic attribute.
13603 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13604 indirect branch via thunk if requested.
13605 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13606 ("*indirect_jump"): Disable for branch conversion using out of
13608 ("indirect_jump_via_thunk<mode>_z10")
13609 ("indirect_jump_via_thunk<mode>")
13610 ("indirect_jump_via_inlinethunk<mode>_z10")
13611 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13612 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13613 ("casesi_jump_via_inlinethunk<mode>_z10")
13614 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13615 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13616 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13617 ("*indirect2_jump"): Disable for branch conversion.
13618 ("casesi_jump"): Turn into expander and expand patterns for branch
13620 ("return_use"): New expander.
13621 ("*return"): Emit return via thunk and rename it to ...
13622 ("*return<mode>"): ... this one.
13623 * config/s390/s390.opt: Add new options and and enum for the
13626 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13628 * lra-constraints.c (match_reload): Unconditionally use
13629 gen_lowpart_SUBREG, rather than selecting between that
13630 and equivalent gen_rtx_SUBREG code.
13632 2018-02-08 Richard Biener <rguenther@suse.de>
13634 PR tree-optimization/84233
13635 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13636 changed flag instead of boguously re-using phi_inserted.
13638 2018-02-08 Martin Jambor <mjambor@suse.cz>
13640 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13641 static local variables.
13643 2018-02-08 Richard Biener <rguenther@suse.de>
13645 PR tree-optimization/84278
13646 * tree-vect-stmts.c (vectorizable_store): When looking for
13647 smaller vector types to perform grouped strided loads/stores
13648 make sure the mode is supported by the target.
13649 (vectorizable_load): Likewise.
13651 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13653 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13654 Increase LDP/STP opportunities by adding adjacent callee-saves.
13656 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13658 PR rtl-optimization/84068
13659 PR rtl-optimization/83459
13660 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13662 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13664 PR tree-optimization/84224
13665 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13666 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13667 non-zero arguments.
13669 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13672 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13673 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13674 count, remove one USE.
13676 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13678 * doc/install.texi (Configuration): Document the
13679 --with-long-double-format={ibm,ieee} PowerPC configuration
13683 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13684 Convert from define_expand to be define_insn_and_split. Rework
13685 float/double/_Float128 conversions to QI/HI/SImode to work with
13686 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13687 conversions to QI/HImode types did a store and then a load to
13688 truncate the value. For conversions to VSX registers, don't split
13689 the insn, instead emit the code directly. Use the code iterator
13690 any_fix to combine signed and unsigned conversions.
13691 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13692 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13693 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13694 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13695 (fix_<mode>di2_hw): Likewise.
13696 (fixuns_<mode>di2_hw): Likewise.
13697 (fix_<mode>si2_hw): Likewise.
13698 (fixuns_<mode>si2_hw): Likewise.
13699 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13700 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13701 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13702 fix<uns>_trunc<SFDF:mode>si2_p8.
13703 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13704 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13705 (fix<uns>_<mode>_mem): Likewise.
13706 (fctiw<u>z_<mode>_mem): Likewise.
13707 (fix<uns>_<mode>_mem): Likewise.
13708 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13709 the register allocator from doing a direct move to the GPRs to do
13710 a store, and instead use the ISA 3.0 store byte/half-word from
13711 vector register instruction. For IEEE 128-bit floating point,
13712 also optimize stores of 32-bit ints.
13713 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13715 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13717 * genextract.c (push_pathstr_operand): New function to support
13719 (walk_rtx): Call push_pathstr_operand.
13720 (print_path): Support [a-zA-Z].
13722 2018-02-07 Richard Biener <rguenther@suse.de>
13724 PR tree-optimization/84037
13725 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13726 (cse_and_gimplify_to_preheader): Declare.
13727 (vect_get_place_in_interleaving_chain): Likewise.
13728 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13730 (_loop_vec_info::~_loop_vec_info): Delete it.
13731 (cse_and_gimplify_to_preheader): New function.
13732 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13733 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13734 (vectorizable_load): Likewise. For grouped stores always base
13735 the IV on the first element.
13736 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13737 condition before gimplifying.
13739 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13741 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13742 *DIV_EXPR and *MOD_EXPR.
13744 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13747 * config/i386/i386.c (ix86_option_override_internal): Mask out
13748 the CF_SET bit when checking -fcf-protection.
13750 2018-02-07 Tom de Vries <tom@codesourcery.com>
13753 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13756 2018-02-07 Richard Biener <rguenther@suse.de>
13758 PR tree-optimization/84204
13759 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13762 PR tree-optimization/84205
13763 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13764 special-case isl_ast_op_zdiv_r.
13766 PR tree-optimization/84223
13767 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13768 Only add conditions from within the region.
13769 (gather_bbs::after_dom_children): Adjust.
13771 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13774 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13775 * config/avr/avr.md: Only post-reload split REG-REG moves if
13776 either register is GENERAL_REG_P.
13778 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13780 PR tree-optimization/84235
13781 * tree-ssa-scopedtables.c
13782 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13783 if the subtraction is performed in floating point type where NaNs are
13784 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13785 build 1. Formatting fix.
13787 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13790 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13791 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13792 and skip it regardless of bb boundaries. Use CALL_P macro,
13793 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13794 unnecessarily, formatting fix.
13796 2018-02-06 Michael Collison <michael.collison@arm.com>
13798 * config/arm/thumb2.md:
13799 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13800 (*thumb_mov_notscc): Ditto.
13802 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13805 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13806 unsigned_fix, not "s".
13808 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13810 * configure.ac (gcc_fn_eh_frame_ro): New function.
13811 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13812 correct .eh_frame permissions.
13813 * configure: Regenerate.
13815 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13817 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13818 irrelevant options.
13820 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13822 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13823 Display warning message for -mno-speculate-indirect-jumps.
13825 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13827 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13829 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13831 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13833 PR tree-optimization/84225
13834 * tree-eh.c (find_trapping_overflow): Only call
13835 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13837 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13840 * config/i386/i386.c: Reimplement the check of possible options
13841 -mibt/-mshstk conbination. Change error messages.
13842 * doc/invoke.texi: Fix a typo: remove extra '='.
13844 2018-02-06 Marek Polacek <polacek@redhat.com>
13846 PR tree-optimization/84228
13847 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13849 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13852 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13853 emitted arch directives.
13854 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13855 __ARM_FEATURE_COPROC before changing architectures.
13857 2018-02-06 Richard Biener <rguenther@suse.de>
13859 * config/i386/i386.c (print_reg): Fix typo.
13860 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13862 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13864 * configure: Regenerate.
13866 2018-02-05 Martin Sebor <msebor@redhat.com>
13868 PR tree-optimization/83369
13869 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13872 2018-02-05 Martin Liska <mliska@suse.cz>
13874 * doc/invoke.texi: Cherry-pick upstream r323995.
13876 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13878 * ira.c (ira_init_register_move_cost): Adjust comment.
13880 2018-02-05 Martin Liska <mliska@suse.cz>
13882 PR gcov-profile/84137
13883 * doc/gcov.texi: Fix typo in documentation.
13885 2018-02-05 Martin Liska <mliska@suse.cz>
13887 PR gcov-profile/83879
13888 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13889 using dlopen functionality.
13891 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13893 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13894 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13895 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13896 _mm_maskz_range_round_ss): New intrinsics.
13897 (__builtin_ia32_rangesd128_round)
13898 (__builtin_ia32_rangess128_round): Remove.
13899 (__builtin_ia32_rangesd128_mask_round,
13900 __builtin_ia32_rangess128_mask_round): New builtins.
13901 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13902 __builtin_ia32_rangess128_round): Remove.
13903 (__builtin_ia32_rangesd128_mask_round,
13904 __builtin_ia32_rangess128_mask_round): New builtins.
13905 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13906 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13907 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13908 "<round_saeonly_constraint>")): Changed to ...
13909 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13910 "<round_saeonly_scalar_constraint>")): ... this.
13911 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13912 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13913 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13914 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13915 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13917 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13919 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13921 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13922 Remove all values except native, 8540 and 8548.
13924 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13926 * config/i386/i386.c (ix86_output_function_return): Pass
13927 INVALID_REGNUM, instead of -1, as invalid register number to
13928 indirect_thunk_name and output_indirect_thunk.
13930 2018-02-02 Julia Koval <julia.koval@intel.com>
13932 * config.gcc: Add -march=icelake.
13933 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13934 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13935 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13936 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13937 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13938 (processor_target_table): Add icelake.
13939 (ix86_option_override_internal): Handle new PTAs.
13940 (get_builtin_code_for_version): Handle icelake.
13941 (M_INTEL_COREI7_ICELAKE): New.
13942 (fold_builtin_cpu): Handle icelake.
13943 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13944 * doc/invoke.texi: Add -march=icelake.
13946 2018-02-02 Julia Koval <julia.koval@intel.com>
13948 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13949 to wide_int_bitmask.
13950 * wide-int-bitmask.h: New.
13952 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13955 * config/i386/i386.md: Replace Pmode with word_mode in
13956 builtin_setjmp_setup and builtin_longjmp to support x32.
13958 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13962 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13964 (rs6000_supported_cpu_names): New static variable.
13965 (linux_cpu_translation_table): Likewise.
13966 (elf_platform) <cpu>: Define new static variable and use it.
13967 Translate kernel AT_PLATFORM name to canonical name if needed.
13968 Error if platform name is unknown.
13970 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13973 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13975 2018-02-01 Jeff Law <law@redhat.com>
13978 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13979 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13980 the scratch if RELEASE_VIA_POP is false.
13981 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13982 If we have to save a temporary register, decrement SIZE appropriately.
13983 Pass new arguments to release_scratch_register_on_entry.
13984 (ix86_adjust_stack_and_probe): Likewise.
13985 (ix86_emit_probe_stack_range): Pass new arguments to
13986 release_scratch_register_on_entry.
13988 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13990 PR rtl-optimization/84157
13991 * combine.c (change_zero_ext): Use REG_P predicate in
13992 front of HARD_REGISTER_P predicate.
13994 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13996 * config/avr/avr.c (avr_option_override): Move disabling of
13997 -fdelete-null-pointer-checks to...
13998 * common/config/avr/avr-common.c (avr_option_optimization_table):
14001 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14003 PR tree-optimization/81635
14004 * tree-data-ref.c (split_constant_offset_1): For types that
14005 wrap on overflow, try to use range info to prove that wrapping
14008 2018-02-01 Renlin Li <renlin.li@arm.com>
14011 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14012 TAILCALL_ADDR_REGS.
14013 (aarch64_register_move_cost): Likewise.
14014 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14015 TAILCALL_ADDR_REGS.
14016 (REG_CLASS_NAMES): Likewise.
14017 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14018 TAILCALL_ADDR_REGS. Remove IP registers.
14019 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14021 2018-02-01 Richard Biener <rguenther@suse.de>
14023 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14024 for specifying RPO order and allow NULL for that.
14025 * domwalk.c (dom_walker::dom_walker): Likewise.
14026 (dom_walker::walk): Handle NULL RPO order.
14027 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14029 (rewrite_update_dom_walker): Likewise.
14030 (mark_def_dom_walker): Likewise.
14032 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14034 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14035 (aarch64_maybe_expand_sve_subreg_move): Declare.
14036 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14037 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14039 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14040 that are semantically a reverse operation.
14041 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14042 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14043 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14045 (aarch64_can_change_mode_class): For big-endian, forbid changes
14046 between two SVE modes if they have different element sizes.
14048 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14050 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14051 the TImode handling for big-endian targets.
14053 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14055 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14056 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14058 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14059 Remove BSWAP handing for big-endian targets and use the form of
14060 LD1RQ appropariate for the mode.
14062 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14064 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14065 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14066 duplicated element.
14068 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14071 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14072 check for operands that need to go through aarch64_sve_reload_be.
14074 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14076 PR tree-optimization/81661
14077 PR tree-optimization/84117
14078 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14079 * tree-eh.c: Include gimplify.h.
14080 (find_trapping_overflow, replace_trapping_overflow,
14081 rewrite_to_non_trapping_overflow): New functions.
14082 * tree-vect-loop.c: Include tree-eh.h.
14083 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14084 * tree-data-ref.c: Include tree-eh.h.
14085 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14087 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14089 PR rtl-optimization/84123
14090 * combine.c (change_zero_ext): Check if hard register satisfies
14091 can_change_dest_mode before calling gen_lowpart_SUBREG.
14093 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14096 * ira.c (ira_init_register_move_cost): Remove assert.
14098 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14100 PR rtl-optimization/84071
14101 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14102 * doc/tm.texi: Regenerate.
14104 2018-01-31 Richard Biener <rguenther@suse.de>
14106 PR tree-optimization/84132
14107 * tree-data-ref.c (analyze_miv_subscript): Properly
14108 check whether evolution_function_is_affine_multivariate_p
14109 before calling gcd_of_steps_may_divide_p.
14111 2018-01-31 Julia Koval <julia.koval@intel.com>
14114 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14115 * config/i386/i386.md (rdpid_rex64) New.
14116 (rdpid): Make 32bit only.
14118 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14121 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14122 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14124 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14127 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14129 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14131 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14133 PR rtl-optimization/84071
14134 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14135 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14137 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14139 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14140 (arc_attribute_table): Add 'aux' attribute.
14141 (arc_in_small_data_p): Consider aux like variables.
14142 (arc_is_aux_reg_p): New function.
14143 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14144 (arc_get_aux_arg): New function.
14145 (prepare_move_operands): Handle aux-register access.
14146 (arc_handle_aux_attribute): New function.
14147 * doc/extend.texi (ARC Variable attributes): Add subsection.
14149 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14151 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14152 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14153 (arc_attribute_table): Add 'uncached' attribute.
14154 (arc_print_operand): Print '.di' flag for uncached memory
14156 (arc_in_small_data_p): Do not consider for small data the uncached
14158 (arc_is_uncached_mem_p): New function.
14159 * config/arc/predicates.md (compact_store_memory_operand): Check
14160 for uncached memory accesses.
14161 (nonvol_nonimm_operand): Likewise.
14162 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14164 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14167 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14168 falign-loops=): Add Optimization flag.
14170 2018-01-30 Jeff Law <law@redhat.com>
14173 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14174 INT_REGISTERS_SAVED. Check it prior to calling
14175 get_scratch_register_on_entry.
14176 (ix86_adjust_stack_and_probe): Similarly.
14177 (ix86_emit_probe_stack_range): Similarly.
14178 (ix86_expand_prologue): Corresponding changes.
14180 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14183 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14184 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14186 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14189 * lra-constraints.c (curr_insn_transform): Process AND in the
14192 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14194 PR rtl-optimization/83986
14195 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14196 dependence against last_pending_memory_flush in addition to
14197 pending_jump_insns.
14199 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14201 PR tree-optimization/81611
14202 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14205 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14208 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14211 2018-01-30 Richard Biener <rguenther@suse.de>
14212 Jakub Jelinek <jakub@redhat.com>
14214 PR tree-optimization/84111
14215 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14216 inner loops added during recursion, as they don't have up-to-date
14219 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14222 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14223 (can_inline_edge_by_limits_p): ... here.
14224 (can_early_inline_edge_p, check_callers,
14225 update_caller_keys, update_callee_keys, recursive_inlining,
14226 add_new_edges_to_heap, speculation_useful_p,
14227 inline_small_functions,
14228 inline_small_functions, flatten_function,
14229 inline_to_all_callers_1): Update.
14231 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14233 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14236 2018-01-30 Richard Biener <rguenther@suse.de>
14238 PR tree-optimization/83008
14239 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14240 invariant and constant vector uses in stmts when they need
14241 more than one stmt.
14243 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14246 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14247 * configure: Regenerate.
14249 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14251 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14253 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14254 Use gen_rtx_REG rather than gen_lowpart.
14256 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14258 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14259 rather than 0 when creating partial subregs.
14261 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14263 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14266 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14269 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14270 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14271 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14272 flags. This restores the settings used before the 2017-07-24.
14273 Turning off pre increment/decrement/modify allows IVOPTS to
14274 optimize DF/SF loops where the index is an int.
14276 2018-01-29 Richard Biener <rguenther@suse.de>
14277 Kelvin Nilsen <kelvin@gcc.gnu.org>
14280 * tree-vect-stmts.c (vectorizable_call): Don't call
14281 targetm.vectorize_builtin_md_vectorized_function if callee is
14284 2018-01-22 Carl Love <cel@us.ibm.com>
14286 * doc/extend.tex: Fix typo in second arg in
14287 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14289 2018-01-29 Richard Biener <rguenther@suse.de>
14291 PR tree-optimization/84086
14292 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14293 (flush_ssaname_freelist): When SSA names were released reset
14294 the SCEV hash table.
14296 2018-01-29 Richard Biener <rguenther@suse.de>
14298 PR tree-optimization/84057
14299 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14300 removed paths when removing edges.
14302 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14304 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14305 -mfunction-return=@var{choice}.
14307 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14309 PR diagnostic/84034
14310 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14311 Handle CR like TAB.
14312 (layout::print_source_line): Likewise.
14313 (test_get_line_width_without_trailing_whitespace): Add test cases.
14315 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14317 PR middle-end/84040
14318 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14321 2018-01-26 Jim Wilson <jimw@sifive.com>
14323 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14325 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14328 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14330 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14331 and CMP + SUB-immediate -> SUBS.
14333 2018-01-26 Martin Sebor <msebor@redhat.com>
14335 PR tree-optimization/83896
14336 * tree-ssa-strlen.c (get_string_len): Rename...
14337 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14338 Avoid assuming length is constant.
14339 (handle_char_store): Use HOST_WIDE_INT for string length.
14341 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14344 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14345 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14347 2018-01-26 Richard Biener <rguenther@suse.de>
14349 PR rtl-optimization/84003
14350 * dse.c (record_store): Only record redundant stores when
14351 the earlier store aliases at least all accesses the later one does.
14353 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14355 PR rtl-optimization/83985
14356 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14357 REG_CFA_RESTORE insns.
14358 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14359 REG_CFA_RESTORE insns here.
14362 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14363 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14365 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14367 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14368 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14369 (arc_init): Likewise.
14370 (arc_override_options): Likewise.
14371 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14373 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14375 * config/arc/arc.h (TARGET_DBNZ): Define.
14376 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14377 properly set the tune attribute.
14378 (dbnz): Use TARGET_DBNZ guard.
14379 * config/arc/arc.opt (mtune): Add core3 option.
14381 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14383 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14384 recognize new pic like addresses.
14385 (arc_delegitimize_address): Clean up.
14387 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14389 * config/arc/arc-arches.def: Option mrf16 valid for all
14391 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14392 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14393 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14394 * config/arc/arc-tables.opt: Regenerate.
14395 * config/arc/arc.c (arc_conditional_register_usage): Handle
14396 reduced register file case.
14397 (arc_file_start): Set must have build attributes.
14398 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14399 mrf16 option value.
14400 * config/arc/arc.opt (mrf16): Add new option.
14401 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14402 * config/arc/genmultilib.awk: Handle new mrf16 option.
14403 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14404 * config/arc/t-multilib: Regenerate.
14405 * doc/invoke.texi (ARC Options): Document mrf16 option.
14407 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14409 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14410 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14411 (arc_attribute_table): Add 'secure_call' attribute.
14412 (arc_print_operand): Print secure call operand.
14413 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14415 (arc_is_secure_call_p): New function. * config/arc/arc.md
14416 (call_i): Add support for sjli instruction.
14417 (call_value_i): Likewise.
14418 * config/arc/constraints.md (Csc): New constraint.
14420 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14421 John Eric Martin <John.Martin@emmicro-us.com>
14423 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14424 * config/arc/arc.c (_arc_jli_section): New struct.
14425 (arc_jli_section): New type.
14426 (rc_jli_sections): New static variable.
14427 (arc_handle_jli_attribute): New function.
14428 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14429 (arc_file_end): New function.
14430 (TARGET_ASM_FILE_END): Define.
14431 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14432 (arc_add_jli_section): New function.
14433 (jli_call_scan): Likewise.
14434 (arc_reorg): Call jli_call_scan.
14435 (arc_output_addsi): Remove 'S' from printing asm operand.
14436 (arc_is_jli_call_p): New function.
14437 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14439 (movhi_insn): Likewise.
14440 (movsi_insn): Likewise.
14441 (movsi_set_cc_insn): Likewise.
14442 (loadqi_update): Likewise.
14443 (load_zeroextendqisi_update): Likewise.
14444 (load_signextendqisi_update): Likewise.
14445 (loadhi_update): Likewise.
14446 (load_zeroextendhisi_update): Likewise.
14447 (load_signextendhisi_update): Likewise.
14448 (loadsi_update): Likewise.
14449 (loadsf_update): Likewise.
14450 (movsicc_insn): Likewise.
14451 (bset_insn): Likewise.
14452 (bxor_insn): Likewise.
14453 (bclr_insn): Likewise.
14454 (bmsk_insn): Likewise.
14455 (bicsi3_insn): Likewise.
14456 (cmpsi_cc_c_insn): Likewise.
14457 (movsi_ne): Likewise.
14458 (movsi_cond_exec): Likewise.
14459 (clrsbsi2): Likewise.
14460 (norm_f): Likewise.
14470 (call_i): Remove 'S' asm letter, add jli instruction.
14471 (call_value_i): Likewise.
14472 * config/arc/arc.op (mjli-always): New option.
14473 * config/arc/constraints.md (Cji): New constraint.
14474 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14476 (subsf3_fpx): Likewise.
14477 (mulsf3_fpx): Likewise.
14478 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14480 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14481 function attrbutes.
14482 * doc/invoke.texi (ARC): Document mjli-always option.
14484 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14486 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14487 avoid addition with 0 and use incw and decw where possible.
14489 2018-01-26 Richard Biener <rguenther@suse.de>
14491 PR tree-optimization/81082
14492 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14493 association if it requires casting to unsigned.
14494 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14495 from fold_plusminus_mult_expr to catch important cases late when
14496 range info is available.
14498 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14500 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14501 * configure.ac (hidden_linkonce): New test.
14502 * configure: Regenerate.
14503 * config.in: Regenerate.
14505 2018-01-26 Julia Koval <julia.koval@intel.com>
14507 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14508 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14509 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14510 _mm_mask_bitshuffle_epi64_mask): Fix type.
14511 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14512 USI_FTYPE_V4DI_V4DI_USI): Remove.
14513 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14514 __builtin_ia32_vpshufbitqmb256_mask,
14515 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14516 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14517 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14519 2018-01-26 Alan Modra <amodra@gmail.com>
14522 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14523 UNSPEC_VBPERMQ. Sort other unspecs.
14525 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14527 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14529 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14531 PR middle-end/83055
14532 * predict.c (drop_profile): Do not push/pop cfun; update also
14534 (handle_missing_profiles): Fix logic looking for zero profiles.
14536 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14538 PR middle-end/83977
14539 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14540 on functions with #pragma omp declare simd or functions with simd
14542 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14543 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14544 Remove trailing \n from warning_at calls.
14546 2018-01-25 Tom de Vries <tom@codesourcery.com>
14549 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14550 for neutered workers.
14552 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14555 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14556 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14558 2018-01-24 Jeff Law <law@redhat.com>
14561 * i386.c (get_probe_interval): Move to earlier point.
14562 (ix86_compute_frame_layout): If -fstack-clash-protection and
14563 the frame is larger than the probe interval, then use pushes
14564 to save registers rather than reg->mem moves.
14565 (ix86_expand_prologue): Remove conditional for int_registers_saved
14568 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14571 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14572 min/max for never referenced object.
14574 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14576 PR middle-end/83977
14577 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14579 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14580 attributes from DECL_ATTRIBUTES (decl) without affecting
14581 DECL_ATTRIBUTES (current_function_decl).
14582 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14583 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14585 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14587 PR tree-optimization/83979
14588 * fold-const.c (fold_comparison): Use constant_boolean_node
14589 instead of boolean_{true,false}_node.
14591 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14593 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14596 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14598 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14599 Simplify the clause that sets the length attribute.
14600 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14601 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14602 clause that sets the length attribute.
14603 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14605 2018-01-24 Tom de Vries <tom@codesourcery.com>
14608 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14609 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14610 Add strict parameter.
14611 (prevent_branch_around_nothing): Insert dummy insn between branch to
14612 label and label with no ptx insn inbetween.
14613 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14615 2018-01-24 Tom de Vries <tom@codesourcery.com>
14618 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14619 for neutered threads in warp.
14620 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14622 2018-01-24 Richard Biener <rguenther@suse.de>
14624 PR tree-optimization/83176
14625 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14628 2018-01-24 Richard Biener <rguenther@suse.de>
14630 PR tree-optimization/82819
14631 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14632 code generating pluses that are no-ops in the target precision.
14634 2018-01-24 Richard Biener <rguenther@suse.de>
14636 PR middle-end/84000
14637 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14639 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14641 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14642 to merge probabilities.
14643 * predict.c (probably_never_executed): Also mark as cold functions
14644 with global 0 profile and guessed local profile.
14645 * profile-count.c (profile_probability::combine_with_count): New
14647 * profile-count.h (profile_probability::operator*,
14648 profile_probability::operator*=, profile_probability::operator/,
14649 profile_probability::operator/=): Reduce precision to adjusted
14650 and set value to guessed on contradictory divisions.
14651 (profile_probability::combine_with_freq): Remove.
14652 (profile_probability::combine_wiht_count): Declare.
14653 (profile_count::force_nonzero):: Set to adjusted.
14654 (profile_count::probability_in):: Set quality to adjusted.
14655 * tree-ssa-tail-merge.c (replace_block_by): Use
14656 combine_with_count.
14658 2018-01-23 Andrew Waterman <andrew@sifive.com>
14659 Jim Wilson <jimw@sifive.com>
14661 * config/riscv/riscv.c (riscv_stack_boundary): New.
14662 (riscv_option_override): Set riscv_stack_boundary. Handle
14663 riscv_preferred_stack_boundary_arg.
14664 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14665 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14666 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14667 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14668 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14669 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14671 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14674 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14675 of struct ix86_frame.
14676 (ix86_expand_epilogue): Likewise. Add a local variable for
14677 the reg_save_offset field in struct ix86_frame.
14679 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14681 PR tree-optimization/82604
14682 * tree-loop-distribution.c (enum partition_kind): New enum item
14683 PKIND_PARTIAL_MEMSET.
14684 (partition_builtin_p): Support above new enum item.
14685 (generate_code_for_partition): Ditto.
14686 (compute_access_range): Differentiate cases that equality can be
14687 proven at all loops, the innermost loops or no loops.
14688 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14689 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14690 (finalize_partitions, distribute_loop): Don't fuse partition of
14691 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14692 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14693 parloop is enabled.
14695 2018-01-23 Martin Liska <mliska@suse.cz>
14697 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14698 order to ignore the predictor.
14699 (PRED_POLYMORPHIC_CALL): Likewise.
14700 (PRED_RECURSIVE_CALL): Likewise.
14702 2018-01-23 Martin Liska <mliska@suse.cz>
14704 * tree-profile.c (tree_profiling): Print function header to
14705 aware reader which function we are working on.
14706 * value-prof.c (gimple_find_values_to_profile): Do not print
14707 not interesting value histograms.
14709 2018-01-23 Martin Liska <mliska@suse.cz>
14711 * profile-count.h (enum profile_quality): Add
14712 profile_uninitialized as the first value. Do not number values
14713 as they are zero based.
14714 (profile_count::verify): Update sanity check.
14715 (profile_probability::verify): Likewise.
14717 2018-01-23 Nathan Sidwell <nathan@acm.org>
14719 * doc/invoke.texi (ffor-scope): Deprecate.
14721 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14723 PR tree-optimization/83510
14724 * domwalk.c (set_all_edges_as_executable): New function.
14725 (dom_walker::dom_walker): Convert bool param
14726 "skip_unreachable_blocks" to enum reachability. Move setup of
14727 edge flags to set_all_edges_as_executable and only do it when
14728 reachability is REACHABLE_BLOCKS.
14729 * domwalk.h (enum dom_walker::reachability): New enum.
14730 (dom_walker::dom_walker): Convert bool param
14731 "skip_unreachable_blocks" to enum reachability.
14732 (set_all_edges_as_executable): New decl.
14733 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14734 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14736 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14737 but converting true to REACHABLE_BLOCKS.
14738 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14740 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14741 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14742 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14744 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14745 if check_all_array_refs will be called.
14747 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14749 * tree.c (selftest::test_location_wrappers): Add more test
14752 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14754 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14755 (selftest::test_bit_in_range): Likewise.
14757 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14760 * doc/sourcebuild.texi (vect_float): Say that the selector
14761 only describes the situation when -funsafe-math-optimizations is on.
14762 (vect_float_strict): Document.
14764 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14766 PR tree-optimization/83965
14767 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14768 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14769 instead of checking only for a reduction.
14770 (vect_recog_widen_sum_pattern): Likewise.
14772 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14774 * predict.c (probably_never_executed): Only use precise profile info.
14775 (compute_function_frequency): Skip after inlining hack since we now
14776 have quality checking.
14778 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14780 * profile-count.h (profile_probability::very_unlikely,
14781 profile_probability::unlikely, profile_probability::even): Set
14782 precision to guessed.
14784 2018-01-23 Richard Biener <rguenther@suse.de>
14786 PR tree-optimization/83963
14787 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14788 Properly terminate dominator walk when crossing the exit edge not
14789 when visiting its source block.
14791 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14794 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14795 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14797 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14799 PR tree-optimization/83957
14800 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14801 semicolon after for body surrounded by braces.
14803 PR tree-optimization/83081
14804 * profile-count.h (profile_probability::split): New method.
14805 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14806 Use profile_probability::split.
14807 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14808 when splitting a single conditional jump into 2.
14810 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14812 PR tree-optimization/69452
14813 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14816 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14818 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14819 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14820 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14822 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14824 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14826 * config/rl78/rl78.md (movdi): New define_expand.
14827 * config/rl78/rl78.c (rl78_split_movdi): New function.
14829 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14832 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14834 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14835 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14836 128-bit to produce an UNSPEC move to get the double word with the
14837 signbit and then a shift directly to do signbit.
14838 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14839 implementation with a new version that just does either a direct
14840 move or a regular move. Move memory interface to separate insns.
14841 Move insns so they are next to the expander.
14842 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14843 with signbit move. Split big and little endian case.
14844 (signbit<mode>2_dm_mem_le): Likewise.
14845 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14846 (signbit<mode>2_dm2): Likewise.
14848 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14850 * config/rl78/rl78.md (anddi3): New define_expand.
14852 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14854 * config/rl78/rl78.md (umindi3): New define_expand.
14856 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14858 * config/rl78/rl78.md (smindi3): New define_expand.
14860 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14862 * config/rl78/rl78.md (smaxdi3): New define_expand.
14864 2018-01-22 Carl Love <cel@us.ibm.com>
14866 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14867 LVX_V1TI): Add macro expansion.
14868 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14869 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14870 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14871 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14872 Change check to determine if the instruction is a byte reversing
14873 entry. Fix typo in comment.
14874 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14875 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14876 Add def_builtin calls for new builtins.
14877 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14878 Add define_insn expansion.
14880 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14882 * config/rl78/rl78.md (umaxdi3): New define_expand.
14884 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14886 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14887 for non-QImode registers.
14889 2018-01-22 Richard Biener <rguenther@suse.de>
14891 PR tree-optimization/83963
14892 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14893 including the loop exit block.
14894 (scop_detection::merge_sese): Likewise.
14895 (scop_detection::add_scop): Do it here instead.
14897 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14899 * doc/sourcebuild.texi (arm_softfloat): Document.
14901 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14904 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14905 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14906 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14908 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14909 David Edelsohn <dje.gcc@gmail.com>
14912 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14913 Change "crset eq" to "crset 2".
14914 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14915 (*call_indirect_aix<mode>_nospec): Likewise.
14916 (*call_value_indirect_aix<mode>_nospec): Likewise.
14917 (*call_indirect_elfv2<mode>_nospec): Likewise.
14918 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14919 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14920 change assembly output from . to $.
14921 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14922 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14923 (*tablejump<mode>_internal1_nospec): Likewise.
14925 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14928 * config/sh/sh_optimize_sett_clrt.cc:
14929 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14931 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14933 PR tree-optimization/83940
14934 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14935 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14936 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14937 use it to pass back the definition type.
14938 (vect_check_store_rhs): Likewise rhs_dt_out.
14939 (vect_build_gather_load_calls): Add a mask_dt argument and use
14940 it instead of a call to vect_is_simple_use.
14941 (vectorizable_store): Update calls to vect_check_load_store_mask
14942 and vect_check_store_rhs. Use the dt returned by the latter instead
14943 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14944 instead of calls to vect_is_simple_use. Pass the scalar rather
14945 than the vector operand to vect_is_simple_use when handling
14946 second and subsequent copies of an rhs value.
14947 (vectorizable_load): Update calls to vect_check_load_store_mask
14948 and vect_build_gather_load_calls. Use the cached mask_dt and
14949 gs_info.offset_dt instead of calls to vect_is_simple_use.
14951 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14953 PR middle-end/83945
14954 * tree-emutls.c: Include gimplify.h.
14955 (lower_emutls_2): New function.
14956 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14957 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14958 it before further processing.
14961 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14962 UINTVAL (trueop1) instead of INTVAL (op1).
14964 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14968 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14969 INCOMING_FRAME_SP_OFFSET if not defined.
14970 (scan_trace): Add ENTRY argument. If true and
14971 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14972 emit a note to adjust the CFA offset.
14973 (create_cfi_notes): Adjust scan_trace callers.
14974 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14975 INCOMING_FRAME_SP_OFFSET in the CIE.
14976 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14977 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14979 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14980 * doc/tm.texi: Regenerated.
14982 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14984 PR rtl-optimization/83147
14985 * lra-constraints.c (remove_inheritance_pseudos): Use
14986 lra_substitute_pseudo_within_insn.
14988 2018-01-19 Tom de Vries <tom@codesourcery.com>
14989 Cesar Philippidis <cesar@codesourcery.com>
14992 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14994 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14997 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14998 spaces for function labels.
15000 2018-01-19 Martin Liska <mliska@suse.cz>
15002 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15003 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15004 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15005 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15006 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15007 (PRED_CONST_RETURN): Change from 69 to 65.
15008 (PRED_NULL_RETURN): Change from 91 to 71.
15009 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15010 (PRED_LOOP_GUARD): Change from 66 to 73.
15012 2018-01-19 Martin Liska <mliska@suse.cz>
15014 * predict.c (predict_insn_def): Add new assert.
15015 (struct branch_predictor): Change type to signed integer.
15016 (test_prediction_value_range): Amend test to cover
15017 PROB_UNINITIALIZED.
15018 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15019 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15020 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15021 (PRED_LOOP_IV_COMPARE): Likewise.
15022 * predict.h (PROB_UNINITIALIZED): Define new constant.
15024 2018-01-19 Martin Liska <mliska@suse.cz>
15026 * predict.c (dump_prediction): Add new format for
15027 analyze_brprob.py script which is enabled with -details
15029 * profile-count.h (precise_p): New function.
15031 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15033 PR tree-optimization/83922
15034 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15035 there are no statements that need masking.
15036 (vect_active_double_reduction_p): New function.
15037 (vect_analyze_loop_operations): Use it when handling phis that
15038 are not in the loop header.
15040 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15042 PR tree-optimization/83914
15043 * tree-vect-loop.c (vectorizable_induction): Don't convert
15044 init_expr or apply the peeling adjustment for inductions
15045 that are nested within the vectorized loop.
15047 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15049 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15052 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15056 * function.h (gimplify_parameters): Add gimple_seq * argument.
15057 * function.c: Include gimple.h and options.h.
15058 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15059 for the added local temporaries if needed.
15060 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15061 if there are any parameter cleanups, wrap whole body into a
15062 try/finally with the cleanups.
15064 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15067 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15068 Use GET_MODE_CLASS for scalar floating point.
15070 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15074 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15075 Fix call of call_cgraph_insertion_hooks.
15077 2018-01-18 Martin Sebor <msebor@redhat.com>
15079 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15081 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15084 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15087 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15090 * common.opt: (-ffile-prefix-map): New option.
15091 * opts.c (common_handle_option): Defer it.
15092 * opts-global.c (handle_common_deferred_options): Handle it.
15093 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15094 * file-prefix-map.h: New file.
15095 (remap_debug_filename, add_debug_prefix_map): ...here.
15096 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15097 * final.c (debug_prefix_map, add_debug_prefix_map
15098 remap_debug_filename): Move to...
15099 * file-prefix-map.c: New file.
15100 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15101 generalize, get rid of alloca(), use strrchr() instead of strchr().
15102 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15103 Implement in terms of add_prefix_map().
15104 (remap_macro_filename, remap_debug_filename): Implement in term of
15106 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15107 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15108 * dbxout.c: Include file-prefix-map.h.
15109 * varasm.c: Likewise.
15110 * vmsdbgout.c: Likewise.
15111 * xcoffout.c: Likewise.
15112 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15113 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15114 * doc/invoke.texi (-ffile-prefix-map): Document.
15115 (-fdebug-prefix-map): Update description.
15117 2018-01-18 Martin Liska <mliska@suse.cz>
15119 * config/i386/i386.c (indirect_thunk_name): Document that also
15121 (output_indirect_thunk): Document why both instructions
15122 (pause and lfence) are generated.
15124 2018-01-18 Richard Biener <rguenther@suse.de>
15126 PR tree-optimization/83887
15127 * graphite-scop-detection.c
15128 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15129 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15130 (scop_detection::merge_sese): Re-implement with a flood-fill
15131 algorithm that properly finds a SESE region if it exists.
15133 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15136 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15137 pointer_diff optimizations use view_convert instead of convert.
15139 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15141 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15142 Generate different code for -mno-speculate-indirect-jumps.
15143 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15144 (*call_indirect_aix<mode>): Disable for
15145 -mno-speculate-indirect-jumps.
15146 (*call_indirect_aix<mode>_nospec): New define_insn.
15147 (*call_value_indirect_aix<mode>): Disable for
15148 -mno-speculate-indirect-jumps.
15149 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15150 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15151 -mno-speculate-indirect-jumps.
15152 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15154 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15156 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15157 long double type, set the flags for noting the default long double
15158 type, even if we don't pass or return a long double type.
15160 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15163 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15166 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15168 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15169 support for merge[hl].
15170 (fold_mergehl_helper): New helper function.
15171 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15172 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15173 (altivec_vmrglw_direct): Add xxmrglw insn.
15175 2018-01-17 Andrew Waterman <andrew@sifive.com>
15177 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15178 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15180 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15183 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15184 call the lto_location_cache before reading the
15185 DECL_SOURCE_LOCATION of the types.
15187 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15188 Richard Sandiford <richard.sandiford@linaro.org>
15190 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15191 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15192 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15193 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15194 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15196 * config/aarch64/constraints.md (aarch64_movti_operand):
15198 * config/aarch64/predicates.md (Uti): Add new constraint.
15200 2018-01-17 Carl Love <cel@us.ibm.com>
15202 * config/rs6000/vsx.md (define_expand xl_len_r,
15203 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15204 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15206 (define_expand, define_insn): Move the shift left from the
15207 define_insn to the define_expand for lxvl and stxvl instructions.
15208 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15209 and XL_LEN_R definitions to PURE.
15211 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15213 * config/i386/i386.c (indirect_thunk_name): Declare regno
15214 as unsigned int. Compare regno with INVALID_REGNUM.
15215 (output_indirect_thunk): Ditto.
15216 (output_indirect_thunk_function): Ditto.
15217 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15218 in the call to output_indirect_thunk_function.
15220 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15222 PR middle-end/83884
15223 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15224 rather than the size of inner_type to determine the stack slot size
15225 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15227 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15230 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15233 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15235 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15236 endian Linux systems to optionally enable multilibs for selecting
15237 the long double type if the user configured an explicit type.
15238 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15239 have no long double multilibs if not defined.
15240 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15241 warn if the user used -mabi={ieee,ibm}longdouble and we built
15242 multilibs for long double.
15243 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15244 appropriate multilib option.
15245 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15247 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15248 for building long double multilibs.
15249 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15251 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15253 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15256 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15258 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15261 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15264 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15267 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15269 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15270 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15272 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15274 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15275 different rtl trees depending on TARGET_64BIT.
15276 (rs6000_gen_lvx): Likewise.
15278 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15280 * config/visium/visium.md (nop): Tweak comment.
15281 (hazard_nop): Likewise.
15283 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15285 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15286 -mspeculate-indirect-jumps.
15287 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15288 for -mno-speculate-indirect-jumps.
15289 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15290 (*call_value_indirect_elfv2<mode>): Disable for
15291 -mno-speculate-indirect-jumps.
15292 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15293 (indirect_jump): Emit different RTL for
15294 -mno-speculate-indirect-jumps.
15295 (*indirect_jump<mode>): Disable for
15296 -mno-speculate-indirect-jumps.
15297 (*indirect_jump<mode>_nospec): New define_insn.
15298 (tablejump): Emit different RTL for
15299 -mno-speculate-indirect-jumps.
15300 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15301 (tablejumpsi_nospec): New define_expand.
15302 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15303 (tablejumpdi_nospec): New define_expand.
15304 (*tablejump<mode>_internal1): Disable for
15305 -mno-speculate-indirect-jumps.
15306 (*tablejump<mode>_internal1_nospec): New define_insn.
15307 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15310 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15312 * caller-save.c (insert_save): Drop unnecessary parameter. All
15315 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15316 Richard Biener <rguenth@suse.de>
15319 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15320 return early, inline manually is_gimple_sizepos. Make sure if we
15321 call gimplify_expr we don't end up with a gimple constant.
15322 * tree.c (variably_modified_type_p): Don't return true for
15323 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15324 * gimplify.h (is_gimple_sizepos): Remove.
15326 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15328 PR tree-optimization/83857
15329 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15330 vectorizable_live_operation for pure SLP statements.
15331 (vectorizable_live_operation): Handle PHIs.
15333 2018-01-16 Richard Biener <rguenther@suse.de>
15335 PR tree-optimization/83867
15336 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15337 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15339 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15342 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15343 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15344 If off is not INTEGER_CST, issue a may not be aligned warning
15345 rather than isn't aligned. Use isn%'t rather than isn't.
15346 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15348 <case MULT_EXPR>: Improve the case when bottom and one of the
15349 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15350 operand, in that case check if the other operand is multiple of
15351 bottom divided by the INTEGER_CST operand.
15353 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15356 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15357 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15358 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15359 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15360 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15361 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15362 (pa_function_arg_size): New function.
15364 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15366 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15367 in a separate statement.
15369 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15371 PR tree-optimization/83847
15372 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15373 group gathers and scatters.
15375 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15377 PR rtl-optimization/86620
15378 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15380 PR rtl-optimization/83213
15381 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15382 to last if both are JUMP_INSNs.
15384 PR tree-optimization/83843
15385 * gimple-ssa-store-merging.c
15386 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15387 store_immediate_info for bswap/nop orig_stores.
15389 2018-01-15 Andrew Waterman <andrew@sifive.com>
15391 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15393 <UDIV>: Increase cost if !TARGET_DIV.
15395 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15397 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15398 (define_attr "cr_logical_3op"): New.
15399 (cceq_ior_compare): Adjust.
15400 (cceq_ior_compare_complement): Adjust.
15401 (*cceq_rev_compare): Adjust.
15402 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15403 (is_cracked_insn): Adjust.
15404 (insn_must_be_first_in_group): Adjust.
15405 * config/rs6000/40x.md: Adjust.
15406 * config/rs6000/440.md: Adjust.
15407 * config/rs6000/476.md: Adjust.
15408 * config/rs6000/601.md: Adjust.
15409 * config/rs6000/603.md: Adjust.
15410 * config/rs6000/6xx.md: Adjust.
15411 * config/rs6000/7450.md: Adjust.
15412 * config/rs6000/7xx.md: Adjust.
15413 * config/rs6000/8540.md: Adjust.
15414 * config/rs6000/cell.md: Adjust.
15415 * config/rs6000/e300c2c3.md: Adjust.
15416 * config/rs6000/e500mc.md: Adjust.
15417 * config/rs6000/e500mc64.md: Adjust.
15418 * config/rs6000/e5500.md: Adjust.
15419 * config/rs6000/e6500.md: Adjust.
15420 * config/rs6000/mpc.md: Adjust.
15421 * config/rs6000/power4.md: Adjust.
15422 * config/rs6000/power5.md: Adjust.
15423 * config/rs6000/power6.md: Adjust.
15424 * config/rs6000/power7.md: Adjust.
15425 * config/rs6000/power8.md: Adjust.
15426 * config/rs6000/power9.md: Adjust.
15427 * config/rs6000/rs64.md: Adjust.
15428 * config/rs6000/titan.md: Adjust.
15430 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15432 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15433 ix86_indirect_branch_register logic.
15435 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15437 * config/i386/constraints.md (Bs): Update
15438 ix86_indirect_branch_register check. Don't check
15439 ix86_indirect_branch_register with GOT_memory_operand.
15441 * config/i386/predicates.md (GOT_memory_operand): Don't check
15442 ix86_indirect_branch_register here.
15443 (GOT32_symbol_operand): Likewise.
15445 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15447 * config/i386/predicates.md (constant_call_address_operand):
15448 Rewrite ix86_indirect_branch_register logic.
15449 (sibcall_insn_operand): Likewise.
15451 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15453 * config/i386/constraints.md (Bs): Replace
15454 ix86_indirect_branch_thunk_register with
15455 ix86_indirect_branch_register.
15457 * config/i386/i386.md (indirect_jump): Likewise.
15458 (tablejump): Likewise.
15459 (*sibcall_memory): Likewise.
15460 (*sibcall_value_memory): Likewise.
15461 Peepholes of indirect call and jump via memory: Likewise.
15462 * config/i386/i386.opt: Likewise.
15463 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15464 (GOT_memory_operand): Likewise.
15465 (call_insn_operand): Likewise.
15466 (sibcall_insn_operand): Likewise.
15467 (GOT32_symbol_operand): Likewise.
15469 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15471 PR middle-end/83837
15472 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15473 type rather than type addr's type points to.
15474 (expand_omp_atomic_mutex): Likewise.
15475 (expand_omp_atomic): Likewise.
15477 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15480 * config/i386/i386.c (output_indirect_thunk_function): Use
15481 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15482 for __x86_return_thunk.
15484 2018-01-15 Richard Biener <rguenther@suse.de>
15486 PR middle-end/83850
15487 * expmed.c (extract_bit_field_1): Fix typo.
15489 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15492 * config/arm/iterators.md (VF): New mode iterator.
15493 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15494 Remove integer-related logic from pattern.
15495 (neon_vabd<mode>_3): Likewise.
15497 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15499 PR middle-end/82694
15500 * common.opt (fstrict-overflow): No longer an alias.
15501 (fwrapv-pointer): New option.
15502 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15503 also for pointer types based on flag_wrapv_pointer.
15504 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15505 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15506 opts->x_flag_wrapv got set.
15507 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15508 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15509 POINTER_TYPE_OVERFLOW_UNDEFINED.
15510 * match.pd: Likewise in address comparison pattern.
15511 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15513 2018-01-15 Richard Biener <rguenther@suse.de>
15516 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15517 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15518 Reset type names to their identifier if their TYPE_DECL doesn't
15519 have linkage (and thus is used for ODR and devirt).
15520 (save_debug_info_for_decl): Remove.
15521 (save_debug_info_for_type): Likewise.
15522 (add_tree_to_fld_list): Adjust.
15523 * tree-pretty-print.c (dump_generic_node): Make dumping of
15524 type names more robust.
15526 2018-01-15 Richard Biener <rguenther@suse.de>
15528 * BASE-VER: Bump to 8.0.1.
15530 2018-01-14 Martin Sebor <msebor@redhat.com>
15533 * builtins.c (check_access): Avoid warning when the no-warning bit
15536 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15538 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15539 * ira-color (allocno_hard_regs_compare): Likewise.
15541 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15544 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15545 Use .pushsection/.popsection.
15547 2018-01-14 Martin Sebor <msebor@redhat.com>
15550 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15552 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15554 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15555 entry from extra_headers.
15556 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15557 extra_headers, make the list bitwise identical to the i?86-*-* one.
15559 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15561 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15562 -mcmodel=large with -mindirect-branch=thunk,
15563 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15564 -mfunction-return=thunk-extern.
15565 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15566 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15567 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15569 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15571 * config/i386/i386.c (print_reg): Print the name of the full
15572 integer register without '%'.
15573 (ix86_print_operand): Handle 'V'.
15574 * doc/extend.texi: Document 'V' modifier.
15576 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15578 * config/i386/constraints.md (Bs): Disallow memory operand for
15579 -mindirect-branch-register.
15581 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15582 (GOT_memory_operand): Likewise.
15583 (call_insn_operand): Likewise.
15584 (sibcall_insn_operand): Likewise.
15585 (GOT32_symbol_operand): Likewise.
15586 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15587 for -mindirect-branch-register.
15588 (tablejump): Likewise.
15589 (*sibcall_memory): Likewise.
15590 (*sibcall_value_memory): Likewise.
15591 Disallow peepholes of indirect call and jump via memory for
15592 -mindirect-branch-register.
15593 (*call_pop): Replace m with Bw.
15594 (*call_value_pop): Likewise.
15595 (*sibcall_pop_memory): Replace m with Bs.
15596 * config/i386/i386.opt (mindirect-branch-register): New option.
15597 * doc/invoke.texi: Document -mindirect-branch-register option.
15599 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15601 * config/i386/i386-protos.h (ix86_output_function_return): New.
15602 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15603 set function_return_type.
15604 (indirect_thunk_name): Add ret_p to indicate thunk for function
15606 (output_indirect_thunk_function): Pass false to
15607 indirect_thunk_name.
15608 (ix86_output_indirect_branch_via_reg): Likewise.
15609 (ix86_output_indirect_branch_via_push): Likewise.
15610 (output_indirect_thunk_function): Create alias for function
15611 return thunk if regno < 0.
15612 (ix86_output_function_return): New function.
15613 (ix86_handle_fndecl_attribute): Handle function_return.
15614 (ix86_attribute_table): Add function_return.
15615 * config/i386/i386.h (machine_function): Add
15616 function_return_type.
15617 * config/i386/i386.md (simple_return_internal): Use
15618 ix86_output_function_return.
15619 (simple_return_internal_long): Likewise.
15620 * config/i386/i386.opt (mfunction-return=): New option.
15621 (indirect_branch): Mention -mfunction-return=.
15622 * doc/extend.texi: Document function_return function attribute.
15623 * doc/invoke.texi: Document -mfunction-return= option.
15625 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15627 * config/i386/i386-opts.h (indirect_branch): New.
15628 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15629 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15630 with local indirect jump when converting indirect call and jump.
15631 (ix86_set_indirect_branch_type): New.
15632 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15633 (indirectlabelno): New.
15634 (indirect_thunk_needed): Likewise.
15635 (indirect_thunk_bnd_needed): Likewise.
15636 (indirect_thunks_used): Likewise.
15637 (indirect_thunks_bnd_used): Likewise.
15638 (INDIRECT_LABEL): Likewise.
15639 (indirect_thunk_name): Likewise.
15640 (output_indirect_thunk): Likewise.
15641 (output_indirect_thunk_function): Likewise.
15642 (ix86_output_indirect_branch_via_reg): Likewise.
15643 (ix86_output_indirect_branch_via_push): Likewise.
15644 (ix86_output_indirect_branch): Likewise.
15645 (ix86_output_indirect_jmp): Likewise.
15646 (ix86_code_end): Call output_indirect_thunk_function if needed.
15647 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15649 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15650 (ix86_attribute_table): Add indirect_branch.
15651 * config/i386/i386.h (machine_function): Add indirect_branch_type
15652 and has_local_indirect_jump.
15653 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15655 (tablejump): Likewise.
15656 (*indirect_jump): Use ix86_output_indirect_jmp.
15657 (*tablejump_1): Likewise.
15658 (simple_return_indirect_internal): Likewise.
15659 * config/i386/i386.opt (mindirect-branch=): New option.
15660 (indirect_branch): New.
15663 (thunk-inline): Likewise.
15664 (thunk-extern): Likewise.
15665 * doc/extend.texi: Document indirect_branch function attribute.
15666 * doc/invoke.texi: Document -mindirect-branch= option.
15668 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15671 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15673 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15675 * ipa-inline.c (want_inline_small_function_p): Return false if
15676 inlining has already failed with CIF_FINAL_ERROR.
15677 (update_caller_keys): Call want_inline_small_function_p before
15679 (update_callee_keys): Likewise.
15681 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15683 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15685 (rs6000_quadword_masked_address_p): Likewise.
15686 (quad_aligned_load_p): Likewise.
15687 (quad_aligned_store_p): Likewise.
15688 (const_load_sequence_p): Add comment to describe the outer-most loop.
15689 (mimic_memory_attributes_and_flags): New function.
15690 (rs6000_gen_stvx): Likewise.
15691 (replace_swapped_aligned_store): Likewise.
15692 (rs6000_gen_lvx): Likewise.
15693 (replace_swapped_aligned_load): Likewise.
15694 (replace_swapped_load_constant): Capitalize argument name in
15695 comment describing this function.
15696 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15697 and stores that access quad-word aligned addresses and replace
15698 with stvx or lvx instructions when appropriate.
15699 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15700 New function prototype.
15701 (rs6000_quadword_masked_address_p): Likewise.
15702 (rs6000_gen_lvx): Likewise.
15703 (rs6000_gen_stvx): Likewise.
15704 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15705 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15706 when memory address is aligned.
15707 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15708 this split to select lvx instruction when memory address is aligned.
15709 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15710 instruction when memory address is aligned.
15711 (*vsx_le_perm_load_v16qi): Likewise.
15712 (four unnamed splitters): Modify to select the stvx instruction
15713 when memory is aligned.
15715 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15717 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15718 which appears in the queue multiple times.
15720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15721 Alan Hayward <alan.hayward@arm.com>
15722 David Sherwood <david.sherwood@arm.com>
15724 * tree-vectorizer.h (vec_lower_bound): New structure.
15725 (_loop_vec_info): Add check_nonzero and lower_bounds.
15726 (LOOP_VINFO_CHECK_NONZERO): New macro.
15727 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15728 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15729 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15730 fields. Make seg_len the distance travelled, not including the
15732 (dr_direction_indicator): Declare.
15733 (dr_zero_step_indicator): Likewise.
15734 (dr_known_forward_stride_p): Likewise.
15735 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15737 (runtime_alias_check_p): Allow runtime alias checks with
15739 (operator ==): Compare access_size and align.
15740 (prune_runtime_alias_test_list): Rework for new distinction between
15741 the access_size and seg_len.
15742 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15744 (get_segment_min_max): New function.
15745 (create_intersect_range_checks): Use it.
15746 (dr_step_indicator): New function.
15747 (dr_direction_indicator): Likewise.
15748 (dr_zero_step_indicator): Likewise.
15749 (dr_known_forward_stride_p): Likewise.
15750 * tree-loop-distribution.c (data_ref_segment_size): Return
15751 DR_STEP * (niters - 1).
15752 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15754 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15755 (vect_preserves_scalar_order_p): New function, split out from...
15756 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15757 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15758 (vect_vfa_access_size): New function.
15759 (vect_vfa_align): Likewise.
15760 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15761 (dump_lower_bound): New function.
15762 (vect_check_lower_bound): Likewise.
15763 (vect_small_gap_p): Likewise.
15764 (vectorizable_with_step_bound_p): Likewise.
15765 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15766 depencies if the vectorization factor is 1. Convert the checks
15767 for nonzero steps into checks on the bounds of DR_STEP. Try using
15768 a bunds check for variable steps if the minimum required step is
15769 relatively small. Update calls to the dr_with_seg_len
15770 constructor and to vect_compile_time_alias.
15771 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15773 (vect_loop_versioning): Call it.
15774 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15776 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15778 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15779 Alan Hayward <alan.hayward@arm.com>
15780 David Sherwood <david.sherwood@arm.com>
15782 * doc/sourcebuild.texi (vect_scatter_store): Document.
15783 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15785 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15787 * genopinit.c (main): Add supports_vec_scatter_store and
15788 supports_vec_scatter_store_cached to target_optabs.
15789 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15790 IFN_MASK_SCATTER_STORE.
15791 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15793 * internal-fn.h (internal_store_fn_p): Declare.
15794 (internal_fn_stored_value_index): Likewise.
15795 * internal-fn.c (scatter_store_direct): New macro.
15796 (expand_scatter_store_optab_fn): New function.
15797 (direct_scatter_store_optab_supported_p): New macro.
15798 (internal_store_fn_p): New function.
15799 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15800 IFN_MASK_SCATTER_STORE.
15801 (internal_fn_mask_index): Likewise.
15802 (internal_fn_stored_value_index): New function.
15803 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15804 for scatter stores.
15805 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15806 * optabs-query.c (supports_vec_scatter_store_p): New function.
15807 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15808 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15809 true for scatter stores.
15810 (vect_gather_scatter_fn_p): Handle scatter stores too.
15811 (vect_check_gather_scatter): Consider using scatter stores if
15812 supports_vec_scatter_store_p.
15813 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15814 scatter stores too.
15815 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15816 internal_fn_stored_value_index.
15817 (check_load_store_masking): Handle scatter stores too.
15818 (vect_get_store_rhs): Make public.
15819 (vectorizable_call): Use internal_store_fn_p.
15820 (vectorizable_store): Handle scatter store internal functions.
15821 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15822 when deciding whether the end of the group has been reached.
15823 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15824 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15825 (mask_scatter_store<mode>): New insns.
15827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15828 Alan Hayward <alan.hayward@arm.com>
15829 David Sherwood <david.sherwood@arm.com>
15831 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15832 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15833 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15835 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15836 Use vect_truncate_gather_scatter_offset if we can't treat the
15837 operation as a normal gather load or scatter store.
15838 (get_group_load_store_type): Take the gather_scatter_info
15839 as argument. Try using a gather load or scatter store for
15840 single-element groups.
15841 (get_load_store_type): Update calls to get_group_load_store_type
15842 and vect_use_strided_gather_scatters_p.
15844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15845 Alan Hayward <alan.hayward@arm.com>
15846 David Sherwood <david.sherwood@arm.com>
15848 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15849 optional tree argument.
15850 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15852 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15853 but continue to use the current value as a fallback.
15854 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15855 to compare the updates.
15856 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15857 (get_load_store_type): Use it when handling a strided access.
15858 (vect_get_strided_load_store_ops): New function.
15859 (vect_get_data_ptr_increment): Likewise.
15860 (vectorizable_load): Handle strided gather loads. Always pass
15861 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15864 Alan Hayward <alan.hayward@arm.com>
15865 David Sherwood <david.sherwood@arm.com>
15867 * doc/md.texi (gather_load@var{m}): Document.
15868 (mask_gather_load@var{m}): Likewise.
15869 * genopinit.c (main): Add supports_vec_gather_load and
15870 supports_vec_gather_load_cached to target_optabs.
15871 * optabs-tree.c (init_tree_optimization_optabs): Use
15872 ggc_cleared_alloc to allocate target_optabs.
15873 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15874 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15876 * internal-fn.h (internal_load_fn_p): Declare.
15877 (internal_gather_scatter_fn_p): Likewise.
15878 (internal_fn_mask_index): Likewise.
15879 (internal_gather_scatter_fn_supported_p): Likewise.
15880 * internal-fn.c (gather_load_direct): New macro.
15881 (expand_gather_load_optab_fn): New function.
15882 (direct_gather_load_optab_supported_p): New macro.
15883 (direct_internal_fn_optab): New function.
15884 (internal_load_fn_p): Likewise.
15885 (internal_gather_scatter_fn_p): Likewise.
15886 (internal_fn_mask_index): Likewise.
15887 (internal_gather_scatter_fn_supported_p): Likewise.
15888 * optabs-query.c (supports_at_least_one_mode_p): New function.
15889 (supports_vec_gather_load_p): Likewise.
15890 * optabs-query.h (supports_vec_gather_load_p): Declare.
15891 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15892 and memory_type field.
15893 (NUM_PATTERNS): Bump to 15.
15894 * tree-vect-data-refs.c: Include internal-fn.h.
15895 (vect_gather_scatter_fn_p): New function.
15896 (vect_describe_gather_scatter_call): Likewise.
15897 (vect_check_gather_scatter): Try using internal functions for
15898 gather loads. Recognize existing calls to a gather load function.
15899 (vect_analyze_data_refs): Consider using gather loads if
15900 supports_vec_gather_load_p.
15901 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15902 (vect_get_gather_scatter_offset_type): Likewise.
15903 (vect_convert_mask_for_vectype): Likewise.
15904 (vect_add_conversion_to_patterm): Likewise.
15905 (vect_try_gather_scatter_pattern): Likewise.
15906 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15907 (vect_vect_recog_func_ptrs): Add it.
15908 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15909 internal_fn_mask_index and internal_gather_scatter_fn_p.
15910 (check_load_store_masking): Take the gather_scatter_info as an
15911 argument and handle gather loads.
15912 (vect_get_gather_scatter_ops): New function.
15913 (vectorizable_call): Check internal_load_fn_p.
15914 (vectorizable_load): Likewise. Handle gather load internal
15916 (vectorizable_store): Update call to check_load_store_masking.
15917 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15918 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15919 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15920 (aarch64_gather_scale_operand_d): New predicates.
15921 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15922 (mask_gather_load<mode>): New insns.
15924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15925 Alan Hayward <alan.hayward@arm.com>
15926 David Sherwood <david.sherwood@arm.com>
15928 * optabs.def (fold_left_plus_optab): New optab.
15929 * doc/md.texi (fold_left_plus_@var{m}): Document.
15930 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15931 * internal-fn.c (fold_left_direct): Define.
15932 (expand_fold_left_optab_fn): Likewise.
15933 (direct_fold_left_optab_supported_p): Likewise.
15934 * fold-const-call.c (fold_const_fold_left): New function.
15935 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15936 * tree-parloops.c (valid_reduction_p): New function.
15937 (gather_scalar_reductions): Use it.
15938 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15939 (vect_finish_replace_stmt): Declare.
15940 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15941 (needs_fold_left_reduction_p): New function, split out from...
15942 (vect_is_simple_reduction): ...here. Accept reductions that
15943 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15944 (vect_force_simple_reduction): Also store the reduction type in
15945 the assignment's STMT_VINFO_REDUC_TYPE.
15946 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15947 (merge_with_identity): New function.
15948 (vect_expand_fold_left): Likewise.
15949 (vectorize_fold_left_reduction): Likewise.
15950 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15951 scalar phi in place for it. Check for target support and reject
15952 cases that would reassociate the operation. Defer the transform
15953 phase to vectorize_fold_left_reduction.
15954 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15955 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15956 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15960 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15961 call to ifc_temp_var.
15963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15964 Alan Hayward <alan.hayward@arm.com>
15965 David Sherwood <david.sherwood@arm.com>
15967 * target.def (legitimize_address_displacement): Take the original
15968 offset as a poly_int.
15969 * targhooks.h (default_legitimize_address_displacement): Update
15971 * targhooks.c (default_legitimize_address_displacement): Likewise.
15972 * doc/tm.texi: Regenerate.
15973 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15974 as an argument, moving assert of ad->disp == ad->disp_term to...
15975 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15976 Try calling targetm.legitimize_address_displacement before expanding
15977 the address rather than afterwards, and adjust for the new interface.
15978 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15979 Match the new hook interface. Handle SVE addresses.
15980 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15981 new hook interface.
15983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15985 * Makefile.in (OBJS): Add early-remat.o.
15986 * target.def (select_early_remat_modes): New hook.
15987 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15988 * doc/tm.texi: Regenerate.
15989 * targhooks.h (default_select_early_remat_modes): Declare.
15990 * targhooks.c (default_select_early_remat_modes): New function.
15991 * timevar.def (TV_EARLY_REMAT): New timevar.
15992 * passes.def (pass_early_remat): New pass.
15993 * tree-pass.h (make_pass_early_remat): Declare.
15994 * early-remat.c: New file.
15995 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15997 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16000 Alan Hayward <alan.hayward@arm.com>
16001 David Sherwood <david.sherwood@arm.com>
16003 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16004 vfm1 with a bound_epilog parameter.
16005 (vect_do_peeling): Update calls accordingly, and move the prologue
16006 call earlier in the function. Treat the base bound_epilog as 0 for
16007 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16008 this base when peeling for gaps.
16009 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16010 with fully-masked loops.
16011 (vect_estimate_min_profitable_iters): Handle the single peeled
16012 iteration in that case.
16014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16015 Alan Hayward <alan.hayward@arm.com>
16016 David Sherwood <david.sherwood@arm.com>
16018 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16019 single-element interleaving even if the size is not a power of 2.
16020 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16021 accesses for single-element interleaving if the group size is
16024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16025 Alan Hayward <alan.hayward@arm.com>
16026 David Sherwood <david.sherwood@arm.com>
16028 * doc/md.texi (fold_extract_last_@var{m}): Document.
16029 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16030 * optabs.def (fold_extract_last_optab): New optab.
16031 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16032 * internal-fn.c (fold_extract_direct): New macro.
16033 (expand_fold_extract_optab_fn): Likewise.
16034 (direct_fold_extract_optab_supported_p): Likewise.
16035 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16036 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16037 EXTRACT_LAST_REDUCTION.
16038 (get_initial_def_for_reduction): Do not create an initial vector
16039 for EXTRACT_LAST_REDUCTION reductions.
16040 (vectorizable_reduction): Leave the scalar phi in place for
16041 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16042 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16043 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16044 transform phase to vectorizable_condition.
16045 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16047 (vect_finish_stmt_generation): ...here.
16048 (vect_finish_replace_stmt): New function.
16049 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16050 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16052 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16055 Alan Hayward <alan.hayward@arm.com>
16056 David Sherwood <david.sherwood@arm.com>
16058 * doc/md.texi (extract_last_@var{m}): Document.
16059 * optabs.def (extract_last_optab): New optab.
16060 * internal-fn.def (EXTRACT_LAST): New internal function.
16061 * internal-fn.c (cond_unary_direct): New macro.
16062 (expand_cond_unary_optab_fn): Likewise.
16063 (direct_cond_unary_optab_supported_p): Likewise.
16064 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16065 loops using EXTRACT_LAST.
16066 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16067 (extract_last_<mode>): ...this optab.
16068 (vec_extract<mode><Vel>): Update accordingly.
16070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16071 Alan Hayward <alan.hayward@arm.com>
16072 David Sherwood <david.sherwood@arm.com>
16074 * target.def (empty_mask_is_expensive): New hook.
16075 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16076 * doc/tm.texi: Regenerate.
16077 * targhooks.h (default_empty_mask_is_expensive): Declare.
16078 * targhooks.c (default_empty_mask_is_expensive): New function.
16079 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16080 if the target says that empty masks are expensive.
16081 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16083 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16086 Alan Hayward <alan.hayward@arm.com>
16087 David Sherwood <david.sherwood@arm.com>
16089 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16090 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16091 (vect_use_loop_mask_for_alignment_p): New function.
16092 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16093 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16094 niters_skip argument. Make sure that the first niters_skip elements
16095 of the first iteration are inactive.
16096 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16097 Update call to vect_set_loop_masks_directly.
16098 (get_misalign_in_elems): New function, split out from...
16099 (vect_gen_prolog_loop_niters): ...here.
16100 (vect_update_init_of_dr): Take a code argument that specifies whether
16101 the adjustment should be added or subtracted.
16102 (vect_update_init_of_drs): Likewise.
16103 (vect_prepare_for_masked_peels): New function.
16104 (vect_do_peeling): Skip prologue peeling if we're using a mask
16105 instead. Update call to vect_update_inits_of_drs.
16106 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16108 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16109 alignment. Do not include the number of peeled iterations in
16110 the minimum threshold in that case.
16111 (vectorizable_induction): Adjust the start value down by
16112 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16113 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16114 Take the number of skipped iterations into account when calculating
16116 * tree-vect-stmts.c (vect_gen_while_not): New function.
16118 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16119 Alan Hayward <alan.hayward@arm.com>
16120 David Sherwood <david.sherwood@arm.com>
16122 * doc/sourcebuild.texi (vect_fully_masked): Document.
16123 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16124 default value to 0.
16125 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16127 (vect_analyze_loop_2): ...here. Don't check the vectorization
16128 factor against the number of loop iterations if the loop is
16131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16132 Alan Hayward <alan.hayward@arm.com>
16133 David Sherwood <david.sherwood@arm.com>
16135 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16136 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16137 (dump_groups): Update accordingly.
16138 (iv_use::mem_type): New member variable.
16139 (address_p): New function.
16140 (record_use): Add a mem_type argument and initialize the new
16142 (record_group_use): Add a mem_type argument. Use address_p.
16143 Remove obsolete null checks of base_object. Update call to record_use.
16144 (find_interesting_uses_op): Update call to record_group_use.
16145 (find_interesting_uses_cond): Likewise.
16146 (find_interesting_uses_address): Likewise.
16147 (get_mem_type_for_internal_fn): New function.
16148 (find_address_like_use): Likewise.
16149 (find_interesting_uses_stmt): Try find_address_like_use before
16150 calling find_interesting_uses_op.
16151 (addr_offset_valid_p): Use the iv mem_type field as the type
16152 of the addressed memory.
16153 (add_autoinc_candidates): Likewise.
16154 (get_address_cost): Likewise.
16155 (split_small_address_groups_p): Use address_p.
16156 (split_address_groups): Likewise.
16157 (add_iv_candidate_for_use): Likewise.
16158 (autoinc_possible_for_pair): Likewise.
16159 (rewrite_groups): Likewise.
16160 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16161 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16162 (get_alias_ptr_type_for_ptr_address): New function.
16163 (rewrite_use_address): Rewrite address uses in calls that were
16164 identified by find_address_like_use.
16166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16167 Alan Hayward <alan.hayward@arm.com>
16168 David Sherwood <david.sherwood@arm.com>
16170 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16172 * gimple-expr.h (is_gimple_addressable: Likewise.
16173 * gimple-expr.c (is_gimple_address): Likewise.
16174 * internal-fn.c (expand_call_mem_ref): New function.
16175 (expand_mask_load_optab_fn): Use it.
16176 (expand_mask_store_optab_fn): Likewise.
16178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16179 Alan Hayward <alan.hayward@arm.com>
16180 David Sherwood <david.sherwood@arm.com>
16182 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16183 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16184 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16185 (cond_umax@var{mode}): Document.
16186 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16187 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16188 (cond_umin_optab, cond_umax_optab): New optabs.
16189 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16190 (COND_IOR, COND_XOR): New internal functions.
16191 * internal-fn.h (get_conditional_internal_fn): Declare.
16192 * internal-fn.c (cond_binary_direct): New macro.
16193 (expand_cond_binary_optab_fn): Likewise.
16194 (direct_cond_binary_optab_supported_p): Likewise.
16195 (get_conditional_internal_fn): New function.
16196 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16197 Cope with reduction statements that are vectorized as calls rather
16199 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16200 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16201 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16202 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16203 (UNSPEC_COND_EOR): New unspecs.
16204 (optab): Add mappings for them.
16205 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16206 (sve_int_op, sve_fp_op): New int attributes.
16208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16209 Alan Hayward <alan.hayward@arm.com>
16210 David Sherwood <david.sherwood@arm.com>
16212 * optabs.def (while_ult_optab): New optab.
16213 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16214 * internal-fn.def (WHILE_ULT): New internal function.
16215 * internal-fn.h (direct_internal_fn_supported_p): New override
16216 that takes two types as argument.
16217 * internal-fn.c (while_direct): New macro.
16218 (expand_while_optab_fn): New function.
16219 (convert_optab_supported_p): Likewise.
16220 (direct_while_optab_supported_p): New macro.
16221 * wide-int.h (wi::udiv_ceil): New function.
16222 * tree-vectorizer.h (rgroup_masks): New structure.
16223 (vec_loop_masks): New typedef.
16224 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16225 and fully_masked_p.
16226 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16227 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16228 (vect_max_vf): New function.
16229 (slpeel_make_loop_iterate_ntimes): Delete.
16230 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16231 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16232 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16233 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16234 internal-fn.h, stor-layout.h and optabs-query.h.
16235 (vect_set_loop_mask): New function.
16236 (add_preheader_seq): Likewise.
16237 (add_header_seq): Likewise.
16238 (interleave_supported_p): Likewise.
16239 (vect_maybe_permute_loop_masks): Likewise.
16240 (vect_set_loop_masks_directly): Likewise.
16241 (vect_set_loop_condition_masked): Likewise.
16242 (vect_set_loop_condition_unmasked): New function, split out from
16243 slpeel_make_loop_iterate_ntimes.
16244 (slpeel_make_loop_iterate_ntimes): Rename to..
16245 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16246 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16247 (vect_do_peeling): Update call accordingly.
16248 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16250 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16251 mask_compare_type, can_fully_mask_p and fully_masked_p.
16252 (release_vec_loop_masks): New function.
16253 (_loop_vec_info): Use it to free the loop masks.
16254 (can_produce_all_loop_masks_p): New function.
16255 (vect_get_max_nscalars_per_iter): Likewise.
16256 (vect_verify_full_masking): Likewise.
16257 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16258 retries, and free the mask rgroups before retrying. Check loop-wide
16259 reasons for disallowing fully-masked loops. Make the final decision
16260 about whether use a fully-masked loop or not.
16261 (vect_estimate_min_profitable_iters): Do not assume that peeling
16262 for the number of iterations will be needed for fully-masked loops.
16263 (vectorizable_reduction): Disable fully-masked loops.
16264 (vectorizable_live_operation): Likewise.
16265 (vect_halve_mask_nunits): New function.
16266 (vect_double_mask_nunits): Likewise.
16267 (vect_record_loop_mask): Likewise.
16268 (vect_get_loop_mask): Likewise.
16269 (vect_transform_loop): Handle the case in which the final loop
16270 iteration might handle a partial vector. Call vect_set_loop_condition
16271 instead of slpeel_make_loop_iterate_ntimes.
16272 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16273 (check_load_store_masking): New function.
16274 (prepare_load_store_mask): Likewise.
16275 (vectorizable_store): Handle fully-masked loops.
16276 (vectorizable_load): Likewise.
16277 (supportable_widening_operation): Use vect_halve_mask_nunits for
16279 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16280 (vect_gen_while): New function.
16281 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16282 (aarch64_uqdec<mode>): New insn.
16284 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16285 Alan Hayward <alan.hayward@arm.com>
16286 David Sherwood <david.sherwood@arm.com>
16288 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16289 (reduc_xor_scal_optab): New optabs.
16290 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16291 (reduc_xor_scal_@var{m}): Document.
16292 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16293 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16294 internal functions.
16295 * fold-const-call.c (fold_const_call): Handle them.
16296 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16297 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16298 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16299 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16300 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16301 (UNSPEC_XORV): New unspecs.
16302 (optab): Add entries for them.
16303 (BITWISEV): New int iterator.
16304 (bit_reduc_op): New int attributes.
16306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16307 Alan Hayward <alan.hayward@arm.com>
16308 David Sherwood <david.sherwood@arm.com>
16310 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16311 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16312 * optabs.def (vec_shl_insert_optab): New optab.
16313 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16314 (duplicate_and_interleave): Likewise.
16315 * tree-vect-loop.c: Include internal-fn.h.
16316 (neutral_op_for_slp_reduction): New function, split out from
16317 get_initial_defs_for_reduction.
16318 (get_initial_def_for_reduction): Handle option 2 for variable-length
16319 vectors by loading the neutral value into a vector and then shifting
16320 the initial value into element 0.
16321 (get_initial_defs_for_reduction): Replace the code argument with
16322 the neutral value calculated by neutral_op_for_slp_reduction.
16323 Use gimple_build_vector for constant-length vectors.
16324 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16325 but the first group_size elements have a neutral value.
16326 Use duplicate_and_interleave otherwise.
16327 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16328 Update call to get_initial_defs_for_reduction. Handle SLP
16329 reductions for variable-length vectors by creating one vector
16330 result for each scalar result, with the elements associated
16331 with other scalar results stubbed out with the neutral value.
16332 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16333 Require IFN_VEC_SHL_INSERT for double reductions on
16334 variable-length vectors, or SLP reductions that have
16335 a neutral value. Require can_duplicate_and_interleave_p
16336 support for variable-length unchained SLP reductions if there
16337 is no neutral value, such as for MIN/MAX reductions. Also require
16338 the number of vector elements to be a multiple of the number of
16339 SLP statements when doing variable-length unchained SLP reductions.
16340 Update call to vect_create_epilog_for_reduction.
16341 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16342 and remove initial values.
16343 (duplicate_and_interleave): Make public.
16344 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16345 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16348 Alan Hayward <alan.hayward@arm.com>
16349 David Sherwood <david.sherwood@arm.com>
16351 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16352 (can_duplicate_and_interleave_p): New function.
16353 (vect_get_and_check_slp_defs): Take the vector of statements
16354 rather than just the current one. Remove excess parentheses.
16355 Restriction rejectinon of vect_constant_def and vect_external_def
16356 for variable-length vectors to boolean types, or types for which
16357 can_duplicate_and_interleave_p is false.
16358 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16359 (duplicate_and_interleave): New function.
16360 (vect_get_constant_vectors): Use gimple_build_vector for
16361 constant-length vectors and suitable variable-length constant
16362 vectors. Use duplicate_and_interleave for other variable-length
16363 vectors. Don't defer the update when inserting new statements.
16365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16366 Alan Hayward <alan.hayward@arm.com>
16367 David Sherwood <david.sherwood@arm.com>
16369 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16370 min_profitable_iters doesn't go negative.
16372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16373 Alan Hayward <alan.hayward@arm.com>
16374 David Sherwood <david.sherwood@arm.com>
16376 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16377 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16378 * optabs.def (vec_mask_load_lanes_optab): New optab.
16379 (vec_mask_store_lanes_optab): Likewise.
16380 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16381 (MASK_STORE_LANES): Likewise.
16382 * internal-fn.c (mask_load_lanes_direct): New macro.
16383 (mask_store_lanes_direct): Likewise.
16384 (expand_mask_load_optab_fn): Handle masked operations.
16385 (expand_mask_load_lanes_optab_fn): New macro.
16386 (expand_mask_store_optab_fn): Handle masked operations.
16387 (expand_mask_store_lanes_optab_fn): New macro.
16388 (direct_mask_load_lanes_optab_supported_p): Likewise.
16389 (direct_mask_store_lanes_optab_supported_p): Likewise.
16390 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16392 (vect_load_lanes_supported): Likewise.
16393 * tree-vect-data-refs.c (strip_conversion): New function.
16394 (can_group_stmts_p): Likewise.
16395 (vect_analyze_data_ref_accesses): Use it instead of checking
16396 for a pair of assignments.
16397 (vect_store_lanes_supported): Take a masked_p parameter.
16398 (vect_load_lanes_supported): Likewise.
16399 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16400 vect_store_lanes_supported and vect_load_lanes_supported.
16401 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16402 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16403 parameter. Don't allow gaps for masked accesses.
16404 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16405 and vect_load_lanes_supported.
16406 (get_load_store_type): Take a masked_p parameter and update
16407 call to get_group_load_store_type.
16408 (vectorizable_store): Update call to get_load_store_type.
16409 Handle IFN_MASK_STORE_LANES.
16410 (vectorizable_load): Update call to get_load_store_type.
16411 Handle IFN_MASK_LOAD_LANES.
16413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16414 Alan Hayward <alan.hayward@arm.com>
16415 David Sherwood <david.sherwood@arm.com>
16417 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16419 * config/aarch64/aarch64-protos.h
16420 (aarch64_sve_struct_memory_operand_p): Declare.
16421 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16422 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16423 (VPRED, vpred): Handle SVE structure modes.
16424 * config/aarch64/constraints.md (Utx): New constraint.
16425 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16426 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16427 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16428 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16429 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16430 structure modes. Split into pieces after RA.
16431 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16432 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16434 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16435 SVE structure modes.
16436 (aarch64_classify_address): Likewise.
16437 (sizetochar): Move earlier in file.
16438 (aarch64_print_operand): Handle SVE register lists.
16439 (aarch64_array_mode): New function.
16440 (aarch64_sve_struct_memory_operand_p): Likewise.
16441 (TARGET_ARRAY_MODE): Redefine.
16443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16444 Alan Hayward <alan.hayward@arm.com>
16445 David Sherwood <david.sherwood@arm.com>
16447 * target.def (array_mode): New target hook.
16448 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16449 * doc/tm.texi: Regenerate.
16450 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16451 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16452 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16453 targetm.array_mode.
16454 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16458 Alan Hayward <alan.hayward@arm.com>
16459 David Sherwood <david.sherwood@arm.com>
16461 * fold-const.c (fold_binary_loc): Check the argument types
16462 rather than the result type when testing for a vector operation.
16464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16466 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16467 * doc/tm.texi: Regenerate.
16469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16470 Alan Hayward <alan.hayward@arm.com>
16471 David Sherwood <david.sherwood@arm.com>
16473 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16474 (sve): Document new AArch64 extension.
16475 * doc/md.texi (w): Extend the description of the AArch64
16476 constraint to include SVE vectors.
16477 (Upl, Upa): Document new AArch64 predicate constraints.
16478 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16480 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16481 (msve-vector-bits=): New option.
16482 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16483 SVE when these are disabled.
16484 (sve): New extension.
16485 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16486 modes. Adjust their number of units based on aarch64_sve_vg.
16487 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16488 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16489 aarch64_addr_query_type.
16490 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16491 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16492 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16493 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16494 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16495 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16496 (aarch64_simd_imm_zero_p): Delete.
16497 (aarch64_check_zero_based_sve_index_immediate): Declare.
16498 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16499 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16500 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16501 (aarch64_sve_float_mul_immediate_p): Likewise.
16502 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16503 rather than an rtx.
16504 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16505 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16506 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16507 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16508 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16509 (aarch64_regmode_natural_size): Likewise.
16510 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16511 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16513 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16514 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16515 for VG and the SVE predicate registers.
16516 (V_ALIASES): Add a "z"-prefixed alias.
16517 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16518 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16519 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16520 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16521 (REG_CLASS_NAMES): Add entries for them.
16522 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16523 and the predicate registers.
16524 (aarch64_sve_vg): Declare.
16525 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16526 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16527 (REGMODE_NATURAL_SIZE): Define.
16528 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16530 * config/aarch64/aarch64.c: Include cfgrtl.h.
16531 (simd_immediate_info): Add a constructor for series vectors,
16532 and an associated step field.
16533 (aarch64_sve_vg): New variable.
16534 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16535 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16536 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16537 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16538 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16539 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16540 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16541 (aarch64_get_mask_mode): New functions.
16542 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16543 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16544 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16545 predicate modes and predicate registers. Explicitly restrict
16546 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16547 to store a vector mode if it is recognized by
16548 aarch64_classify_vector_mode.
16549 (aarch64_regmode_natural_size): New function.
16550 (aarch64_hard_regno_caller_save_mode): Return the original mode
16552 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16553 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16554 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16555 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16557 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16558 does not overlap dest if the function is frame-related. Handle
16560 (aarch64_split_add_offset): New function.
16561 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16562 them aarch64_add_offset.
16563 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16564 and update call to aarch64_sub_sp.
16565 (aarch64_add_cfa_expression): New function.
16566 (aarch64_expand_prologue): Pass extra temporary registers to the
16567 functions above. Handle the case in which we need to emit new
16568 DW_CFA_expressions for registers that were originally saved
16569 relative to the stack pointer, but now have to be expressed
16570 relative to the frame pointer.
16571 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16573 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16574 IP0 and IP1 values for SVE frames.
16575 (aarch64_expand_vec_series): New function.
16576 (aarch64_expand_sve_widened_duplicate): Likewise.
16577 (aarch64_expand_sve_const_vector): Likewise.
16578 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16579 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16580 into the register, rather than emitting a SET directly.
16581 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16582 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16583 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16584 (offset_9bit_signed_scaled_p): New functions.
16585 (aarch64_replicate_bitmask_imm): New function.
16586 (aarch64_bitmask_imm): Use it.
16587 (aarch64_cannot_force_const_mem): Reject expressions involving
16588 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16589 (aarch64_classify_index): Handle SVE indices, by requiring
16590 a plain register index with a scale that matches the element size.
16591 (aarch64_classify_address): Handle SVE addresses. Assert that
16592 the mode of the address is VOIDmode or an integer mode.
16593 Update call to aarch64_classify_symbol.
16594 (aarch64_classify_symbolic_expression): Update call to
16595 aarch64_classify_symbol.
16596 (aarch64_const_vec_all_in_range_p): New function.
16597 (aarch64_print_vector_float_operand): Likewise.
16598 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16599 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16600 and the FP immediates 1.0 and 0.5.
16601 (aarch64_print_address_internal): Handle SVE addresses.
16602 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16603 (aarch64_regno_regclass): Handle predicate registers.
16604 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16606 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16607 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16608 (aarch64_convert_sve_vector_bits): New function.
16609 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16610 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16611 rather than an rtx.
16612 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16613 Handle SVE vector and predicate modes. Accept VL-based constants
16614 that need only one temporary register, and VL offsets that require
16615 no temporary registers.
16616 (aarch64_conditional_register_usage): Mark the predicate registers
16617 as fixed if SVE isn't available.
16618 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16619 Return true for SVE vector and predicate modes.
16620 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16621 rather than an unsigned int. Handle SVE modes.
16622 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16624 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16626 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16627 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16628 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16629 (aarch64_sve_float_mul_immediate_p): New functions.
16630 (aarch64_sve_valid_immediate): New function.
16631 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16632 Explicitly reject structure modes. Check for INDEX constants.
16633 Handle PTRUE and PFALSE constants.
16634 (aarch64_check_zero_based_sve_index_immediate): New function.
16635 (aarch64_simd_imm_zero_p): Delete.
16636 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16637 vector modes. Accept constants in the range of CNT[BHWD].
16638 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16639 ask for an Advanced SIMD mode.
16640 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16641 (aarch64_simd_vector_alignment): Handle SVE predicates.
16642 (aarch64_vectorize_preferred_vector_alignment): New function.
16643 (aarch64_simd_vector_alignment_reachable): Use it instead of
16645 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16646 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16648 (MAX_VECT_LEN): Delete.
16649 (expand_vec_perm_d): Add a vec_flags field.
16650 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16651 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16652 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16654 (aarch64_evpc_rev): Rename to...
16655 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16656 (aarch64_evpc_rev_global): New function.
16657 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16658 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16660 (aarch64_evpc_sve_tbl): New function.
16661 (aarch64_expand_vec_perm_const_1): Update after rename of
16662 aarch64_evpc_rev. Handle SVE permutes too, trying
16663 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16664 than aarch64_evpc_tbl.
16665 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16666 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16667 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16668 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16669 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16670 (aarch64_expand_sve_vcond): New functions.
16671 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16672 of aarch64_vector_mode_p.
16673 (aarch64_dwarf_poly_indeterminate_value): New function.
16674 (aarch64_compute_pressure_classes): Likewise.
16675 (aarch64_can_change_mode_class): Likewise.
16676 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16677 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16678 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16679 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16680 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16681 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16682 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16683 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16685 (Dn, Dl, Dr): Accept const as well as const_vector.
16686 (Dz): Likewise. Compare against CONST0_RTX.
16687 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16688 of "vector" where appropriate.
16689 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16690 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16691 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16692 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16693 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16694 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16695 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16696 (v_int_equiv): Extend to SVE modes.
16697 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16699 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16700 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16701 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16702 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16703 (SVE_COND_FP_CMP): New int iterators.
16704 (perm_hilo): Handle the new unpack unspecs.
16705 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16707 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16708 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16709 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16710 (aarch64_equality_operator, aarch64_constant_vector_operand)
16711 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16712 (aarch64_sve_nonimmediate_operand): Likewise.
16713 (aarch64_sve_general_operand): Likewise.
16714 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16715 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16716 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16717 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16718 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16719 (aarch64_sve_float_arith_immediate): Likewise.
16720 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16721 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16722 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16723 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16724 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16725 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16726 (aarch64_sve_float_arith_operand): Likewise.
16727 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16728 (aarch64_sve_float_mul_operand): Likewise.
16729 (aarch64_sve_vec_perm_operand): Likewise.
16730 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16731 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16732 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16733 as well as const_vector.
16734 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16735 in file. Use CONST0_RTX and CONSTM1_RTX.
16736 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16737 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16738 Use aarch64_simd_imm_zero.
16739 * config/aarch64/aarch64-sve.md: New file.
16740 * config/aarch64/aarch64.md: Include it.
16741 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16742 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16743 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16744 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16745 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16746 (sve): New attribute.
16747 (enabled): Disable instructions with the sve attribute unless
16749 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16750 aarch64_expand_mov_immediate.
16751 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16752 CNT[BHSD] immediates.
16753 (movti): Split CONST_POLY_INT moves into two halves.
16754 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16755 Split additions that need a temporary here if the destination
16756 is the stack pointer.
16757 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16758 (*add<mode>3_poly_1): New instruction.
16759 (set_clobber_cc): New expander.
16761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16763 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16764 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16765 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16766 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16767 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16768 Change innermode from fixed_mode_size to machine_mode.
16769 (simplify_subreg): Update call accordingly. Handle a constant-sized
16770 subreg of a variable-length CONST_VECTOR.
16772 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16773 Alan Hayward <alan.hayward@arm.com>
16774 David Sherwood <david.sherwood@arm.com>
16776 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16777 (add_offset_to_base): New function, split out from...
16778 (create_mem_ref): ...here. When handling a scale other than 1,
16779 check first whether the address is valid without the offset.
16780 Add it into the base if so, leaving the index and scale as-is.
16782 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16785 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16786 fold_for_warn before checking if arg2 is INTEGER_CST.
16788 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16790 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16791 (store_multiple_operation): Delete.
16792 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16793 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16794 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16795 guarded by TARGET_STRING.
16796 (rs6000_output_load_multiple): Delete.
16797 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16798 OPTION_MASK_STRING / TARGET_STRING handling.
16799 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16800 (const rs6000_opt_masks) <"string">: Change mask to 0.
16801 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16802 (MASK_STRING): Delete.
16803 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16805 (load_multiple): Delete.
16812 (store_multiple): Delete.
16819 (movmemsi_8reg): Delete.
16820 (corresponding unnamed define_insn): Delete.
16821 (movmemsi_6reg): Delete.
16822 (corresponding unnamed define_insn): Delete.
16823 (movmemsi_4reg): Delete.
16824 (corresponding unnamed define_insn): Delete.
16825 (movmemsi_2reg): Delete.
16826 (corresponding unnamed define_insn): Delete.
16827 (movmemsi_1reg): Delete.
16828 (corresponding unnamed define_insn): Delete.
16829 * config/rs6000/rs6000.opt (mno-string): New.
16830 (mstring): Replace by deprecation warning stub.
16831 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16833 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16835 * regrename.c (regrename_do_replace): If replacing the same
16836 reg multiple times, try to reuse last created gen_raw_REG.
16839 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16840 main to workaround a bug in GDB.
16842 2018-01-12 Tom de Vries <tom@codesourcery.com>
16845 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16847 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16849 PR rtl-optimization/80481
16850 * ira-color.c (get_cap_member): New function.
16851 (allocnos_conflict_by_live_ranges_p): Use it.
16852 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16853 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16855 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16858 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16859 (*saddl_se_1): Ditto.
16860 (*ssubsi_1): Ditto.
16861 (*ssubl_se_1): Ditto.
16863 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16865 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16866 rather than wi::to_widest for DR_INITs.
16867 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16868 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16869 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16871 (vect_analyze_group_access_1): Note that here.
16873 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16875 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16876 polynomial type sizes.
16878 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16880 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16881 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16882 (gimple_add_tmp_var): Likewise.
16884 2018-01-12 Martin Liska <mliska@suse.cz>
16886 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16887 (gimple_alloc_sizes): Likewise.
16888 (dump_gimple_statistics): Use PRIu64 in printf format.
16889 * gimple.h: Change uint64_t to int.
16891 2018-01-12 Martin Liska <mliska@suse.cz>
16893 * tree-core.h: Use uint64_t instead of int.
16894 * tree.c (tree_node_counts): Likewise.
16895 (tree_node_sizes): Likewise.
16896 (dump_tree_statistics): Use PRIu64 in printf format.
16898 2018-01-12 Martin Liska <mliska@suse.cz>
16900 * Makefile.in: As qsort_chk is implemented in vec.c, add
16901 vec.o to linkage of gencfn-macros.
16902 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16903 passing the info to record_node_allocation_statistics.
16904 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16906 * ggc-common.c (struct ggc_usage): Add operator== and use
16907 it in operator< and compare function.
16908 * mem-stats.h (struct mem_usage): Likewise.
16909 * vec.c (struct vec_usage): Remove operator< and compare
16910 function. Can be simply inherited.
16912 2018-01-12 Martin Jambor <mjambor@suse.cz>
16915 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16916 * tree-ssa-math-opts.c: Include domwalk.h.
16917 (convert_mult_to_fma_1): New function.
16918 (fma_transformation_info): New type.
16919 (fma_deferring_state): Likewise.
16920 (cancel_fma_deferring): New function.
16921 (result_of_phi): Likewise.
16922 (last_fma_candidate_feeds_initial_phi): Likewise.
16923 (convert_mult_to_fma): Added deferring logic, split actual
16924 transformation to convert_mult_to_fma_1.
16925 (math_opts_dom_walker): New type.
16926 (math_opts_dom_walker::after_dom_children): New method, body moved
16927 here from pass_optimize_widening_mul::execute, added deferring logic
16929 (pass_optimize_widening_mul::execute): Moved most of code to
16930 math_opts_dom_walker::after_dom_children.
16931 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16932 * config/i386/i386.c (ix86_option_override_internal): Added
16933 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16935 2018-01-12 Richard Biener <rguenther@suse.de>
16938 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16939 inline instance vars.
16941 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16944 * config/rx/rx.c (rx_is_restricted_memory_address):
16945 Handle SUBREG case.
16947 2018-01-12 Richard Biener <rguenther@suse.de>
16949 PR tree-optimization/80846
16950 * target.def (split_reduction): New target hook.
16951 * targhooks.c (default_split_reduction): New function.
16952 * targhooks.h (default_split_reduction): Declare.
16953 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16954 target requests first reduce vectors by combining low and high
16956 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16957 (get_vectype_for_scalar_type_and_size): Export.
16958 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16959 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16960 * doc/tm.texi: Regenerate.
16961 * config/i386/i386.c (ix86_split_reduction): Implement
16962 TARGET_VECTORIZE_SPLIT_REDUCTION.
16964 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16967 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16968 in PIC mode except for TARGET_VXWORKS_RTP.
16969 * config/sparc/sparc.c: Include cfgrtl.h.
16970 (TARGET_INIT_PIC_REG): Define.
16971 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16972 (sparc_pic_register_p): New predicate.
16973 (sparc_legitimate_address_p): Use it.
16974 (sparc_legitimize_pic_address): Likewise.
16975 (sparc_delegitimize_address): Likewise.
16976 (sparc_mode_dependent_address_p): Likewise.
16977 (gen_load_pcrel_sym): Remove 4th parameter.
16978 (load_got_register): Adjust call to above. Remove obsolete stuff.
16979 (sparc_expand_prologue): Do not call load_got_register here.
16980 (sparc_flat_expand_prologue): Likewise.
16981 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16982 (sparc_use_pseudo_pic_reg): New function.
16983 (sparc_init_pic_reg): Likewise.
16984 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16985 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16987 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16989 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16990 Add item for branch_cost.
16992 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16994 PR rtl-optimization/83565
16995 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16996 not extend the result to a larger mode for rotate operations.
16997 (num_sign_bit_copies1): Likewise.
16999 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17002 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17004 Use values-Xc.o for -pedantic.
17005 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17007 2018-01-12 Martin Liska <mliska@suse.cz>
17010 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17012 (possible_polymorphic_call_targets): Use it.
17013 (ipa_devirt): Likewise.
17015 2018-01-12 Martin Liska <mliska@suse.cz>
17017 * profile-count.h (enum profile_quality): Use 0 as invalid
17018 enum value of profile_quality.
17020 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17022 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17023 -mext-string options.
17025 2018-01-12 Richard Biener <rguenther@suse.de>
17027 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17028 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17029 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17031 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17033 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17035 * configure.ac (--with-long-double-format): Add support for the
17036 configuration option to change the default long double format on
17038 * config.gcc (powerpc*-linux*-*): Likewise.
17039 * configure: Regenerate.
17040 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17041 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17042 used without modification.
17044 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17046 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17047 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17048 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17049 MISC_BUILTIN_SPEC_BARRIER.
17050 (rs6000_init_builtins): Likewise.
17051 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17053 (speculation_barrier): New define_insn.
17054 * doc/extend.texi: Document __builtin_speculation_barrier.
17056 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17059 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17060 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17061 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17063 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17064 integral modes instead of "ss" and "sd".
17065 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17066 vectors with 32-bit and 64-bit elements.
17067 (vecdupssescalarmodesuffix): New mode attribute.
17068 (vec_dup<mode>): Use it.
17070 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17073 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17074 frame if argument is passed on stack.
17076 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17079 * ree.c (combine_reaching_defs): Optimize also
17080 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17081 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17083 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17085 PR middle-end/83189
17086 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17088 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17090 PR middle-end/83718
17091 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17092 after they are computed.
17094 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17096 PR tree-optimization/83695
17097 * gimple-loop-linterchange.cc
17098 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17099 reset cached scev information after interchange.
17100 (pass_linterchange::execute): Remove call to scev_reset_htab.
17102 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17104 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17105 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17106 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17107 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17108 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17109 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17110 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17111 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17112 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17113 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17114 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17115 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17116 (V_lane_reg): Likewise.
17117 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17119 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17120 (vfmal_lane_low<mode>_intrinsic,
17121 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17122 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17123 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17124 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17125 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17126 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17128 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17130 * config/arm/arm-cpus.in (fp16fml): New feature.
17131 (ALL_SIMD): Add fp16fml.
17132 (armv8.2-a): Add fp16fml as an option.
17133 (armv8.3-a): Likewise.
17134 (armv8.4-a): Add fp16fml as part of fp16.
17135 * config/arm/arm.h (TARGET_FP16FML): Define.
17136 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17138 * config/arm/arm-modes.def (V2HF): Define.
17139 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17140 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17141 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17142 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17143 vfmsl_low, vfmsl_high): New set of builtins.
17144 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17145 (vfml_op): New code attribute.
17146 (VFMLHALVES): New int iterator.
17147 (VFML, VFMLSEL): New mode attributes.
17148 (V_reg): Define mapping for V2HF.
17149 (V_hi, V_lo): New mode attributes.
17150 (VF_constraint): Likewise.
17151 (vfml_half, vfml_half_selector): New int attributes.
17152 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17154 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17155 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17157 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17158 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17159 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17160 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17162 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17163 Document new effective target and option set.
17165 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17167 * config/arm/arm-cpus.in (armv8_4): New feature.
17168 (ARMv8_4a): New fgroup.
17169 (armv8.4-a): New arch.
17170 * config/arm/arm-tables.opt: Regenerate.
17171 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17172 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17173 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17174 Add matching rules for -march=armv8.4-a and extensions.
17175 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17177 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17180 * config/rx/rx.md (BW): New mode attribute.
17181 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17183 2018-01-11 Richard Biener <rguenther@suse.de>
17185 PR tree-optimization/83435
17186 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17187 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17188 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17190 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17191 Alan Hayward <alan.hayward@arm.com>
17192 David Sherwood <david.sherwood@arm.com>
17194 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17196 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17197 (aarch64_print_address_internal): Use it to check for a zero offset.
17199 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17200 Alan Hayward <alan.hayward@arm.com>
17201 David Sherwood <david.sherwood@arm.com>
17203 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17204 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17205 Return a poly_int64 rather than a HOST_WIDE_INT.
17206 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17207 rather than a HOST_WIDE_INT.
17208 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17209 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17210 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17211 final_offset from HOST_WIDE_INT to poly_int64.
17212 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17213 to_constant when getting the number of units in an Advanced SIMD
17215 (aarch64_builtin_vectorized_function): Check for a constant number
17217 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17219 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17220 attribute instead of GET_MODE_NUNITS.
17221 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17222 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17223 GET_MODE_SIZE for fixed-size registers.
17224 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17225 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17226 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17227 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17228 (aarch64_print_operand, aarch64_print_address_internal)
17229 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17230 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17231 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17232 Handle polynomial GET_MODE_SIZE.
17233 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17234 wider than SImode without modification.
17235 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17236 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17237 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17238 passing and returning SVE modes.
17239 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17240 rather than GEN_INT.
17241 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17242 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17243 (aarch64_allocate_and_probe_stack_space): Likewise.
17244 (aarch64_layout_frame): Cope with polynomial offsets.
17245 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17246 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17247 polynomial offsets.
17248 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17249 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17250 poly_int64 rather than a HOST_WIDE_INT.
17251 (aarch64_get_separate_components, aarch64_process_components)
17252 (aarch64_expand_prologue, aarch64_expand_epilogue)
17253 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17254 (aarch64_anchor_offset): New function, split out from...
17255 (aarch64_legitimize_address): ...here.
17256 (aarch64_builtin_vectorization_cost): Handle polynomial
17257 TYPE_VECTOR_SUBPARTS.
17258 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17260 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17261 number of elements from the PARALLEL rather than the mode.
17262 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17263 rather than GET_MODE_BITSIZE.
17264 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17265 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17266 (aarch64_expand_vec_perm_const_1): Handle polynomial
17267 d->perm.length () and d->perm elements.
17268 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17269 Apply to_constant to d->perm elements.
17270 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17271 polynomial CONST_VECTOR_NUNITS.
17272 (aarch64_move_pointer): Take amount as a poly_int64 rather
17274 (aarch64_progress_pointer): Avoid temporary variable.
17275 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17276 the mode attribute instead of GET_MODE.
17278 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17279 Alan Hayward <alan.hayward@arm.com>
17280 David Sherwood <david.sherwood@arm.com>
17282 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17283 x exists before using it.
17284 (aarch64_add_constant_internal): Rename to...
17285 (aarch64_add_offset_1): ...this. Replace regnum with separate
17286 src and dest rtxes. Handle the case in which they're different,
17287 including when the offset is zero. Replace scratchreg with an rtx.
17288 Use 2 additions if there is no spare register into which we can
17289 move a 16-bit constant.
17290 (aarch64_add_constant): Delete.
17291 (aarch64_add_offset): Replace reg with separate src and dest
17292 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17293 Use aarch64_add_offset_1.
17294 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17295 an rtx rather than an int. Take the delta as a poly_int64
17296 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17297 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17298 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17299 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17300 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17301 and aarch64_add_sp.
17302 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17303 aarch64_add_constant.
17305 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17307 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17308 Use scalar_float_mode.
17310 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17312 * config/aarch64/aarch64-simd.md
17313 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17314 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17315 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17316 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17317 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17318 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17319 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17320 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17321 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17322 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17324 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17327 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17328 targ_options->x_arm_arch_string is non NULL.
17330 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17332 * config/aarch64/aarch64.h
17333 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17335 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17338 * expmed.c (emit_store_flag_force): Swap if const op0
17339 and change VOIDmode to mode of op0.
17341 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17343 PR rtl-optimization/83761
17344 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17345 than bytes to mode_for_size.
17347 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17349 PR middle-end/83189
17350 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17351 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17354 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17356 PR middle-end/83575
17357 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17358 when in layout mode.
17359 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17360 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17363 2018-01-10 Michael Collison <michael.collison@arm.com>
17365 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17366 * config/aarch64/aarch64-option-extension.def: Add
17367 AARCH64_OPT_EXTENSION of 'fp16fml'.
17368 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17369 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17370 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17371 * config/aarch64/constraints.md (Ui7): New constraint.
17372 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17373 (VFMLA_SEL_W): Ditto.
17376 (VFMLA16_LOW): New int iterator.
17377 (VFMLA16_HIGH): Ditto.
17378 (UNSPEC_FMLAL): New unspec.
17379 (UNSPEC_FMLSL): Ditto.
17380 (UNSPEC_FMLAL2): Ditto.
17381 (UNSPEC_FMLSL2): Ditto.
17382 (f16mac): New code attribute.
17383 * config/aarch64/aarch64-simd-builtins.def
17384 (aarch64_fmlal_lowv2sf): Ditto.
17385 (aarch64_fmlsl_lowv2sf): Ditto.
17386 (aarch64_fmlalq_lowv4sf): Ditto.
17387 (aarch64_fmlslq_lowv4sf): Ditto.
17388 (aarch64_fmlal_highv2sf): Ditto.
17389 (aarch64_fmlsl_highv2sf): Ditto.
17390 (aarch64_fmlalq_highv4sf): Ditto.
17391 (aarch64_fmlslq_highv4sf): Ditto.
17392 (aarch64_fmlal_lane_lowv2sf): Ditto.
17393 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17394 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17395 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17396 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17397 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17398 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17399 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17400 (aarch64_fmlal_lane_highv2sf): Ditto.
17401 (aarch64_fmlsl_lane_highv2sf): Ditto.
17402 (aarch64_fmlal_laneq_highv2sf): Ditto.
17403 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17404 (aarch64_fmlalq_lane_highv4sf): Ditto.
17405 (aarch64_fmlsl_lane_highv4sf): Ditto.
17406 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17407 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17408 * config/aarch64/aarch64-simd.md:
17409 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17410 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17411 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17412 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17413 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17414 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17415 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17416 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17417 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17418 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17419 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17420 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17421 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17422 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17423 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17424 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17425 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17426 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17427 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17428 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17429 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17430 (vfmlsl_low_u32): Ditto.
17431 (vfmlalq_low_u32): Ditto.
17432 (vfmlslq_low_u32): Ditto.
17433 (vfmlal_high_u32): Ditto.
17434 (vfmlsl_high_u32): Ditto.
17435 (vfmlalq_high_u32): Ditto.
17436 (vfmlslq_high_u32): Ditto.
17437 (vfmlal_lane_low_u32): Ditto.
17438 (vfmlsl_lane_low_u32): Ditto.
17439 (vfmlal_laneq_low_u32): Ditto.
17440 (vfmlsl_laneq_low_u32): Ditto.
17441 (vfmlalq_lane_low_u32): Ditto.
17442 (vfmlslq_lane_low_u32): Ditto.
17443 (vfmlalq_laneq_low_u32): Ditto.
17444 (vfmlslq_laneq_low_u32): Ditto.
17445 (vfmlal_lane_high_u32): Ditto.
17446 (vfmlsl_lane_high_u32): Ditto.
17447 (vfmlal_laneq_high_u32): Ditto.
17448 (vfmlsl_laneq_high_u32): Ditto.
17449 (vfmlalq_lane_high_u32): Ditto.
17450 (vfmlslq_lane_high_u32): Ditto.
17451 (vfmlalq_laneq_high_u32): Ditto.
17452 (vfmlslq_laneq_high_u32): Ditto.
17453 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17454 (AARCH64_FL_FOR_ARCH8_4): New.
17455 (AARCH64_ISA_F16FML): New ISA flag.
17456 (TARGET_F16FML): New feature flag for fp16fml.
17457 (doc/invoke.texi): Document new fp16fml option.
17459 2018-01-10 Michael Collison <michael.collison@arm.com>
17461 * config/aarch64/aarch64-builtins.c:
17462 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17463 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17464 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17465 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17466 (AARCH64_ISA_SHA3): New ISA flag.
17467 (TARGET_SHA3): New feature flag for sha3.
17468 * config/aarch64/iterators.md (sha512_op): New int attribute.
17469 (CRYPTO_SHA512): New int iterator.
17470 (UNSPEC_SHA512H): New unspec.
17471 (UNSPEC_SHA512H2): Ditto.
17472 (UNSPEC_SHA512SU0): Ditto.
17473 (UNSPEC_SHA512SU1): Ditto.
17474 * config/aarch64/aarch64-simd-builtins.def
17475 (aarch64_crypto_sha512hqv2di): New builtin.
17476 (aarch64_crypto_sha512h2qv2di): Ditto.
17477 (aarch64_crypto_sha512su0qv2di): Ditto.
17478 (aarch64_crypto_sha512su1qv2di): Ditto.
17479 (aarch64_eor3qv8hi): Ditto.
17480 (aarch64_rax1qv2di): Ditto.
17481 (aarch64_xarqv2di): Ditto.
17482 (aarch64_bcaxqv8hi): Ditto.
17483 * config/aarch64/aarch64-simd.md:
17484 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17485 (aarch64_crypto_sha512su0qv2di): Ditto.
17486 (aarch64_crypto_sha512su1qv2di): Ditto.
17487 (aarch64_eor3qv8hi): Ditto.
17488 (aarch64_rax1qv2di): Ditto.
17489 (aarch64_xarqv2di): Ditto.
17490 (aarch64_bcaxqv8hi): Ditto.
17491 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17492 (vsha512h2q_u64): Ditto.
17493 (vsha512su0q_u64): Ditto.
17494 (vsha512su1q_u64): Ditto.
17495 (veor3q_u16): Ditto.
17496 (vrax1q_u64): Ditto.
17497 (vxarq_u64): Ditto.
17498 (vbcaxq_u16): Ditto.
17499 * config/arm/types.md (crypto_sha512): New type attribute.
17500 (crypto_sha3): Ditto.
17501 (doc/invoke.texi): Document new sha3 option.
17503 2018-01-10 Michael Collison <michael.collison@arm.com>
17505 * config/aarch64/aarch64-builtins.c:
17506 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17507 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17508 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17509 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17510 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17511 (AARCH64_ISA_SM4): New ISA flag.
17512 (TARGET_SM4): New feature flag for sm4.
17513 * config/aarch64/aarch64-simd-builtins.def
17514 (aarch64_sm3ss1qv4si): Ditto.
17515 (aarch64_sm3tt1aq4si): Ditto.
17516 (aarch64_sm3tt1bq4si): Ditto.
17517 (aarch64_sm3tt2aq4si): Ditto.
17518 (aarch64_sm3tt2bq4si): Ditto.
17519 (aarch64_sm3partw1qv4si): Ditto.
17520 (aarch64_sm3partw2qv4si): Ditto.
17521 (aarch64_sm4eqv4si): Ditto.
17522 (aarch64_sm4ekeyqv4si): Ditto.
17523 * config/aarch64/aarch64-simd.md:
17524 (aarch64_sm3ss1qv4si): Ditto.
17525 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17526 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17527 (aarch64_sm4eqv4si): Ditto.
17528 (aarch64_sm4ekeyqv4si): Ditto.
17529 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17530 (sm3part_op): Ditto.
17531 (CRYPTO_SM3TT): Ditto.
17532 (CRYPTO_SM3PART): Ditto.
17533 (UNSPEC_SM3SS1): New unspec.
17534 (UNSPEC_SM3TT1A): Ditto.
17535 (UNSPEC_SM3TT1B): Ditto.
17536 (UNSPEC_SM3TT2A): Ditto.
17537 (UNSPEC_SM3TT2B): Ditto.
17538 (UNSPEC_SM3PARTW1): Ditto.
17539 (UNSPEC_SM3PARTW2): Ditto.
17540 (UNSPEC_SM4E): Ditto.
17541 (UNSPEC_SM4EKEY): Ditto.
17542 * config/aarch64/constraints.md (Ui2): New constraint.
17543 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17544 * config/arm/types.md (crypto_sm3): New type attribute.
17545 (crypto_sm4): Ditto.
17546 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17547 (vsm3tt1aq_u32): Ditto.
17548 (vsm3tt1bq_u32): Ditto.
17549 (vsm3tt2aq_u32): Ditto.
17550 (vsm3tt2bq_u32): Ditto.
17551 (vsm3partw1q_u32): Ditto.
17552 (vsm3partw2q_u32): Ditto.
17553 (vsm4eq_u32): Ditto.
17554 (vsm4ekeyq_u32): Ditto.
17555 (doc/invoke.texi): Document new sm4 option.
17557 2018-01-10 Michael Collison <michael.collison@arm.com>
17559 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17560 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17561 (AARCH64_FL_FOR_ARCH8_4): New.
17562 (AARCH64_FL_V8_4): New flag.
17563 (doc/invoke.texi): Document new armv8.4-a option.
17565 2018-01-10 Michael Collison <michael.collison@arm.com>
17567 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17568 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17569 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17570 * config/aarch64/aarch64-option-extension.def: Add
17571 AARCH64_OPT_EXTENSION of 'sha2'.
17572 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17573 (crypto): Disable sha2 and aes if crypto disabled.
17574 (crypto): Enable aes and sha2 if enabled.
17575 (simd): Disable sha2 and aes if simd disabled.
17576 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17578 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17579 (TARGET_SHA2): New feature flag for sha2.
17580 (TARGET_AES): New feature flag for aes.
17581 * config/aarch64/aarch64-simd.md:
17582 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17583 conditional on TARGET_AES.
17584 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17585 (aarch64_crypto_sha1hsi): Make pattern conditional
17587 (aarch64_crypto_sha1hv4si): Ditto.
17588 (aarch64_be_crypto_sha1hv4si): Ditto.
17589 (aarch64_crypto_sha1su1v4si): Ditto.
17590 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17591 (aarch64_crypto_sha1su0v4si): Ditto.
17592 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17593 (aarch64_crypto_sha256su0v4si): Ditto.
17594 (aarch64_crypto_sha256su1v4si): Ditto.
17595 (doc/invoke.texi): Document new aes and sha2 options.
17597 2018-01-10 Martin Sebor <msebor@redhat.com>
17599 PR tree-optimization/83781
17600 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17603 2018-01-11 Martin Sebor <msebor@gmail.com>
17604 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17606 PR tree-optimization/83501
17607 PR tree-optimization/81703
17609 * tree-ssa-strlen.c (get_string_cst): Rename...
17610 (get_string_len): ...to this. Handle global constants.
17611 (handle_char_store): Adjust.
17613 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17614 Jim Wilson <jimw@sifive.com>
17616 * config/riscv/riscv-protos.h (riscv_output_return): New.
17617 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17618 (riscv_attribute_table, riscv_output_return),
17619 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17620 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17621 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17623 (riscv_expand_prologue): Add early return for naked function.
17624 (riscv_expand_epilogue): Likewise.
17625 (riscv_function_ok_for_sibcall): Return false for naked function.
17626 (riscv_set_current_function): New.
17627 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17628 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17629 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17630 * doc/extend.texi (RISC-V Function Attributes): New.
17632 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17634 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17635 check for 128-bit long double before checking TCmode.
17636 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17637 128-bit long doubles before checking TFmode or TCmode.
17638 (FLOAT128_IBM_P): Likewise.
17640 2018-01-10 Martin Sebor <msebor@redhat.com>
17642 PR tree-optimization/83671
17643 * builtins.c (c_strlen): Unconditionally return zero for the empty
17645 Use -Warray-bounds for warnings.
17646 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17647 for non-constant array indices with COMPONENT_REF, arrays of
17648 arrays, and pointers to arrays.
17649 (gimple_fold_builtin_strlen): Determine and set length range for
17650 non-constant character arrays.
17652 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17654 PR middle-end/81897
17655 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17658 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17660 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17662 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17665 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17666 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17667 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17668 indexed_or_indirect_operand predicate.
17669 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17670 (*vsx_le_perm_load_v8hi): Likewise.
17671 (*vsx_le_perm_load_v16qi): Likewise.
17672 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17673 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17674 (*vsx_le_perm_store_v8hi): Likewise.
17675 (*vsx_le_perm_store_v16qi): Likewise.
17676 (eight unnamed splitters): Likewise.
17678 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17680 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17681 * config/rs6000/emmintrin.h: Likewise.
17682 * config/rs6000/mmintrin.h: Likewise.
17683 * config/rs6000/xmmintrin.h: Likewise.
17685 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17688 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17690 * tree.c (tree_nop_conversion): Return true for location wrapper
17692 (maybe_wrap_with_location): New function.
17693 (selftest::check_strip_nops): New function.
17694 (selftest::test_location_wrappers): New function.
17695 (selftest::tree_c_tests): Call it.
17696 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17697 (maybe_wrap_with_location): New decl.
17698 (EXPR_LOCATION_WRAPPER_P): New macro.
17699 (location_wrapper_p): New inline function.
17700 (tree_strip_any_location_wrapper): New inline function.
17702 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17705 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17706 stack_realign_offset for the largest alignment of stack slot
17708 (ix86_find_max_used_stack_alignment): New function.
17709 (ix86_finalize_stack_frame_flags): Use it. Set
17710 max_used_stack_alignment if we don't realign stack.
17711 * config/i386/i386.h (machine_function): Add
17712 max_used_stack_alignment.
17714 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17716 * config/arm/arm.opt (-mbranch-cost): New option.
17717 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17720 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17723 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17724 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17726 2018-01-10 Richard Biener <rguenther@suse.de>
17729 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17730 early out so it also covers the case where we have a non-NULL
17733 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17735 PR tree-optimization/83753
17736 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17737 for non-strided grouped accesses if the number of elements is 1.
17739 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17742 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17743 * i386.h (TARGET_USE_GATHER): Define.
17744 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17746 2018-01-10 Martin Liska <mliska@suse.cz>
17749 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17750 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17752 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17753 CLEANUP_NO_PARTITIONING is not set.
17755 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17757 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17758 for vectors, as a partial revert of r254296.
17759 * rtl.h (const_vec_p): Delete.
17760 (const_vec_duplicate_p): Don't test for vector CONSTs.
17761 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17762 * expmed.c (make_tree): Likewise.
17765 * common.md (E, F): Use CONSTANT_P instead of checking for
17767 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17768 checking for CONST_VECTOR.
17770 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17772 PR middle-end/83575
17773 * predict.c (force_edge_cold): Handle in more sane way edges
17774 with no prediction.
17776 2018-01-09 Carl Love <cel@us.ibm.com>
17778 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17780 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17781 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17782 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17783 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17784 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17785 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17786 * config/rs6000/rs6000-protos.h: Add extern defition for
17787 rs6000_generate_float2_double_code.
17788 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17790 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17791 (float2_v2df): Add define_expand.
17793 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17796 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17797 op_mode in the force_to_mode call.
17799 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17801 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17802 instead of checking each element individually.
17803 (aarch64_evpc_uzp): Likewise.
17804 (aarch64_evpc_zip): Likewise.
17805 (aarch64_evpc_ext): Likewise.
17806 (aarch64_evpc_rev): Likewise.
17807 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17808 instead of checking each element individually. Return true without
17810 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17811 whether all selected elements come from the same input, instead of
17812 checking each element individually. Remove calls to gen_rtx_REG,
17813 start_sequence and end_sequence and instead assert that no rtl is
17816 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17818 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17819 order of HIGH and CONST checks.
17821 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17823 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17824 if the destination isn't an SSA_NAME.
17826 2018-01-09 Richard Biener <rguenther@suse.de>
17828 PR tree-optimization/83668
17829 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17831 (canonicalize_loop_form): ... here, renamed from ...
17832 (canonicalize_loop_closed_ssa_form): ... this and amended to
17833 swap successor edges for loop exit blocks to make us use
17834 the RPO order we need for initial schedule generation.
17836 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17838 PR tree-optimization/64811
17839 * match.pd: When optimizing comparisons with Inf, avoid
17840 introducing or losing exceptions from comparisons with NaN.
17842 2018-01-09 Martin Liska <mliska@suse.cz>
17845 * asan.c (shadow_mem_size): Add gcc_assert.
17847 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17849 Don't save registers in main().
17852 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17853 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17854 * config/avr/avr.c (avr_set_current_function): Don't error if
17855 naked, OS_task or OS_main are specified at the same time.
17856 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17858 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17860 * common/config/avr/avr-common.c (avr_option_optimization_table):
17861 Switch on -mmain-is-OS_task for optimizing compilations.
17863 2018-01-09 Richard Biener <rguenther@suse.de>
17865 PR tree-optimization/83572
17866 * graphite.c: Include cfganal.h.
17867 (graphite_transform_loops): Connect infinite loops to exit
17868 and remove fake edges at the end.
17870 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17872 * ipa-inline.c (edge_badness): Revert accidental checkin.
17874 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17877 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17878 symbols; not inline clones.
17880 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17883 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17884 hard registers. Formatting fixes.
17886 PR preprocessor/83722
17887 * gcc.c (try_generate_repro): Pass
17888 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17889 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17892 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17893 Kito Cheng <kito.cheng@gmail.com>
17895 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17896 (riscv_leaf_function_p): Delete.
17897 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17899 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17901 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17903 (do_ifelse): New function.
17904 (do_isel): New function.
17905 (do_sub3): New function.
17906 (do_add3): New function.
17907 (do_load_mask_compare): New function.
17908 (do_overlap_load_compare): New function.
17909 (expand_compare_loop): New function.
17910 (expand_block_compare): Call expand_compare_loop() when appropriate.
17911 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17912 option description.
17913 (-mblock-compare-inline-loop-limit): New option.
17915 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17918 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17919 Reverse order of second and third operands in first alternative.
17920 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17921 of first and second elements in UNSPEC_VPERMR vector.
17922 (altivec_expand_vec_perm_le): Likewise.
17924 2018-01-08 Jeff Law <law@redhat.com>
17926 PR rtl-optimizatin/81308
17927 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17928 (process_switch): If group_case_labels makes a change, then set
17930 (pass_convert_switch::execute): If a switch is converted, then
17931 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17933 PR rtl-optimization/81308
17934 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17937 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17939 PR target/83663 - Revert r255946
17940 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17941 generation for cases where splatting a value is not useful.
17942 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17943 across a vec_duplicate and a paradoxical subreg forming a vector
17944 mode to a vec_concat.
17946 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17948 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17949 -march=armv8.3-a variants.
17950 * config/arm/t-multilib: Likewise.
17951 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17953 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17955 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17957 (cceq_ior_compare_complement): Give it a name so I can use it, and
17958 change boolean_or_operator predicate to boolean_operator so it can
17959 be used to generate a crand.
17960 (eqne): New code iterator.
17961 (bd/bd_neg): New code_attrs.
17962 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17963 a single define_insn.
17964 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17965 decrement (bdnzt/bdnzf/bdzt/bdzf).
17966 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17967 with the new names of the branch decrement patterns, and added the
17968 names of the branch decrement conditional patterns.
17970 2018-01-08 Richard Biener <rguenther@suse.de>
17972 PR tree-optimization/83563
17973 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17976 2018-01-08 Richard Biener <rguenther@suse.de>
17978 PR middle-end/83713
17979 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17981 2018-01-08 Richard Biener <rguenther@suse.de>
17983 PR tree-optimization/83685
17984 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17985 references to abnormals.
17987 2018-01-08 Richard Biener <rguenther@suse.de>
17990 * dwarf2out.c (output_indirect_strings): Handle empty
17991 skeleton_debug_str_hash.
17992 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17994 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17996 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17997 (emit_store_direct): Likewise.
17998 (arc_trampoline_adjust_address): Likewise.
17999 (arc_asm_trampoline_template): New function.
18000 (arc_initialize_trampoline): Use asm_trampoline_template.
18001 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18002 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18003 * config/arc/arc.md (flush_icache): Delete pattern.
18005 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18007 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18008 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18011 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18014 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18015 by not USED_FOR_TARGET.
18016 (make_pass_resolve_sw_modes): Likewise.
18018 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18020 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18023 2018-01-08 Richard Biener <rguenther@suse.de>
18025 PR middle-end/83580
18026 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18028 2018-01-08 Richard Biener <rguenther@suse.de>
18030 PR middle-end/83517
18031 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18033 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18035 PR middle-end/81897
18036 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18037 basic blocks with a small number of successors.
18038 (convert_control_dep_chain_into_preds): Improve handling of
18040 (dump_predicates): Split apart into...
18041 (dump_pred_chain): ...here...
18042 (dump_pred_info): ...and here.
18043 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18044 (can_chain_union_be_invalidated_p): Improve check for invalidation
18046 (uninit_uses_cannot_happen): Avoid unnecessary if
18047 convert_control_dep_chain_into_preds yielded nothing.
18049 2018-01-06 Martin Sebor <msebor@redhat.com>
18051 PR tree-optimization/83640
18052 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18053 subtracting negative offset from size.
18054 (builtin_access::overlap): Adjust offset bounds of the access to fall
18055 within the size of the object if possible.
18057 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18059 PR rtl-optimization/83699
18060 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18061 extract_bit_field_as_subreg to cases in which the extracted
18062 value is also a vector.
18064 * lra-constraints.c (process_alt_operands): Test for the equivalence
18065 substitutions when detecting a possible reload cycle.
18067 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18070 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18071 by default if flag_selective_schedling{,2}. Formatting fixes.
18073 PR rtl-optimization/83682
18074 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18075 if it has non-VECTOR_MODE element mode.
18076 (vec_duplicate_p): Likewise.
18078 PR middle-end/83694
18079 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18080 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18082 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18085 * config/i386/i386-builtin.def
18086 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18087 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18088 Require also OPTION_MASK_ISA_AVX512F in addition to
18089 OPTION_MASK_ISA_GFNI.
18090 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18091 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18092 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18093 to OPTION_MASK_ISA_GFNI.
18094 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18095 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18096 OPTION_MASK_ISA_AVX512BW.
18097 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18098 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18099 addition to OPTION_MASK_ISA_GFNI.
18100 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18101 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18102 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18103 to OPTION_MASK_ISA_GFNI.
18104 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18105 a requirement for all ISAs rather than any of them with a few
18107 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18109 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18110 bitmasks to be enabled with 3 exceptions, instead of requiring any
18111 enabled ISA with lots of exceptions.
18112 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18113 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18114 Change avx512bw in isa attribute to avx512f.
18115 * config/i386/sgxintrin.h: Add license boilerplate.
18116 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18117 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18118 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18119 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18121 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18122 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18123 temporarily sse2 rather than sse if not enabled already.
18126 * config/i386/sse.md (VI248_VLBW): Rename to ...
18127 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18128 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18129 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18130 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18131 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18132 mode iterator instead of VI248_VLBW.
18134 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18136 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18137 (record_modified): Skip clobbers; add debug output.
18138 (param_change_prob): Use sreal frequencies.
18140 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18142 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18143 punt for user-aligned variables.
18145 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18147 * tree-chrec.c (chrec_contains_symbols): Return true for
18150 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18153 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18154 of (x|y) == x for BICS pattern.
18156 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18158 PR tree-optimization/83605
18159 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18160 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18163 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18165 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18166 * config/epiphany/rtems.h: New file.
18168 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18169 Uros Bizjak <ubizjak@gmail.com>
18172 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18173 QIreg_operand instead of register_operand predicate.
18174 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18175 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18176 comments instead of -fmitigate[-_]rop.
18178 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18181 * cgraphunit.c (symbol_table::compile): Switch to text_section
18182 before calling assembly_start debug hook.
18183 * run-rtl-passes.c (run_rtl_passes): Likewise.
18186 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18188 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18189 range_int_cst_p rather than !symbolic_range_p before calling
18190 extract_range_from_multiplicative_op_1.
18192 2018-01-04 Jeff Law <law@redhat.com>
18194 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18195 redundant test in assertion.
18197 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18199 * doc/rtl.texi: Document machine_mode wrapper classes.
18201 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18203 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18204 using tree_to_uhwi.
18206 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18208 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18209 the VEC_PERM_EXPR fold to fail.
18211 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18214 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18215 to switched_sections.
18217 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18220 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18221 test for d.testing.
18223 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18226 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18227 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18229 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18232 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18233 is BLKmode and bitpos not zero or mode change is needed.
18235 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18238 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18241 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18244 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18245 instead of MULT rtx. Update all corresponding splitters.
18246 (*saddl_se): Ditto.
18247 (*ssub<modesuffix>): Ditto.
18248 (*ssubl_se): Ditto.
18249 (*cmp_sadd_di): Update split patterns.
18250 (*cmp_sadd_si): Ditto.
18251 (*cmp_sadd_sidi): Ditto.
18252 (*cmp_ssub_di): Ditto.
18253 (*cmp_ssub_si): Ditto.
18254 (*cmp_ssub_sidi): Ditto.
18255 * config/alpha/predicates.md (const23_operand): New predicate.
18256 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18257 Look for ASHIFT, not MULT inner operand.
18258 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18260 2018-01-04 Martin Liska <mliska@suse.cz>
18262 PR gcov-profile/83669
18263 * gcov.c (output_intermediate_file): Add version to intermediate
18265 * doc/gcov.texi: Document new field 'version' in intermediate
18266 file format. Fix location of '-k' option of gcov command.
18268 2018-01-04 Martin Liska <mliska@suse.cz>
18271 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18273 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18275 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18277 2018-01-03 Martin Sebor <msebor@redhat.com>
18279 PR tree-optimization/83655
18280 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18281 checking calls with invalid arguments.
18283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18285 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18286 (vectorizable_mask_load_store): Delete.
18287 (vectorizable_call): Return false for masked loads and stores.
18288 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18289 instead of gimple_assign_rhs1.
18290 (vectorizable_load): Handle IFN_MASK_LOAD.
18291 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18295 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18297 (vectorizable_mask_load_store): ...here.
18298 (vectorizable_load): ...and here.
18300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18302 * tree-vect-stmts.c (vect_build_all_ones_mask)
18303 (vect_build_zero_merge_argument): New functions, split out from...
18304 (vectorizable_load): ...here.
18306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18308 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18310 (vectorizable_mask_load_store): ...here.
18311 (vectorizable_store): ...and here.
18313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18315 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18317 (vectorizable_mask_load_store): ...here.
18319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18321 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18322 (vect_model_store_cost): Take a vec_load_store_type instead of a
18324 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18325 (vect_model_store_cost): Take a vec_load_store_type instead of a
18327 (vectorizable_mask_load_store): Update accordingly.
18328 (vectorizable_store): Likewise.
18329 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18333 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18334 IFN_MASK_LOAD calls here rather than...
18335 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18338 Alan Hayward <alan.hayward@arm.com>
18339 David Sherwood <david.sherwood@arm.com>
18341 * expmed.c (extract_bit_field_1): For vector extracts,
18342 fall back to extract_bit_field_as_subreg if vec_extract
18345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18346 Alan Hayward <alan.hayward@arm.com>
18347 David Sherwood <david.sherwood@arm.com>
18349 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18350 they are variable or constant sized.
18351 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18352 slots for constant-sized data.
18354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18355 Alan Hayward <alan.hayward@arm.com>
18356 David Sherwood <david.sherwood@arm.com>
18358 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18359 handling COND_EXPRs with boolean comparisons, try to find a better
18360 basis for the mask type than the boolean itself.
18362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18364 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18365 is calculated and how it can be overridden.
18366 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18367 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18369 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18373 Alan Hayward <alan.hayward@arm.com>
18374 David Sherwood <david.sherwood@arm.com>
18376 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18377 Remove the mode argument.
18378 (aarch64_simd_valid_immediate): Remove the mode and inverse
18380 * config/aarch64/iterators.md (bitsize): New iterator.
18381 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18382 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18383 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18384 aarch64_simd_valid_immediate.
18385 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18386 (aarch64_reg_or_bic_imm): Likewise.
18387 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18388 with an insn_type enum and msl with a modifier_type enum.
18389 Replace element_width with a scalar_mode. Change the shift
18390 to unsigned int. Add constructors for scalar_float_mode and
18391 scalar_int_mode elements.
18392 (aarch64_vect_float_const_representable_p): Delete.
18393 (aarch64_can_const_movi_rtx_p)
18394 (aarch64_simd_scalar_immediate_valid_for_move)
18395 (aarch64_simd_make_constant): Update call to
18396 aarch64_simd_valid_immediate.
18397 (aarch64_advsimd_valid_immediate_hs): New function.
18398 (aarch64_advsimd_valid_immediate): Likewise.
18399 (aarch64_simd_valid_immediate): Remove mode and inverse
18400 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18401 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18402 and aarch64_float_const_representable_p on the result.
18403 (aarch64_output_simd_mov_immediate): Remove mode argument.
18404 Update call to aarch64_simd_valid_immediate and use of
18405 simd_immediate_info.
18406 (aarch64_output_scalar_simd_mov_immediate): Update call
18409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18410 Alan Hayward <alan.hayward@arm.com>
18411 David Sherwood <david.sherwood@arm.com>
18413 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18414 (mode_nunits): Likewise CONST_MODE_NUNITS.
18415 * machmode.def (ADJUST_NUNITS): Document.
18416 * genmodes.c (mode_data::need_nunits_adj): New field.
18417 (blank_mode): Update accordingly.
18418 (adj_nunits): New variable.
18419 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18421 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18422 listed in adj_nunits.
18423 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18424 listed in adj_nunits. Don't emit case statements for such modes.
18425 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18426 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18427 nothing if adj_nunits is nonnull.
18428 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18429 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18430 (emit_mode_fbit): Update use of print_maybe_const_decl.
18431 (emit_move_size): Likewise. Treat the array as non-const
18433 (emit_mode_adjustments): Handle adj_nunits.
18435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18437 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18438 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18439 (VECTOR_MODES): Use it.
18440 (make_vector_modes): Take the prefix as an argument.
18442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18443 Alan Hayward <alan.hayward@arm.com>
18444 David Sherwood <david.sherwood@arm.com>
18446 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18447 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18448 for MODE_VECTOR_BOOL.
18449 * machmode.def (VECTOR_BOOL_MODE): Document.
18450 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18451 (make_vector_bool_mode): New function.
18452 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18454 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18455 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18457 * stor-layout.c (int_mode_for_mode): Likewise.
18458 * tree.c (build_vector_type_for_mode): Likewise.
18459 * varasm.c (output_constant_pool_2): Likewise.
18460 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18461 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18462 for MODE_VECTOR_BOOL.
18463 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18464 of mode class checks.
18465 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18466 instead of a list of mode class checks.
18467 (expand_vector_scalar_condition): Likewise.
18468 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18471 Alan Hayward <alan.hayward@arm.com>
18472 David Sherwood <david.sherwood@arm.com>
18474 * machmode.h (mode_size): Change from unsigned short to
18476 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18477 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18478 or if measurement_type is not polynomial.
18479 (fixed_size_mode::includes_p): Check for constant-sized modes.
18480 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18481 return a poly_uint16 rather than an unsigned short.
18482 (emit_mode_size): Change the type of mode_size from unsigned short
18483 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18484 (emit_mode_adjustments): Cope with polynomial vector sizes.
18485 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18487 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18489 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18490 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18491 * caller-save.c (setup_save_areas): Likewise.
18492 (replace_reg_with_saved_mem): Likewise.
18493 * calls.c (emit_library_call_value_1): Likewise.
18494 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18495 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18496 (gen_lowpart_for_combine): Likewise.
18497 * convert.c (convert_to_integer_1): Likewise.
18498 * cse.c (equiv_constant, cse_insn): Likewise.
18499 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18500 (cselib_subst_to_values): Likewise.
18501 * dce.c (word_dce_process_block): Likewise.
18502 * df-problems.c (df_word_lr_mark_ref): Likewise.
18503 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18504 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18505 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18506 (rtl_for_decl_location): Likewise.
18507 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18508 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18509 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18510 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18511 (expand_expr_real_1): Likewise.
18512 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18513 (pad_below): Likewise.
18514 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18515 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18516 * ira.c (get_subreg_tracking_sizes): Likewise.
18517 * ira-build.c (ira_create_allocno_objects): Likewise.
18518 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18519 (ira_sort_regnos_for_alter_reg): Likewise.
18520 * ira-costs.c (record_operand_costs): Likewise.
18521 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18522 (resolve_simple_move): Likewise.
18523 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18524 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18525 (lra_constraints): Likewise.
18526 (CONST_POOL_OK_P): Reject variable-sized modes.
18527 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18528 (add_pseudo_to_slot, lra_spill): Likewise.
18529 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18530 * optabs-query.c (get_best_extraction_insn): Likewise.
18531 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18532 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18533 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18534 * recog.c (offsettable_address_addr_space_p): Likewise.
18535 * regcprop.c (maybe_mode_change): Likewise.
18536 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18537 * regrename.c (build_def_use): Likewise.
18538 * regstat.c (dump_reg_info): Likewise.
18539 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18540 (find_reloads, find_reloads_subreg_address): Likewise.
18541 * reload1.c (eliminate_regs_1): Likewise.
18542 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18543 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18544 (simplify_binary_operation_1, simplify_subreg): Likewise.
18545 * targhooks.c (default_function_arg_padding): Likewise.
18546 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18547 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18548 (verify_gimple_assign_ternary): Likewise.
18549 * tree-inline.c (estimate_move_cost): Likewise.
18550 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18551 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18552 (get_address_cost_ainc): Likewise.
18553 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18554 (vect_supportable_dr_alignment): Likewise.
18555 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18556 (vectorizable_reduction): Likewise.
18557 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18558 (vectorizable_operation, vectorizable_load): Likewise.
18559 * tree.c (build_same_sized_truth_vector_type): Likewise.
18560 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18561 * var-tracking.c (emit_note_insn_var_location): Likewise.
18562 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18563 (ADDR_VEC_ALIGN): Likewise.
18565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18566 Alan Hayward <alan.hayward@arm.com>
18567 David Sherwood <david.sherwood@arm.com>
18569 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18571 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18572 or if measurement_type is polynomial.
18573 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18574 * combine.c (make_extraction): Likewise.
18575 * dse.c (find_shift_sequence): Likewise.
18576 * dwarf2out.c (mem_loc_descriptor): Likewise.
18577 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18578 (extract_bit_field, extract_low_bits): Likewise.
18579 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18580 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18581 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18582 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18583 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18584 * reload.c (find_reloads): Likewise.
18585 * reload1.c (alter_reg): Likewise.
18586 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18587 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18588 * tree-if-conv.c (predicate_mem_writes): Likewise.
18589 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18590 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18591 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18592 * valtrack.c (dead_debug_insert_temp): Likewise.
18593 * varasm.c (mergeable_constant_section): Likewise.
18594 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18597 Alan Hayward <alan.hayward@arm.com>
18598 David Sherwood <david.sherwood@arm.com>
18600 * expr.c (expand_assignment): Cope with polynomial mode sizes
18601 when assigning to a CONCAT.
18603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18604 Alan Hayward <alan.hayward@arm.com>
18605 David Sherwood <david.sherwood@arm.com>
18607 * machmode.h (mode_precision): Change from unsigned short to
18609 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18611 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18612 or if measurement_type is not polynomial.
18613 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18614 in which the mode is already known to be a scalar_int_mode.
18615 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18616 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18618 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18619 for GET_MODE_PRECISION.
18620 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18621 for GET_MODE_PRECISION.
18622 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18624 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18625 (expand_field_assignment, make_extraction): Likewise.
18626 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18627 (get_last_value): Likewise.
18628 * convert.c (convert_to_integer_1): Likewise.
18629 * cse.c (cse_insn): Likewise.
18630 * expr.c (expand_expr_real_1): Likewise.
18631 * lra-constraints.c (simplify_operand_subreg): Likewise.
18632 * optabs-query.c (can_atomic_load_p): Likewise.
18633 * optabs.c (expand_atomic_load): Likewise.
18634 (expand_atomic_store): Likewise.
18635 * ree.c (combine_reaching_defs): Likewise.
18636 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18637 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18638 * tree.h (type_has_mode_precision_p): Likewise.
18639 * ubsan.c (instrument_si_overflow): Likewise.
18641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18642 Alan Hayward <alan.hayward@arm.com>
18643 David Sherwood <david.sherwood@arm.com>
18645 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18646 polynomial numbers of units.
18647 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18648 (valid_vector_subparts_p): New function.
18649 (build_vector_type): Remove temporary shim and take the number
18650 of units as a poly_uint64 rather than an int.
18651 (build_opaque_vector_type): Take the number of units as a
18652 poly_uint64 rather than an int.
18653 * tree.c (build_vector_from_ctor): Handle polynomial
18654 TYPE_VECTOR_SUBPARTS.
18655 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18656 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18657 (build_vector_from_val): If the number of units is variable,
18658 use build_vec_duplicate_cst for constant operands and
18659 VEC_DUPLICATE_EXPR otherwise.
18660 (make_vector_type): Remove temporary is_constant ().
18661 (build_vector_type, build_opaque_vector_type): Take the number of
18662 units as a poly_uint64 rather than an int.
18663 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18665 * cfgexpand.c (expand_debug_expr): Likewise.
18666 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18667 (store_constructor, expand_expr_real_1): Likewise.
18668 (const_scalar_mask_from_tree): Likewise.
18669 * fold-const-call.c (fold_const_reduction): Likewise.
18670 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18671 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18672 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18673 (fold_relational_const): Likewise.
18674 (native_interpret_vector): Likewise. Change the size from an
18675 int to an unsigned int.
18676 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18677 TYPE_VECTOR_SUBPARTS.
18678 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18679 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18680 duplicating a non-constant operand into a variable-length vector.
18681 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18682 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18683 * ipa-icf.c (sem_variable::equals): Likewise.
18684 * match.pd: Likewise.
18685 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18686 * print-tree.c (print_node): Likewise.
18687 * stor-layout.c (layout_type): Likewise.
18688 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18689 * tree-cfg.c (verify_gimple_comparison): Likewise.
18690 (verify_gimple_assign_binary): Likewise.
18691 (verify_gimple_assign_ternary): Likewise.
18692 (verify_gimple_assign_single): Likewise.
18693 * tree-pretty-print.c (dump_generic_node): Likewise.
18694 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18695 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18696 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18697 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18698 (vect_shift_permute_load_chain): Likewise.
18699 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18700 (expand_vector_condition, optimize_vector_constructor): Likewise.
18701 (lower_vec_perm, get_compute_type): Likewise.
18702 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18703 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18704 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18705 (vect_recog_mask_conversion_pattern): Likewise.
18706 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18707 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18708 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18709 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18710 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18711 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18712 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18713 (supportable_widening_operation): Likewise.
18714 (supportable_narrowing_operation): Likewise.
18715 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18717 * varasm.c (output_constant): Likewise.
18719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18720 Alan Hayward <alan.hayward@arm.com>
18721 David Sherwood <david.sherwood@arm.com>
18723 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18724 so that both the length == 3 and length != 3 cases set up their
18725 own permute vectors. Add comments explaining why we know the
18726 number of elements is constant.
18727 (vect_permute_load_chain): Likewise.
18729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18730 Alan Hayward <alan.hayward@arm.com>
18731 David Sherwood <david.sherwood@arm.com>
18733 * machmode.h (mode_nunits): Change from unsigned char to
18735 (ONLY_FIXED_SIZE_MODES): New macro.
18736 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18737 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18738 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18740 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18741 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18742 or if measurement_type is not polynomial.
18743 * genmodes.c (ZERO_COEFFS): New macro.
18744 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18746 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18747 Use ZERO_COEFFS when emitting initializers.
18748 * data-streamer.h (bp_pack_poly_value): New function.
18749 (bp_unpack_poly_value): Likewise.
18750 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18751 for GET_MODE_NUNITS.
18752 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18753 for GET_MODE_NUNITS.
18754 * tree.c (make_vector_type): Remove temporary shim and make
18755 the real function take the number of units as a poly_uint64
18756 rather than an int.
18757 (build_vector_type_for_mode): Handle polynomial nunits.
18758 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18759 * emit-rtl.c (const_vec_series_p_1): Likewise.
18760 (gen_rtx_CONST_VECTOR): Likewise.
18761 * fold-const.c (test_vec_duplicate_folding): Likewise.
18762 * genrecog.c (validate_pattern): Likewise.
18763 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18764 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18765 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18766 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18767 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18768 * rtlanal.c (subreg_get_info): Likewise.
18769 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18770 (vect_grouped_load_supported): Likewise.
18771 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18772 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18773 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18774 (simplify_const_unary_operation, simplify_binary_operation_1)
18775 (simplify_const_binary_operation, simplify_ternary_operation)
18776 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18777 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18778 instead of CONST_VECTOR_NUNITS.
18779 * varasm.c (output_constant_pool_2): Likewise.
18780 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18781 explicit-encoded elements in the XVEC for variable-length vectors.
18783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18785 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18788 Alan Hayward <alan.hayward@arm.com>
18789 David Sherwood <david.sherwood@arm.com>
18791 * coretypes.h (fixed_size_mode): Declare.
18792 (fixed_size_mode_pod): New typedef.
18793 * builtins.h (target_builtins::x_apply_args_mode)
18794 (target_builtins::x_apply_result_mode): Change type to
18795 fixed_size_mode_pod.
18796 * builtins.c (apply_args_size, apply_result_size, result_vector)
18797 (expand_builtin_apply_args_1, expand_builtin_apply)
18798 (expand_builtin_return): Update accordingly.
18800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18802 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18803 * cselib.c (cselib_hash_rtx): Likewise.
18804 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18805 CONST_VECTOR encoding.
18807 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18808 Jeff Law <law@redhat.com>
18811 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18812 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18813 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18814 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18817 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18818 explicitly probe *sp in a noreturn function if there were any callee
18819 register saves or frame pointer is needed.
18821 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18824 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18825 BLKmode for ternary, binary or unary expressions.
18828 * var-tracking.c (delete_vta_debug_insn): New inline function.
18829 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18830 insns from get_insns () to NULL instead of each bb separately.
18831 Use delete_vta_debug_insn. No longer static.
18832 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18833 delete_vta_debug_insns callers.
18834 * rtl.h (delete_vta_debug_insns): Declare.
18835 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18836 instead of variable_tracking_main.
18838 2018-01-03 Martin Sebor <msebor@redhat.com>
18840 PR tree-optimization/83603
18841 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18842 arguments past the endof the argument list in functions declared
18843 without a prototype.
18844 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18845 Avoid checking when arguments are null.
18847 2018-01-03 Martin Sebor <msebor@redhat.com>
18850 * doc/extend.texi (attribute const): Fix a typo.
18851 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18852 issuing -Wsuggest-attribute for void functions.
18854 2018-01-03 Martin Sebor <msebor@redhat.com>
18856 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18857 offset_int::from instead of wide_int::to_shwi.
18858 (maybe_diag_overlap): Remove assertion.
18859 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18860 * gimple-ssa-sprintf.c (format_directive): Same.
18861 (parse_directive): Same.
18862 (sprintf_dom_walker::compute_format_length): Same.
18863 (try_substitute_return_value): Same.
18865 2018-01-03 Jeff Law <law@redhat.com>
18867 PR middle-end/83654
18868 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18869 non-constant residual for zero at runtime and avoid probing in
18870 that case. Reorganize code for trailing problem to mirror handling
18873 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18875 PR tree-optimization/83501
18876 * tree-ssa-strlen.c (get_string_cst): New.
18877 (handle_char_store): Call get_string_cst.
18879 2018-01-03 Martin Liska <mliska@suse.cz>
18881 PR tree-optimization/83593
18882 * tree-ssa-strlen.c: Include tree-cfg.h.
18883 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18884 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18885 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18887 (strlen_dom_walker::before_dom_children): Call
18888 gimple_purge_dead_eh_edges. Dump tranformation with details
18890 (strlen_dom_walker::before_dom_children): Update call by adding
18891 new argument cleanup_eh.
18892 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18894 2018-01-03 Martin Liska <mliska@suse.cz>
18897 * cif-code.def (VARIADIC_THUNK): New enum value.
18898 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18901 2018-01-03 Jan Beulich <jbeulich@suse.com>
18903 * sse.md (mov<mode>_internal): Tighten condition for when to use
18904 vmovdqu<ssescalarsize> for TI and OI modes.
18906 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18908 Update copyright years.
18910 2018-01-03 Martin Liska <mliska@suse.cz>
18913 * ipa-visibility.c (function_and_variable_visibility): Skip
18914 functions with noipa attribure.
18916 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18918 * gcc.c (process_command): Update copyright notice dates.
18919 * gcov-dump.c (print_version): Ditto.
18920 * gcov.c (print_version): Ditto.
18921 * gcov-tool.c (print_version): Ditto.
18922 * gengtype.c (create_file): Ditto.
18923 * doc/cpp.texi: Bump @copying's copyright year.
18924 * doc/cppinternals.texi: Ditto.
18925 * doc/gcc.texi: Ditto.
18926 * doc/gccint.texi: Ditto.
18927 * doc/gcov.texi: Ditto.
18928 * doc/install.texi: Ditto.
18929 * doc/invoke.texi: Ditto.
18931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18933 * vector-builder.h (vector_builder::m_full_nelts): Change from
18934 unsigned int to poly_uint64.
18935 (vector_builder::full_nelts): Update prototype accordingly.
18936 (vector_builder::new_vector): Likewise.
18937 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18938 (vector_builder::operator ==): Likewise.
18939 (vector_builder::finalize): Likewise.
18940 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18941 Take the number of elements as a poly_uint64 rather than an
18943 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18944 from unsigned int to poly_uint64.
18945 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18946 (vec_perm_indices::new_vector): Likewise.
18947 (vec_perm_indices::length): Likewise.
18948 (vec_perm_indices::nelts_per_input): Likewise.
18949 (vec_perm_indices::input_nelts): Likewise.
18950 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18951 number of elements per input as a poly_uint64 rather than an
18952 unsigned int. Use the original encoding for variable-length
18953 vectors, rather than clamping each individual element.
18954 For the second and subsequent elements in each pattern,
18955 clamp the step and base before clamping their sum.
18956 (vec_perm_indices::series_p): Handle polynomial element counts.
18957 (vec_perm_indices::all_in_range_p): Likewise.
18958 (vec_perm_indices_to_tree): Likewise.
18959 (vec_perm_indices_to_rtx): Likewise.
18960 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18961 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18962 (tree_vector_builder::new_binary_operation): Handle polynomial
18963 element counts. Return false if we need to know the number
18964 of elements at compile time.
18965 * fold-const.c (fold_vec_perm): Punt if the number of elements
18966 isn't known at compile time.
18968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18970 * vec-perm-indices.h (vec_perm_builder): Change element type
18971 from HOST_WIDE_INT to poly_int64.
18972 (vec_perm_indices::element_type): Update accordingly.
18973 (vec_perm_indices::clamp): Handle polynomial element_types.
18974 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18975 (vec_perm_indices::all_in_range_p): Likewise.
18976 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18978 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18979 polynomial vec_perm_indices element types.
18980 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18981 * fold-const.c (fold_vec_perm): Likewise.
18982 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18983 * tree-vect-generic.c (lower_vec_perm): Likewise.
18984 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18985 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18986 element type to HOST_WIDE_INT.
18988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18989 Alan Hayward <alan.hayward@arm.com>
18990 David Sherwood <david.sherwood@arm.com>
18992 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18993 rather than an int. Use plus_constant.
18994 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18995 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18998 Alan Hayward <alan.hayward@arm.com>
18999 David Sherwood <david.sherwood@arm.com>
19001 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19002 a HOST_WIDE_INT to a poly_int64.
19004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19005 Alan Hayward <alan.hayward@arm.com>
19006 David Sherwood <david.sherwood@arm.com>
19008 * calls.c (load_register_parameters): Cope with polynomial
19009 mode sizes. Require a constant size for BLKmode parameters
19010 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19011 forces a parameter to be padded at the lsb end in order to
19012 fill a complete number of words, require the parameter size
19013 to be ordered wrt UNITS_PER_WORD.
19015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19016 Alan Hayward <alan.hayward@arm.com>
19017 David Sherwood <david.sherwood@arm.com>
19019 * reload1.c (spill_stack_slot_width): Change element type
19020 from unsigned int to poly_uint64_pod.
19021 (alter_reg): Treat mode sizes as polynomial.
19023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19024 Alan Hayward <alan.hayward@arm.com>
19025 David Sherwood <david.sherwood@arm.com>
19027 * reload.c (complex_word_subreg_p): New function.
19028 (reload_inner_reg_of_subreg, push_reload): Use it.
19030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19031 Alan Hayward <alan.hayward@arm.com>
19032 David Sherwood <david.sherwood@arm.com>
19034 * lra-constraints.c (process_alt_operands): Reject matched
19035 operands whose sizes aren't ordered.
19036 (match_reload): Refer to this check here.
19038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19039 Alan Hayward <alan.hayward@arm.com>
19040 David Sherwood <david.sherwood@arm.com>
19042 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19043 that the mode size is in the set {1, 2, 4, 8, 16}.
19045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19046 Alan Hayward <alan.hayward@arm.com>
19047 David Sherwood <david.sherwood@arm.com>
19049 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19050 Use plus_constant instead of gen_rtx_PLUS.
19052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19053 Alan Hayward <alan.hayward@arm.com>
19054 David Sherwood <david.sherwood@arm.com>
19056 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19057 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19058 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19059 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19060 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19061 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19062 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19063 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19064 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19065 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19067 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19068 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19069 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19070 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19071 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19072 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19073 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19074 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19075 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19076 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19078 * expr.c (emit_move_resolve_push): Treat the input and result
19079 of PUSH_ROUNDING as a poly_int64.
19080 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19081 (emit_push_insn): Likewise.
19082 * lra-eliminations.c (mark_not_eliminable): Likewise.
19083 * recog.c (push_operand): Likewise.
19084 * reload1.c (elimination_effects): Likewise.
19085 * rtlanal.c (nonzero_bits1): Likewise.
19086 * calls.c (store_one_arg): Likewise. Require the padding to be
19087 known at compile time.
19089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19090 Alan Hayward <alan.hayward@arm.com>
19091 David Sherwood <david.sherwood@arm.com>
19093 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19094 Use plus_constant instead of gen_rtx_PLUS.
19096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19097 Alan Hayward <alan.hayward@arm.com>
19098 David Sherwood <david.sherwood@arm.com>
19100 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19101 rather than an int.
19103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19104 Alan Hayward <alan.hayward@arm.com>
19105 David Sherwood <david.sherwood@arm.com>
19107 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19108 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19109 via stack temporaries. Treat the mode size as polynomial too.
19111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19112 Alan Hayward <alan.hayward@arm.com>
19113 David Sherwood <david.sherwood@arm.com>
19115 * expr.c (expand_expr_real_2): When handling conversions involving
19116 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19117 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19118 as a poly_uint64 too.
19120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19121 Alan Hayward <alan.hayward@arm.com>
19122 David Sherwood <david.sherwood@arm.com>
19124 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19127 Alan Hayward <alan.hayward@arm.com>
19128 David Sherwood <david.sherwood@arm.com>
19130 * combine.c (can_change_dest_mode): Handle polynomial
19131 REGMODE_NATURAL_SIZE.
19132 * expmed.c (store_bit_field_1): Likewise.
19133 * expr.c (store_constructor): Likewise.
19134 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19135 and polynomial REGMODE_NATURAL_SIZE.
19136 (gen_lowpart_common): Likewise.
19137 * reginfo.c (record_subregs_of_mode): Likewise.
19138 * rtlanal.c (read_modify_subreg_p): Likewise.
19140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19141 Alan Hayward <alan.hayward@arm.com>
19142 David Sherwood <david.sherwood@arm.com>
19144 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19145 numbers of elements.
19147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19148 Alan Hayward <alan.hayward@arm.com>
19149 David Sherwood <david.sherwood@arm.com>
19151 * match.pd: Cope with polynomial numbers of vector elements.
19153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19154 Alan Hayward <alan.hayward@arm.com>
19155 David Sherwood <david.sherwood@arm.com>
19157 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19158 in a POINTER_PLUS_EXPR.
19160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19161 Alan Hayward <alan.hayward@arm.com>
19162 David Sherwood <david.sherwood@arm.com>
19164 * omp-simd-clone.c (simd_clone_subparts): New function.
19165 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19166 (ipa_simd_modify_function_body): Likewise.
19168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19169 Alan Hayward <alan.hayward@arm.com>
19170 David Sherwood <david.sherwood@arm.com>
19172 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19173 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19174 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19175 (expand_vector_condition, vector_element): Likewise.
19176 (subparts_gt): New function.
19177 (get_compute_type): Use subparts_gt.
19178 (count_type_subparts): Delete.
19179 (expand_vector_operations_1): Use subparts_gt instead of
19180 count_type_subparts.
19182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19183 Alan Hayward <alan.hayward@arm.com>
19184 David Sherwood <david.sherwood@arm.com>
19186 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19187 (vect_compile_time_alias): ...this new function. Do the calculation
19188 on poly_ints rather than trees.
19189 (vect_prune_runtime_alias_test_list): Update call accordingly.
19191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19192 Alan Hayward <alan.hayward@arm.com>
19193 David Sherwood <david.sherwood@arm.com>
19195 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19197 (vect_schedule_slp_instance): Likewise.
19199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19200 Alan Hayward <alan.hayward@arm.com>
19201 David Sherwood <david.sherwood@arm.com>
19203 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19204 constant and extern definitions for variable-length vectors.
19205 (vect_get_constant_vectors): Note that the number of units
19206 is known to be constant.
19208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19209 Alan Hayward <alan.hayward@arm.com>
19210 David Sherwood <david.sherwood@arm.com>
19212 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19213 of units as polynomial. Choose between WIDE and NARROW based
19216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19217 Alan Hayward <alan.hayward@arm.com>
19218 David Sherwood <david.sherwood@arm.com>
19220 * tree-vect-stmts.c (simd_clone_subparts): New function.
19221 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19224 Alan Hayward <alan.hayward@arm.com>
19225 David Sherwood <david.sherwood@arm.com>
19227 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19228 vectors as polynomial. Use build_index_vector for
19229 IFN_GOMP_SIMD_LANE.
19231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19232 Alan Hayward <alan.hayward@arm.com>
19233 David Sherwood <david.sherwood@arm.com>
19235 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19236 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19237 for variable-length vectors.
19238 (vectorizable_mask_load_store): Treat the number of units as
19239 polynomial, asserting that it is constant if the condition has
19240 already been enforced.
19241 (vectorizable_store, vectorizable_load): Likewise.
19243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19244 Alan Hayward <alan.hayward@arm.com>
19245 David Sherwood <david.sherwood@arm.com>
19247 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19248 of units as polynomial. Punt if we can't tell at compile time
19249 which vector contains the final result.
19251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19252 Alan Hayward <alan.hayward@arm.com>
19253 David Sherwood <david.sherwood@arm.com>
19255 * tree-vect-loop.c (vectorizable_induction): Treat the number
19256 of units as polynomial. Punt on SLP inductions. Use an integer
19257 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19258 cast of such a series for variable-length floating-point
19261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19262 Alan Hayward <alan.hayward@arm.com>
19263 David Sherwood <david.sherwood@arm.com>
19265 * tree.h (build_index_vector): Declare.
19266 * tree.c (build_index_vector): New function.
19267 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19268 of units as polynomial, forcibly converting it to a constant if
19269 vectorizable_reduction has already enforced the condition.
19270 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19271 to create a {1,2,3,...} vector.
19272 (vectorizable_reduction): Treat the number of units as polynomial.
19273 Choose vectype_in based on the largest scalar element size rather
19274 than the smallest number of units. Enforce the restrictions
19277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19278 Alan Hayward <alan.hayward@arm.com>
19279 David Sherwood <david.sherwood@arm.com>
19281 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19282 number of units as polynomial.
19284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19285 Alan Hayward <alan.hayward@arm.com>
19286 David Sherwood <david.sherwood@arm.com>
19288 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19289 * target.def (autovectorize_vector_sizes): Return the vector sizes
19290 by pointer, using vector_sizes rather than a bitmask.
19291 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19292 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19293 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19295 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19296 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19297 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19298 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19299 * omp-general.c (omp_max_vf): Likewise.
19300 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19301 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19302 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19303 * tree-vect-slp.c (vect_slp_bb): Likewise.
19304 * doc/tm.texi: Regenerate.
19305 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19307 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19308 the vector size as a poly_uint64 rather than an unsigned int.
19309 (current_vector_size): Change from an unsigned int to a poly_uint64.
19310 (get_vectype_for_scalar_type): Update accordingly.
19311 * tree.h (build_truth_vector_type): Take the size and number of
19312 units as a poly_uint64 rather than an unsigned int.
19313 (build_vector_type): Add a temporary overload that takes
19314 the number of units as a poly_uint64 rather than an unsigned int.
19315 * tree.c (make_vector_type): Likewise.
19316 (build_truth_vector_type): Take the number of units as a poly_uint64
19317 rather than an unsigned int.
19319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19320 Alan Hayward <alan.hayward@arm.com>
19321 David Sherwood <david.sherwood@arm.com>
19323 * target.def (get_mask_mode): Take the number of units and length
19324 as poly_uint64s rather than unsigned ints.
19325 * targhooks.h (default_get_mask_mode): Update accordingly.
19326 * targhooks.c (default_get_mask_mode): Likewise.
19327 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19328 * doc/tm.texi: Regenerate.
19330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19331 Alan Hayward <alan.hayward@arm.com>
19332 David Sherwood <david.sherwood@arm.com>
19334 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19335 * omp-general.c (omp_max_vf): Likewise.
19336 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19337 (expand_omp_simd): Handle polynomial safelen.
19338 * omp-low.c (omplow_simd_context): Add a default constructor.
19339 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19340 (lower_rec_simd_input_clauses): Update accordingly.
19341 (lower_rec_input_clauses): Likewise.
19343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19344 Alan Hayward <alan.hayward@arm.com>
19345 David Sherwood <david.sherwood@arm.com>
19347 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19348 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19349 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19350 (vect_analyze_slp_cost): Likewise.
19351 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19352 (vect_model_load_cost): Likewise.
19354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19355 Alan Hayward <alan.hayward@arm.com>
19356 David Sherwood <david.sherwood@arm.com>
19358 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19359 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19360 from an unsigned int * to a poly_uint64_pod *.
19361 (calculate_unrolling_factor): New function.
19362 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19365 Alan Hayward <alan.hayward@arm.com>
19366 David Sherwood <david.sherwood@arm.com>
19368 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19369 from an unsigned int to a poly_uint64.
19370 (_loop_vec_info::slp_unrolling_factor): Likewise.
19371 (_loop_vec_info::vectorization_factor): Change from an int
19373 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19374 (vect_get_num_vectors): New function.
19375 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19376 (vect_get_num_copies): Use vect_get_num_vectors.
19377 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19378 to an unsigned int *.
19379 (vect_analyze_data_refs): Change min_vf from an int * to a
19381 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19382 than an unsigned HOST_WIDE_INT.
19383 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19384 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19385 to an unsigned int *.
19386 (vect_analyze_data_ref_dependences): Likewise.
19387 (vect_compute_data_ref_alignment): Handle polynomial vf.
19388 (vect_enhance_data_refs_alignment): Likewise.
19389 (vect_prune_runtime_alias_test_list): Likewise.
19390 (vect_shift_permute_load_chain): Likewise.
19391 (vect_supportable_dr_alignment): Likewise.
19392 (dependence_distance_ge_vf): Take the vectorization factor as a
19393 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19394 (vect_analyze_data_refs): Change min_vf from an int * to a
19396 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19397 vfm1 as a poly_uint64 rather than an int. Make the same change
19398 for the returned bound_scalar.
19399 (vect_gen_vector_loop_niters): Handle polynomial vf.
19400 (vect_do_peeling): Likewise. Update call to
19401 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19402 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19404 * tree-vect-loop.c (vect_determine_vectorization_factor)
19405 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19406 (vect_get_known_peeling_cost): Likewise.
19407 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19408 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19409 (vect_transform_loop): Likewise. Use the lowest possible VF when
19410 updating the upper bounds of the loop.
19411 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19412 rather than an int.
19413 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19414 polynomial unroll factors.
19415 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19416 (vect_make_slp_decision): Likewise.
19417 (vect_supported_load_permutation_p): Likewise, and polynomial
19419 (vect_analyze_slp_cost): Handle polynomial vf.
19420 (vect_slp_analyze_node_operations): Likewise.
19421 (vect_slp_analyze_bb_1): Likewise.
19422 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19423 than an unsigned HOST_WIDE_INT.
19424 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19425 (vectorizable_load): Handle polynomial vf.
19426 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19428 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19431 Alan Hayward <alan.hayward@arm.com>
19432 David Sherwood <david.sherwood@arm.com>
19434 * match.pd: Handle bit operations involving three constants
19435 and try to fold one pair.
19437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19439 * tree-vect-loop-manip.c: Include gimple-fold.h.
19440 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19441 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19442 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19443 Add a path that uses a step of VF instead of 1, but disable it
19445 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19446 and niters_no_overflow parameters. Update calls to
19447 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19448 Create a new SSA name if the latter choses to use a ste other
19449 than zero, and return it via niters_vector_mult_vf_var.
19450 * tree-vect-loop.c (vect_transform_loop): Update calls to
19451 vect_do_peeling, vect_gen_vector_loop_niters and
19452 slpeel_make_loop_iterate_ntimes.
19453 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19454 (vect_gen_vector_loop_niters): Update declarations after above changes.
19456 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19458 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19459 128-bit round to integer instructions.
19460 (ceil<mode>2): Likewise.
19461 (btrunc<mode>2): Likewise.
19462 (round<mode>2): Likewise.
19464 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19466 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19467 unaligned VSX load/store on P8/P9.
19468 (expand_block_clear): Allow the use of unaligned VSX
19469 load/store on P8/P9.
19471 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19473 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19475 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19476 swap associated with both a load and a store.
19478 2018-01-02 Andrew Waterman <andrew@sifive.com>
19480 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19481 * config/riscv/riscv.md (clear_cache): Use it.
19483 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19485 * web.c: Remove out-of-date comment.
19487 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19489 * expr.c (fixup_args_size_notes): Check that any existing
19490 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19491 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19492 (emit_single_push_insn): ...here.
19494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19496 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19497 (const_vector_encoded_nelts): New function.
19498 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19499 (const_vector_int_elt, const_vector_elt): Declare.
19500 * emit-rtl.c (const_vector_int_elt_1): New function.
19501 (const_vector_elt): Likewise.
19502 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19503 of CONST_VECTOR_ELT.
19505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19507 * expr.c: Include rtx-vector-builder.h.
19508 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19509 directly on the tree encoding.
19510 (const_vector_from_tree): Likewise.
19511 * optabs.c: Include rtx-vector-builder.h.
19512 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19513 sequence of "u" values.
19514 * vec-perm-indices.c: Include rtx-vector-builder.h.
19515 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19516 directly on the vec_perm_indices encoding.
19518 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19520 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19521 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19522 * rtx-vector-builder.h: New file.
19523 * rtx-vector-builder.c: Likewise.
19524 * rtl.h (rtx_def::u2): Add a const_vector field.
19525 (CONST_VECTOR_NPATTERNS): New macro.
19526 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19527 (CONST_VECTOR_DUPLICATE_P): Likewise.
19528 (CONST_VECTOR_STEPPED_P): Likewise.
19529 (CONST_VECTOR_ENCODED_ELT): Likewise.
19530 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19531 (unwrap_const_vec_duplicate): Likewise.
19532 (const_vec_series_p): Check for a non-duplicated vector encoding.
19533 Say that the function only returns true for integer vectors.
19534 * emit-rtl.c: Include rtx-vector-builder.h.
19535 (gen_const_vec_duplicate_1): Delete.
19536 (gen_const_vector): Call gen_const_vec_duplicate instead of
19537 gen_const_vec_duplicate_1.
19538 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19539 (gen_const_vec_duplicate): Use rtx_vector_builder.
19540 (gen_const_vec_series): Likewise.
19541 (gen_rtx_CONST_VECTOR): Likewise.
19542 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19543 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19544 Build a new vector rather than modifying a CONST_VECTOR in-place.
19545 (handle_special_swappables): Update call accordingly.
19546 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19547 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19548 Build a new vector rather than modifying a CONST_VECTOR in-place.
19549 (handle_special_swappables): Update call accordingly.
19551 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19553 * simplify-rtx.c (simplify_const_binary_operation): Use
19554 CONST_VECTOR_ELT instead of XVECEXP.
19556 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19558 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19559 the selector elements to be different from the data elements
19560 if the selector is a VECTOR_CST.
19561 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19562 ssizetype for the selector.
19564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19566 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19567 before testing each element individually.
19568 * tree-vect-generic.c (lower_vec_perm): Likewise.
19570 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19572 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19573 * selftest-run-tests.c (selftest::run_tests): Call it.
19574 * vector-builder.h (vector_builder::operator ==): New function.
19575 (vector_builder::operator !=): Likewise.
19576 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19577 (vec_perm_indices::all_from_input_p): New function.
19578 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19579 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19580 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19581 instead of reading the VECTOR_CST directly. Detect whether both
19582 vector inputs are the same before constructing the vec_perm_indices,
19583 and update the number of inputs argument accordingly. Use the
19584 utility functions added above. Only construct sel2 if we need to.
19586 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19588 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19589 the broadcast of the low byte.
19590 (expand_mult_highpart): Use an explicit encoding for the permutes.
19591 * optabs-query.c (can_mult_highpart_p): Likewise.
19592 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19593 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19594 (vectorizable_bswap): Likewise.
19595 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19596 explicit encoding for the power-of-2 permutes.
19597 (vect_permute_store_chain): Likewise.
19598 (vect_grouped_load_supported): Likewise.
19599 (vect_permute_load_chain): Likewise.
19601 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19603 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19604 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19605 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19606 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19607 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19608 (vect_gen_perm_mask_any): Likewise.
19610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19612 * int-vector-builder.h: New file.
19613 * vec-perm-indices.h: Include int-vector-builder.h.
19614 (vec_perm_indices): Redefine as an int_vector_builder.
19615 (auto_vec_perm_indices): Delete.
19616 (vec_perm_builder): Redefine as a stand-alone class.
19617 (vec_perm_indices::vec_perm_indices): New function.
19618 (vec_perm_indices::clamp): Likewise.
19619 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19620 (vec_perm_indices::new_vector): New function.
19621 (vec_perm_indices::new_expanded_vector): Update for new
19622 vec_perm_indices class.
19623 (vec_perm_indices::rotate_inputs): New function.
19624 (vec_perm_indices::all_in_range_p): Operate directly on the
19625 encoded form, without computing elided elements.
19626 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19627 encoding. Update for new vec_perm_indices class.
19628 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19629 the given vec_perm_builder.
19630 (expand_vec_perm_var): Update vec_perm_builder constructor.
19631 (expand_mult_highpart): Use vec_perm_builder instead of
19632 auto_vec_perm_indices.
19633 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19634 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19635 or double series encoding as appropriate.
19636 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19637 vec_perm_indices instead of auto_vec_perm_indices.
19638 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19639 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19640 (vect_permute_store_chain): Likewise.
19641 (vect_grouped_load_supported): Likewise.
19642 (vect_permute_load_chain): Likewise.
19643 (vect_shift_permute_load_chain): Likewise.
19644 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19645 (vect_transform_slp_perm_load): Likewise.
19646 (vect_schedule_slp_instance): Likewise.
19647 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19648 (vectorizable_mask_load_store): Likewise.
19649 (vectorizable_bswap): Likewise.
19650 (vectorizable_store): Likewise.
19651 (vectorizable_load): Likewise.
19652 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19653 vec_perm_indices instead of auto_vec_perm_indices. Use
19654 tree_to_vec_perm_builder to read the vector from a tree.
19655 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19656 vec_perm_builder instead of a vec_perm_indices.
19657 (have_whole_vector_shift): Use vec_perm_builder and
19658 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19659 truncation to calc_vec_perm_mask_for_shift.
19660 (vect_create_epilog_for_reduction): Likewise.
19661 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19662 from auto_vec_perm_indices to vec_perm_indices.
19663 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19664 instead of changing individual elements.
19665 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19666 the vector in d.perm.
19667 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19668 from auto_vec_perm_indices to vec_perm_indices.
19669 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19670 instead of changing individual elements.
19671 (arm_vectorize_vec_perm_const): Use new_vector to install
19672 the vector in d.perm.
19673 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19674 Update vec_perm_builder constructor.
19675 (rs6000_expand_interleave): Likewise.
19676 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19677 (rs6000_expand_interleave): Likewise.
19679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19681 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19682 to qimode could truncate the indices.
19683 * optabs.c (expand_vec_perm_var): Likewise.
19685 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19687 * Makefile.in (OBJS): Add vec-perm-indices.o.
19688 * vec-perm-indices.h: New file.
19689 * vec-perm-indices.c: Likewise.
19690 * target.h (vec_perm_indices): Replace with a forward class
19692 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19693 * optabs.h: Include vec-perm-indices.h.
19694 (expand_vec_perm): Delete.
19695 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19696 (expand_vec_perm_const): Declare.
19697 * target.def (vec_perm_const_ok): Replace with...
19698 (vec_perm_const): ...this new hook.
19699 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19700 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19701 * doc/tm.texi: Regenerate.
19702 * optabs.def (vec_perm_const): Delete.
19703 * doc/md.texi (vec_perm_const): Likewise.
19704 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19705 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19706 expand_vec_perm for constant permutation vectors. Assert that
19707 the mode of variable permutation vectors is the integer equivalent
19708 of the mode that is being permuted.
19709 * optabs-query.h (selector_fits_mode_p): Declare.
19710 * optabs-query.c: Include vec-perm-indices.h.
19711 (selector_fits_mode_p): New function.
19712 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19713 is defined, instead of checking whether the vec_perm_const_optab
19714 exists. Use targetm.vectorize.vec_perm_const instead of
19715 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19716 fit in the vector mode before using a variable permute.
19717 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19718 vec_perm_indices instead of an rtx.
19719 (expand_vec_perm): Replace with...
19720 (expand_vec_perm_const): ...this new function. Take the selector
19721 as a vec_perm_indices rather than an rtx. Also take the mode of
19722 the selector. Update call to shift_amt_for_vec_perm_mask.
19723 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19724 Use vec_perm_indices::new_expanded_vector to expand the original
19725 selector into bytes. Check whether the indices fit in the vector
19726 mode before using a variable permute.
19727 (expand_vec_perm_var): Make global.
19728 (expand_mult_highpart): Use expand_vec_perm_const.
19729 * fold-const.c: Includes vec-perm-indices.h.
19730 * tree-ssa-forwprop.c: Likewise.
19731 * tree-vect-data-refs.c: Likewise.
19732 * tree-vect-generic.c: Likewise.
19733 * tree-vect-loop.c: Likewise.
19734 * tree-vect-slp.c: Likewise.
19735 * tree-vect-stmts.c: Likewise.
19736 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19738 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19739 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19740 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19741 (aarch64_vectorize_vec_perm_const): ...this new function.
19742 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19743 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19744 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19745 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19746 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19747 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19748 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19750 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19751 check for NEON modes.
19752 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19753 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19754 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19755 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19757 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19758 the old VEC_PERM_CONST conditions.
19759 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19760 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19761 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19762 (ia64_vectorize_vec_perm_const_ok): Merge into...
19763 (ia64_vectorize_vec_perm_const): ...this new function.
19764 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19765 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19766 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19767 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19768 * config/mips/mips.c (mips_expand_vec_perm_const)
19769 (mips_vectorize_vec_perm_const_ok): Merge into...
19770 (mips_vectorize_vec_perm_const): ...this new function.
19771 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19772 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19773 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19774 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19775 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19776 (rs6000_expand_vec_perm_const): Delete.
19777 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19779 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19780 (altivec_expand_vec_perm_const_le): Take each operand individually.
19781 Operate on constant selectors rather than rtxes.
19782 (altivec_expand_vec_perm_const): Likewise. Update call to
19783 altivec_expand_vec_perm_const_le.
19784 (rs6000_expand_vec_perm_const): Delete.
19785 (rs6000_vectorize_vec_perm_const_ok): Delete.
19786 (rs6000_vectorize_vec_perm_const): New function.
19787 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19788 an element count and rtx array.
19789 (rs6000_expand_extract_even): Update call accordingly.
19790 (rs6000_expand_interleave): Likewise.
19791 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19792 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19793 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19794 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19795 (rs6000_expand_vec_perm_const): Delete.
19796 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19797 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19798 (altivec_expand_vec_perm_const_le): Take each operand individually.
19799 Operate on constant selectors rather than rtxes.
19800 (altivec_expand_vec_perm_const): Likewise. Update call to
19801 altivec_expand_vec_perm_const_le.
19802 (rs6000_expand_vec_perm_const): Delete.
19803 (rs6000_vectorize_vec_perm_const_ok): Delete.
19804 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19805 reference to the SPE evmerge intructions.
19806 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19807 an element count and rtx array.
19808 (rs6000_expand_extract_even): Update call accordingly.
19809 (rs6000_expand_interleave): Likewise.
19810 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19811 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19813 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19817 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19818 vector mode and that that mode matches the mode of the data
19820 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19821 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19822 directly using expand_vec_perm_1 when forcing selectors into
19824 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19826 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19828 * optabs-query.h (can_vec_perm_p): Delete.
19829 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19830 * optabs-query.c (can_vec_perm_p): Split into...
19831 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19832 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19833 particular selector is valid.
19834 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19835 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19836 (vect_grouped_load_supported): Likewise.
19837 (vect_shift_permute_load_chain): Likewise.
19838 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19839 (vect_transform_slp_perm_load): Likewise.
19840 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19841 (vectorizable_bswap): Likewise.
19842 (vect_gen_perm_mask_checked): Likewise.
19843 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19844 implementations of variable permutation vectors into account
19845 when deciding which selector to use.
19846 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19847 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19848 with a false third argument.
19849 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19850 to test whether the constant selector is valid and can_vec_perm_var_p
19851 to test whether a variable selector is valid.
19853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19855 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19856 * optabs-query.c (can_vec_perm_p): Likewise.
19857 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19858 instead of vec_perm_indices.
19859 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19860 (vect_gen_perm_mask_checked): Likewise,
19861 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19862 (vect_gen_perm_mask_checked): Likewise,
19864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19866 * optabs-query.h (qimode_for_vec_perm): Declare.
19867 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19868 (qimode_for_vec_perm): ...this new function.
19869 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19871 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19873 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19874 does not have a conditional at the top.
19876 2018-01-02 Richard Biener <rguenther@suse.de>
19878 * ipa-inline.c (big_speedup_p): Fix expression.
19880 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19883 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19886 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19890 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19891 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19892 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19893 cond_taken_branch_cost 3->4.
19895 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19897 PR tree-optimization/83581
19898 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19899 TODO_cleanup_cfg if any changes have been made.
19901 PR middle-end/83608
19902 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19903 convert_modes if target mode has the right side, but different mode
19906 PR middle-end/83609
19907 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19908 last argument when extracting from CONCAT. If either from_real or
19909 from_imag is NULL, use expansion through memory. If result is not
19910 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19911 the parts directly to inner mode, if even that fails, use expansion
19914 PR middle-end/83623
19915 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19916 check for bswap in mode rather than HImode and use that in expand_unop
19919 Copyright (C) 2018 Free Software Foundation, Inc.
19921 Copying and distribution of this file, with or without modification,
19922 are permitted in any medium without royalty provided the copyright
19923 notice and this notice are preserved.