1 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/86526
4 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5 (inline_expand_builtin_string_cmp): Likewise.
6 (inline_string_cmp): Likewise. Use c_readstr instead of
7 builtin_memcpy_read_str. Add unit_mode temporary.
9 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
12 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
13 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
15 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
17 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
18 Alphabetize prototypes of built-in functions, separating out
19 built-in functions that are listed in this section but should be
22 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
25 * expmed.c (emit_store_flag): Do not emit setcc followed by a
26 conditional move when trapping comparison was split to a
27 non-trapping one (and vice versa).
29 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
31 * config/s390/s390.c (s390_function_profiler): Generate nops
32 instead of profiler call sequences.
33 * config/s390/s390.opt: Add the new option.
35 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
37 * config/s390/s390.c (s390_function_profiler): Generate
39 * config/s390/s390.opt: Add the new option.
41 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
43 * common.opt: Add the new warning.
44 * config/s390/s390.c (s390_function_profiler): Emit "brasl
45 %r0,__fentry__" when -mfentry is specified.
46 (s390_option_override_internal): Disallow -mfentry for 31-bit
48 * config/s390/s390.opt: Add the new option.
50 2018-07-16 Richard Biener <rguenther@suse.de>
53 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
54 for function-local FUNCTION_DECL and RESULT_DECL immediately.
56 2018-07-16 Martin Liska <mliska@suse.cz>
59 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
62 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
64 * config/arc/arcHS.md: Update ARCHS scheduling rules.
66 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
68 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
70 * config/arc/arc-cpus.def (hs4x): New cpu.
72 * config/arc/arc-tables.opt: Regenerate.
73 * config/arc/arc.c (arc_sched_issue_rate): New function.
74 (TARGET_SCHED_ISSUE_RATE): Define.
75 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
76 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
78 (attr tune): Add ARCHS4x tune values.
79 (attr tune_dspmpy): Define.
80 (*tst): Correct instruction type.
81 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
82 * config/arc/arcHS4x.md: New file.
83 * config/arc/fpu.md: Update instruction type attributes.
84 * config/arc/t-multilib: Regenerate.
86 2018-07-16 Tom de Vries <tdevries@suse.de>
89 * var-tracking.c (vt_initialize): Fix pre_dec handling.
91 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
93 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
96 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
98 PR tree-optimization/86514
99 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
100 conversion to a boolean type from a type with greater precision.
102 2018-07-16 Tom de Vries <tdevries@suse.de>
104 * var-tracking.c (vt_initialize): Print adjusted insn slim if
105 dump_flags request TDF_SLIM.
107 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
109 * fold-const.c (int_const_binop_1): Abstract...
110 (wide_int_binop): ...wide int code here.
111 (poly_int_binop): ...poly int code here.
112 Abstract the rest of int_const_binop_1 into int_const_binop.
113 * fold-const.h (wide_int_binop): New.
114 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
115 Remove useless PLUS/MINUS_EXPR case.
116 (zero_nonzero_bits_from_vr): Move wide int code...
117 (zero_nonzero_bits_from_bounds): ...here.
118 (extract_range_from_binary_expr_1): Move mask optimization code...
119 (range_easy_mask_min_max): ...here.
120 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
121 (range_easy_mask_min_max): New.
123 2018-07-15 Jeff Law <law@redhat.com>
126 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
129 2018-07-14 Jim Wilson <jimw@sifive.com>
131 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
133 2018-07-14 Paul Koning <ni1d@arrl.net>
135 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
137 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
139 * lto-streamer-out.c (copy_function_or_variable): Dump info about
142 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
143 Steve Munroe <munroesj52@gmail.com>
145 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
146 (_mm_andnot_si128): Likewise.
147 (_mm_or_si128): Likewise.
148 (_mm_xor_si128): Likewise.
150 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
153 * builtins.c (expand_builtin_memcmp): Inline the calls first
154 when result_eq is false.
155 (expand_builtin_strcmp): Inline the calls first.
156 (expand_builtin_strncmp): Likewise.
157 (inline_string_cmp): New routine. Expand a string compare
158 call by using a sequence of char comparison.
159 (inline_expand_builtin_string_cmp): New routine. Inline expansion
160 a call to str(n)cmp/memcmp.
161 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
163 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
165 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
167 * config/arm/driver-arm.c: Include arm-native.h.
168 (host_detect_local_cpu): Use auto-generated data tables.
169 (vendors, arm_cpu_table): Delete. Move part information to ...
170 * config/arm/arm-cpus.in: ... here.
171 * config/arm/parsecpu.awk (gen_native): New function.
172 (vendor, part): New CPU fields.
173 (END): Add support for building the native CPU detection tables.
174 * config/arm/t-arm (arm-native.h): Add build rule.
175 (driver-arm.o): Add dependency on arm-native.h.
177 2018-07-13 Richard Biener <rguenther@suse.de>
180 * match.pd (addr1 - addr2): Allow either of the operand to
183 2018-07-13 Tom de Vries <tdevries@suse.de>
185 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
188 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
190 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
191 arrays instead of numbered variables.
193 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
195 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
196 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
198 2018-07-13 Richard Biener <rguenther@suse.de>
201 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
202 instead of get_context_die.
204 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
205 Richard Biener <rguenther@suse.de>
208 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
209 that the loop latch destination where phi is defined.
211 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
213 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
214 (riscv_expand_epilogue): Add assertion to check interrupt mode.
215 (riscv_set_current_function): Extract getting interrupt type to new
217 (riscv_get_interrupt_type): New function.
218 (riscv_merge_decl_attributes): New function, checking interrupt type is
220 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
222 2018-07-12 Paul Koning <ni1d@arrl.net>
224 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
227 2018-07-12 Paul Koning <ni1d@arrl.net>
229 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
230 zero reference, add doloop_end instead.
231 * doc/md.texi (decrement_and_branch_until_zero): Remove.
232 (Looping patterns): Remove decrement_and_branch_until_zero. Add
233 detail for doloop_end.
235 2018-07-12 Martin Sebor <msebor@redhat.com>
238 * attribs.c (decl_attributes): Reject conflicting attributes before
239 calling attribute handlers.
241 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
243 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
245 (gcc::dump_manager::get_dump_file_name): likewise.
246 (dump_begin): Likewise.
247 * dumpfile.h (dump_begin): Update prototype.
248 (gcc::dump_manager::get_dump_file_name,
249 gcc::dump_manager::get_dump_file_name): Update prototype.
251 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
253 * internal-fn.h (vectorizable_internal_fn_p): New function.
254 * tree-vect-slp.c (compatible_calls_p): Likewise.
255 (vect_build_slp_tree_1): Remove nops argument. Handle calls
256 to internal functions.
257 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
259 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
261 * fold-const.h (inverse_conditions_p): Declare.
262 * fold-const.c (inverse_conditions_p): New function.
263 * match.pd: Use inverse_conditions_p. Add folds of view_converts
264 that test the inverse condition of a conditional internal function.
265 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
266 * internal-fn.c (internal_fn_mask_index): Handle conditional
268 (vectorized_internal_fn_supported_p): New function.
269 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
270 (any_pred_load_store): Replace with...
271 (need_to_predicate): ...this new variable.
272 (redundant_ssa_names): New variable.
273 (ifcvt_can_use_mask_load_store): Move initial checks to...
274 (ifcvt_can_predicate): ...this new function. Handle tree codes
275 for which a conditional internal function exists.
276 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
277 instead of ifcvt_can_use_mask_load_store. Update after variable
279 (predicate_load_or_store): New function, split out from
280 predicate_mem_writes.
281 (check_redundant_cond_expr): New function.
282 (value_available_p): Likewise.
283 (predicate_rhs_code): Likewise.
284 (predicate_mem_writes): Rename to...
285 (predicate_statements): ...this. Use predicate_load_or_store
286 and predicate_rhs_code.
287 (combine_blocks, tree_if_conversion): Update after above name changes.
288 (ifcvt_local_dce): Handle redundant_ssa_names.
289 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
290 general conditional functions.
291 * tree-vect-stmts.c (vectorizable_call): Likewise.
293 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
294 Alan Hayward <alan.hayward@arm.com>
295 David Sherwood <david.sherwood@arm.com>
297 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
298 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
299 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
300 plus and minus and convert them into IFN_COND_FMA-based sequences.
301 (convert_mult_to_fma): Handle conditional plus and minus.
303 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
305 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
306 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
307 (cond_fnms_optab): New optabs.
308 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
310 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
311 * internal-fn.h (get_conditional_internal_fn): Declare.
312 (get_unconditional_internal_fn): Likewise.
313 * internal-fn.c (cond_ternary_direct): New macro.
314 (expand_cond_ternary_optab_fn): Likewise.
315 (direct_cond_ternary_optab_supported_p): Likewise.
316 (FOR_EACH_COND_FN_PAIR): Likewise.
317 (get_conditional_internal_fn): New function.
318 (get_unconditional_internal_fn): Likewise.
319 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
320 (gimple_match_op::gimple_match_op): Add a new overload for 5
322 (gimple_match_op::set_op): Likewise.
323 (gimple_resimplify5): Declare.
324 * genmatch.c (decision_tree::gen): Generate simplifications for
326 * gimple-match-head.c (gimple_simplify): Define an overload for
327 5 operands. Handle calls with 5 arguments in the top-level overload.
328 (convert_conditional_op): Handle conversions from unconditional
329 internal functions to conditional ones.
330 (gimple_resimplify5): New function.
331 (build_call_internal): Pass a fifth operand.
332 (maybe_push_res_to_seq): Likewise.
333 (try_conditional_simplification): Try converting conditional
334 internal functions to unconditional internal functions.
335 Handle 3-operand unconditional forms.
336 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
337 Define ternary equivalents of the current rules for binary conditional
339 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
341 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
342 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
343 (optab): Handle them.
344 (SVE_COND_FP_TERNARY): New int iterator.
345 (sve_fmla_op, sve_fmad_op): New int attributes.
346 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
347 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
348 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
350 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
352 * target.def (preferred_else_value): New target hook.
353 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
354 * doc/tm.texi: Regenerate.
355 * targhooks.h (default_preferred_else_value): Declare.
356 * targhooks.c (default_preferred_else_value): New function.
357 * internal-fn.h (conditional_internal_fn_code): Declare.
358 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
359 (get_conditional_internal_fn): Use it.
360 (conditional_internal_fn_code): New function.
361 * gimple-match.h (gimple_match_cond): New struct.
362 (gimple_match_op): Add a cond member function.
363 (gimple_match_op::gimple_match_op): Update all forms to take a
365 * genmatch.c (expr::gen_transform): Use the same condition as res_op
366 for the suboperation, but don't specify a particular else_value.
367 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
368 (visit_nary_op, visit_reference_op_load): Pass
369 gimple_match_cond::UNCOND to the gimple_match_op constructor.
370 * gimple-match-head.c: Include tree-eh.h
371 (convert_conditional_op): New function.
372 (maybe_resimplify_conditional_op): Likewise.
373 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
374 (gimple_resimplify2): Likewise.
375 (gimple_resimplify3): Likewise.
376 (gimple_resimplify4): Likewise.
377 (maybe_push_res_to_seq): Return null for conditional operations.
378 (try_conditional_simplification): New function.
379 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
381 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
383 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
385 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
387 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
389 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
391 (hash_tree): Do not hash DECL_FCONTEXT
392 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
393 Do not stream DECL_FCONTEXT.
394 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
395 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
397 2018-07-12 Richard Biener <rguenther@suse.de>
400 * dwarf2out.c (gen_block_die): Only output blocks when they have
401 at least one !DECL_IGNORED_P variable.
403 2018-07-12 Richard Biener <rguenther@suse.de>
406 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
407 Remove -mieee-fp handling.
409 2018-07-12 Richard Biener <rguenther@suse.de>
411 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
412 left-over from last patch.
414 2018-07-12 Jakub Jelinek <jakub@redhat.com>
416 PR tree-optimization/86492
417 * gimple-ssa-store-merging.c
418 (imm_store_chain_info::coalesce_immediate_stores): Call
419 check_no_overlap even for the merge_overlapping case. Formatting fix.
421 2018-07-12 Richard Biener <rguenther@suse.de>
424 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
425 move possibly trapping operations into the conditional.
427 2018-07-12 Richard Biener <rguenther@suse.de>
429 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
430 (vn_lookup_simplify_result): Remove recursion limit applied
432 (vn_nary_build_or_lookup_1): Adjust.
433 (try_to_simplify): Likewise.
434 * gimple-match-head.c (gimple_resimplify1): Instead apply one
436 (gimple_resimplify2): Likewise.
437 (gimple_resimplify3): Likewise.
438 (gimple_resimplify4): Likewise.
440 2018-07-11 Jakub Jelinek <jakub@redhat.com>
442 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
443 Use __mmask64 type instead of __mmask8 for __M argument.
444 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
445 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
447 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
448 __mmask16 for __M argument.
449 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
450 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
451 to __mmask16 instead of __mmask8.
452 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
453 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
454 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
455 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
456 instead of __mmask16 for __U argument.
457 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
458 __mmask16 instead of __mmask8 for __U argument.
459 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
461 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
463 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
465 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
467 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
469 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
471 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
473 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
475 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
477 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
479 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
480 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
481 return type as well as __M argument type and all casts from __mmask8
483 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
484 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
485 return type as well as __M argument type and all casts from __mmask8
487 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
488 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
489 return type as well as __M argument type and all casts from __mmask8
491 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
492 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
493 return type as well as __M argument type and all casts from __mmask8
495 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
496 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
499 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
501 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
502 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
505 2018-07-11 Paul Koning <ni1d@arrl.net>
507 * doc/md.texi (define_subst): Document how multiple occurrences of
508 the same argument in the replacement pattern are handled.
510 2018-07-11 Paul Koning <ni1d@arrl.net>
512 * doc/extend.texi (Common Variable Attributes): Move "mode" into
514 (Common Type Attributes): Add "mode" attribute.
516 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
518 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
519 stream DECL_ORIGINAL_TYPE.
520 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
521 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
522 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
523 Do not walk original type.
524 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
526 (write_ts_decl_non_common_tree_pointers): Do not stream
528 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
529 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
531 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
533 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
534 thread twice from the same starting edge.
536 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
538 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
539 * gimple.c (gimple_call_nonnull_result_p): ...here...
540 (gimple_call_nonnull_arg): ...and here.
541 * gimple.h (gimple_call_nonnull_result_p): New.
542 (gimple_call_nonnull_arg): New.
544 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
546 * config/arm/arm-cpus.in: Move information from fpu field of each
547 cpu definition to the isa field.
548 * config/arm/parsecpu.awk (fpu): Delete match rule.
549 (gen_comm_data): Don't add bits from the CPU's FPU entry.
551 2018-07-11 Richard Biener <rguenther@suse.de>
554 * dwarf2out.c (init_sections_and_labels): Use
555 output_asm_line_debug_info consistently.
556 (dwarf2out_early_finish): Likewise.
557 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
560 2018-07-11 Richard Biener <rguenther@suse.de>
562 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
563 Rework father_bb setting in a way to avoid propagating constants
564 multiple times on a loop body.
566 2018-07-10 Mark Wielaard <mark@klomp.org>
569 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
571 2018-07-10 Richard Biener <rguenther@suse.de>
573 * hash-map.h (hash_map::iterator::operator*): Return
574 references to key and value.
576 2018-07-10 Jakub Jelinek <jakub@redhat.com>
579 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
580 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
581 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
582 (gimplify_omp_for): For composite loops, move outer
583 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
584 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
585 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
586 TREE_LIST for both the original class iterator and the "last" helper
587 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
588 loop, remember has_decl_expr from outer composite loops for the
589 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
591 2018-07-09 Martin Sebor <msebor@redhat.com>
595 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
596 accessing implicitly initialized array elements.
597 * expr.c (string_constant): Handle string initializers of
598 character arrays within aggregates.
599 * gimple-fold.c (fold_array_ctor_reference): Add argument.
600 Store element offset. As a special case, handle zero size.
601 (fold_nonarray_ctor_reference): Same.
602 (fold_ctor_reference): Add argument. Store subobject offset.
603 * gimple-fold.h (fold_ctor_reference): Add argument.
605 2018-07-09 Paul Koning <ni1d@arrl.net>
607 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
608 (pdp11_insn_cost): New function.
609 (pdp11_md_asm_adjust): New function.
610 (TARGET_INVALID_WITHIN_DOLOOP): Define.
611 (pdp11_rtx_costs): Update to match machine better.
612 (output_addr_const_pdp11): Correct format mismatch warnings.
613 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
614 * config/pdp11/pdp11.md: General change to add base_cost and/or
615 length attributes for use by new pdp11_insn_cost function.
616 (MIN_BRANCH): Correct definition.
618 (doloop_end): Use standard pattern name for looping pattern.
619 (doloop_end_nocc): New.
620 (movsf): Add another constraint alternative.
621 (zero_extendqihi2): Add constraint alternatives for not in place
623 (zero_extendhisi2): Remove.
624 (shift patterns): Add CC handling variants.
628 (define_peephole2): New peephole to recognize mov that sets CC for
631 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
633 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
634 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
637 2018-07-09 Jakub Jelinek <jakub@redhat.com>
640 * real.c (real_nextafter): Return true if result is denormal.
642 2018-07-09 Martin Liska <mliska@suse.cz>
644 * common.opt: Add back wrongly removed attribute.
646 2018-07-09 Richard Biener <rguenther@suse.de>
649 * dwarf2out.c (gen_block_die): For an early generated DIE
650 always output high/low PC attributes.
652 2018-07-09 Tom de Vries <tdevries@suse.de>
654 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
655 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
656 onto VAR_DECL with abstract origin.
658 2018-07-07 Jim Wilson <jimw@sifive.com>
660 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
662 2018-07-07 Tom de Vries <tdevries@suse.de>
664 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
666 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
668 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
670 (combine_bound): Use wide-int overflow calculation instead of
672 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
674 * fold-const.c (int_const_binop_2): Same.
675 (extract_muldiv_1): Same.
676 (fold_div_compare): Same.
677 (fold_abs_const): Same.
679 * poly-int.h (add): Same.
683 * predict.c (predict_iv_comparison): Same.
684 * profile-count.c (slow_safe_scale_64bit): Same.
685 * simplify-rtx.c (simplify_const_binary_operation): Same.
686 * tree-chrec.c (tree_fold_binomial): Same.
687 * tree-data-ref.c (split_constant_offset_1): Same.
688 * tree-if-conv.c (idx_within_array_bound): Same.
689 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
690 * tree-ssa-phiopt.c (minmax_replacement): Same.
691 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
692 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
693 * vr-values.c (vr_values::adjust_range_with_scev): Same.
694 * wide-int.cc (wi::add_large): Same.
695 (wi::mul_internal): Same.
696 (wi::sub_large): Same.
697 (wi::divmod_internal): Same.
698 * wide-int.h: Change overflow type to overflow_type for neg, add,
699 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
700 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
701 mul_internal, divmod_internal.
702 (overflow_type): New enum.
703 (accumulate_overflow): New.
705 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
707 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
708 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
710 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
712 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
713 argument is checked for zero before entering loop, avoid checking again.
715 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
717 * gimplify.h (generic_expr_could_trap_p): Set as global function.
718 * gimplify.h (generic_expr_could_trap_p): Likwise.
719 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
721 2018-07-06 Jakub Jelinek <jakub@redhat.com>
723 PR tree-optimization/86401
724 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
725 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
726 (fold_bit_and_mask): ... here. New helper function for match.pd.
727 * fold-const.h (fold_bit_and_mask): Declare.
728 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
730 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
733 * target.def (translate_mode_attribute): New hook.
734 * targhooks.h (default_translate_mode_attribute): Declare.
735 * targhooks.c (default_translate_mode_attribute): New function.
736 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
737 * doc/tm.texi: Regenerate.
738 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
739 (rs6000_translate_mode_attribute): New function.
741 2018-07-06 Paul Koning <ni1d@arrl.net>
743 * doc/md.texi (define_split): Document DONE and FAIL.
744 (define_peephole2): Ditto.
746 2018-07-05 Jeff Law <law@redhat.com>
748 PR tree-optimization/86010
749 * tree-ssa-dse.c (compute_trims): More aggressively trim at
750 both the head and tail of mem* and str* calls.
752 2018-07-05 Jim Wilson <jimw@sifive.com>
754 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
756 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
758 * config/aarch64/aarch64-simd.md: correct flags text for
759 MIN_EXPR replacement.
761 2018-07-05 James Clarke <jrtc27@jrtc27.com>
763 * configure: Regenerated.
765 2018-07-05 Carl Love <cel@us.ibm.com>
767 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
768 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
769 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
770 VSX_BUILTIN_DOUBLEL_V4SF.
772 2018-07-05 Martin Sebor <msebor@redhat.com>
775 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
776 than its domain to compute its the upper bound of a char array.
778 2018-07-05 Nathan Sidwell <nathan@acm.org>
780 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
781 * doc/cpp.texi: Update comment.
782 * doc/tm.texi: Rebuilt.
783 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
784 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
785 * doc/extend.texi (Backwards Compatibility): Clarify it is system
786 headers affected by extern "C".
787 * system.h: Poison NO_IMPLICIT_EXTERN_C.
788 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
789 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
790 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
791 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
792 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
793 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
794 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
795 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
796 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
797 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
798 config/sparc/sp64-elf.h, config/spu/spu.h,
799 config/stormy16/stormy16.h, config/v850/v850.h,
800 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
801 define NO_IMPLICIT_EXTERN_C.
802 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
804 2018-07-05 Tamar Christina <tamar.christina@arm.com>
807 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
808 instead of GET_MODE_SIZE when comparing Units.
810 2018-07-05 Tamar Christina <tamar.christina@arm.com>
813 * rtlanal.c (set_noop_p): Constrain on mode change,
814 include hard-reg-set.h
816 2018-07-05 Tamar Christina <tamar.christina@arm.com>
818 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
820 2018-07-05 Jakub Jelinek <jakub@redhat.com>
823 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
826 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
828 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
830 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
833 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
835 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
837 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
840 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
842 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
844 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
847 2018-07-04 Martin Liska <mliska@suse.cz>
849 * tree-switch-conversion.c: Define
850 max_ratio_for_speed and max_ratio_for_size constants.
852 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
853 Martin Liska <mliska@suse.cz>
858 * common.opt: Rename align options with 'str_' prefix.
859 * common/config/i386/i386-common.c (set_malign_value): New
861 (ix86_handle_option): Use it to set -falign-* options/
862 * config/aarch64/aarch64-protos.h (struct tune_params): Change
863 type from int to string.
864 * config/aarch64/aarch64.c: Update default values from int
866 * config/alpha/alpha.c (alpha_override_options_after_change):
868 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
869 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
870 max skip conditionally.
871 * config/i386/freebsd.h (SUBALIGN_LOG): New.
872 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
873 max skip conditionally.
874 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
875 max skip conditionally.
876 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
877 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
878 max skip conditionally.
879 * config/i386/i386.c (struct ptt): Change type from int to
881 (ix86_default_align): Set default values.
882 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
883 max skip conditionally.
884 * config/i386/iamcu.h (SUBALIGN_LOG): New.
885 (ASM_OUTPUT_MAX_SKIP_ALIGN):
886 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
887 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
888 max skip conditionally.
889 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
890 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
891 * config/i386/x86-64.h (SUBALIGN_LOG): New.
892 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
893 max skip conditionally.
894 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
895 * config/ia64/ia64.c (ia64_option_override): Set default values
896 for alignment options.
897 * config/m68k/m68k.c: Handle new str_align_* options.
898 * config/mips/mips.c (mips_set_compression_mode): Change
900 (mips_option_override): Set default values for options.
901 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
903 * config/rs6000/rs6000.c (rs6000_option_override_internal):
905 * config/rx/rx.c (rx_option_override): Likewise.
906 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
907 (LABEL_ALIGN): Use align_labels_log.
908 (LOOP_ALIGN): Use align_loops_align.
909 * config/s390/s390.c (s390_asm_output_function_label): Use new
911 * config/sh/sh.c (sh_override_options_after_change):
912 Change type of constants.
913 * config/spu/spu.c (spu_sched_init): Likewise.
914 * config/sparc/sparc.c (sparc_option_override): Set default
916 * config/visium/visium.c (visium_option_override): Likewise.
917 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
918 emit p2align format with last argument if it's not needed.
919 * doc/invoke.texi: Document extended format of -falign-*.
920 * final.c: Use align_labels alignment.
921 * flags.h (struct target_flag_state): Change type to use
923 (struct align_flags_tuple): New.
924 (struct align_flags): Likewise.
925 (align_loops_log): Redefine macro to use new types.
926 (align_loops_max_skip): Redefine macro to use new types.
927 (align_jumps_log): Redefine macro to use new types.
928 (align_jumps_max_skip): Redefine macro to use new types.
929 (align_labels_log): Redefine macro to use new types.
930 (align_labels_max_skip): Redefine macro to use new types.
931 (align_functions_log): Redefine macro to use new types.
932 (align_loops): Redefine macro to use new types.
933 (align_jumps): Redefine macro to use new types.
934 (align_labels): Redefine macro to use new types.
935 (align_functions): Redefine macro to use new types.
936 (align_functions_max_skip): Redefine macro to use new types.
937 (align_loops_value): New macro.
938 (align_jumps_value): New macro.
939 (align_labels_value): New macro.
940 (align_functions_value): New macro.
941 * function.c (invoke_set_current_function_hook): Propagate
942 alignment values from flags to global variables default in
944 * ipa-icf.c (sem_function::equals_wpa): Use
945 cl_optimization_option_eq instead of memcmp.
946 * lto-streamer.h (cl_optimization_stream_out): Support streaming
948 (cl_optimization_stream_in): Likewise.
949 * optc-save-gen.awk: Support strings in cl_optimization.
950 * opth-gen.awk: Likewise.
951 * opts.c (finish_options): Remove error checking of invalid
953 (MAX_CODE_ALIGN): Remove.
954 (MAX_CODE_ALIGN_VALUE): Likewise.
955 (parse_and_check_align_values): New function.
956 (check_alignment_argument): Likewise.
957 (common_handle_option): Use check_alignment_argument.
958 * opts.h (parse_and_check_align_values): Declare.
959 * toplev.c (init_alignments): Remove.
960 (read_log_maxskip): New.
961 (parse_N_M): Likewise.
962 (parse_alignment_opts): Likewise.
963 (backend_init_target): Remove usage of init_alignments.
964 * toplev.h (parse_alignment_opts): Declare.
965 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
967 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
968 * tree.c (cl_option_hasher::equal): New.
969 * varasm.c: Use new global macros.
971 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
973 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
974 Use a simpler align directive also if MAXSKIP = ALIGN-1.
975 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
976 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
977 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
978 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
979 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
980 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
981 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
982 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
983 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
984 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
985 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
987 2018-07-04 Martin Liska <mliska@suse.cz>
988 Jonathan Wakely <jwakely@redhat.com>
990 * coverage.c: Use correct type.
991 * doc/invoke.texi: Language correction.
993 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
996 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
997 ENDBRANCH for non-tail call which may return via indirect branch.
998 * doc/extend.texi: Document indirect_return attribute.
1000 2018-07-03 Martin Sebor <msebor@redhat.com>
1002 PR tree-optimization/86274
1003 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1005 (format_floating): Correct handling of infinities and NaNs.
1007 2018-07-03 Martin Sebor <msebor@redhat.com>
1009 * print-tree.c (print_real_cst): New function.
1010 (print_node_brief): Call it.
1011 (print_node): Ditto.
1013 2018-07-03 Jeff Law <law@redhat.com>
1015 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1016 into a single pattern.
1018 * config/h8300/h8300.md (ors code_iterator): New.
1019 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1020 a single pattern and single splitter.
1021 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1022 (iorqi3_1, xorqi3_1): Likewise.
1023 (iorqi3, xorqi3 expanders): Similarly.
1025 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1026 (movmd_internal) into a single pattern using the P mode iterator.
1027 (movmd splitters): Similarly.
1028 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1029 (movsd splitters): Similarly.
1031 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1032 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1033 right table based on the size of the operand.
1034 * config/h8300/h8300.md (length_table): Corresponding changes. All
1035 references to "addb", "addw" and "addl" changed to "add".
1036 (btst patterns): Merge two variants into a single pattern.
1037 (tstqi, tsthi): Likewise.
1038 (addhi3_incdec, addsi3_incdec): Likewise.
1039 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1040 (mulhi3, mulsi3): Likewise.
1041 (udivhi3, udivsi3): Likewise.
1042 (divhi3, divsi3): Likewise.
1043 (andorqi3, andorhi3, andorsi3): Likewise.
1045 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1048 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1049 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1051 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1053 PR tree-optimization/85694
1054 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1055 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1057 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1058 (<u>avg<mode>3_ceil): New patterns.
1060 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1062 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1063 scan-tree-dump directive.
1064 * gcc.dg/vect/slp-perm-2.c: Likewise.
1065 * gcc.dg/vect/slp-perm-3.c: Likewise.
1066 * gcc.dg/vect/slp-perm-5.c: Likewise.
1067 * gcc.dg/vect/slp-perm-6.c: Likewise.
1068 * gcc.dg/vect/slp-perm-7.c: Likewise.
1069 * gcc.dg/vect/slp-perm-8.c: Likewise.
1071 2018-07-03 Marek Polacek <polacek@redhat.com>
1074 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1076 2018-07-03 Richard Biener <rguenther@suse.de>
1079 * tree-ssa-structalias.c (find_func_clobbers): Properly
1080 handle indirect calls.
1082 2018-07-03 Jeff Law <law@redhat.com>
1084 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1085 (shifts): New code iterator.
1086 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1087 expander. Fix HImode handling on H8/SX.
1088 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1089 (subqi3, subhi3, subsi3 expanders): Likewise.
1090 (andqi3, andhi3, andsi3 expanders): Likewise.
1091 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1092 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1093 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1094 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1095 (zero_extendqihi2, zero_extendqisi2): Likewise.
1096 (extendqihi2, extendqisi2): Likewise.
1097 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1098 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1099 (rotlqi3_1, rotlhi3_1): Likewise.
1100 (logicalhi3_sn, logicalsi3_sn): Likewise.
1101 (logicalhi3, logicalsi3): Likewise.
1103 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1105 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1106 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1107 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1108 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1109 type to append_pattern_def_seq instead of creating a stmt_vec_info
1111 (build_mask_conversion): Likewise. Remove vinfo argument.
1112 (vect_add_conversion_to_patterm): Likewise, renaming to...
1113 (vect_add_conversion_to_pattern): ...this.
1114 (vect_recog_mask_conversion_pattern): Update call to
1115 build_mask_conversion. Pass the vector type to
1116 append_pattern_def_seq here too.
1117 (vect_recog_gather_scatter_pattern): Update call to
1118 vect_add_conversion_to_pattern.
1120 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1122 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1123 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1124 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1125 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1126 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1127 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1128 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1129 append_pattern_def_seq instead of new_pattern_def_seq.
1130 (vect_recog_divmod_pattern): Do both of the above.
1131 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1134 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1136 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1137 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1138 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1139 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1140 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1141 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1142 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1143 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1144 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1145 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1146 parameter with a single stmt_vec_info.
1147 (vect_recog_func_ptr): Likewise.
1148 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1149 (vect_try_gather_scatter_pattern): ...this.
1150 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1151 the stmt_vec_info of the statement to be matched. Don't clear
1152 STMT_VINFO_RELATED_STMT.
1153 (vect_pattern_recog): Update call accordingly.
1155 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1157 PR tree-optimization/85694
1158 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1159 (uavgM3_ceil): Document new optabs.
1160 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1161 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1163 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1164 (savg_ceil_optab): New optabs.
1165 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1166 (vect_vect_recog_func_ptrs): Add it.
1167 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1168 constant directly from the associated lhs.
1170 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1172 * tree-vect-patterns.c (vect_split_statement): New function.
1173 (vect_convert_input): Use it to try to split an existing cast.
1175 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1177 * poly-int.h (print_hex): New function.
1178 * dumpfile.h (dump_dec, dump_hex): Declare.
1179 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1180 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1181 min_input_precision, operation_precision and operation_sign.
1182 * tree-vect-patterns.c (vect_get_range_info): New function.
1183 (vect_same_loop_or_bb_p, vect_single_imm_use)
1184 (vect_operation_fits_smaller_type): Delete.
1185 (vect_look_through_possible_promotion): Add an optional
1186 single_use_p parameter.
1187 (vect_recog_over_widening_pattern): Rewrite to use new
1188 stmt_vec_info infomration. Handle one operation at a time.
1189 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1190 (vect_truncatable_operation_p, vect_set_operation_type)
1191 (vect_set_min_input_precision): New functions.
1192 (vect_determine_min_output_precision_1): Likewise.
1193 (vect_determine_min_output_precision): Likewise.
1194 (vect_determine_precisions_from_range): Likewise.
1195 (vect_determine_precisions_from_users): Likewise.
1196 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1197 (vect_vect_recog_func_ptrs): Put over_widening first.
1199 (vect_pattern_recog): Call vect_determine_precisions.
1201 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1203 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1204 statements that have been replaced by further pattern statements.
1205 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1207 2018-07-03 Richard Biener <rguenther@suse.de>
1209 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1210 always set *dt. Dump vectype in vectype overload.
1211 * dumpfile.h (dump_gimple_expr): New function.
1212 (dump_gimple_expr_loc): Likewise.
1213 * dumpfile.c (dump_gimple_expr): New function.
1214 (dump_gimple_expr_loc): Likewise.
1216 2018-07-02 Jeff Law <law@redhat.com>
1218 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1219 the H8/300, H8/300H and H8/S variants into a single pattern.
1220 (movhi_h8300, movqi_h8300hs): Similarly.
1221 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1222 (QHI mode iterator): New.
1224 * config/h8300/h8300.md: Remove trailing whitespace.
1226 2018-07-02 Jim Wilson <jimw@sifive.com>
1228 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1229 instead of emit_insn for interrupt returns.
1230 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1231 (riscv_sret, riscv_uret): Likewise.
1233 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1235 * pretty-print.c (selftest::test_pp_format): Move save and restore
1236 of quotes to class auto_fix_quotes, and add an instance.
1237 * selftest.c: Include "intl.h".
1238 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1239 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1240 * selftest.h (selftest::auto_fix_quotes): New class.
1242 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1244 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1245 (aarch64_sve_prepare_conditional_op): Remove.
1246 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1247 Allow aarch64_simd_reg_or_zero as select operand; remove
1248 the aarch64_sve_prepare_conditional_op call.
1249 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1250 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1251 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1252 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1253 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1254 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1255 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1256 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1257 and a splitters to match all of the *_any patterns.
1258 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1260 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1261 (SVE_COND_FP_BINARY_REV): Remove.
1262 (sve_int_op_rev, sve_fp_op_rev): New.
1263 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1264 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1265 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1266 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1267 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1268 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1269 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1270 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1271 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1273 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1274 Remove match_dup 1 from the inner unspec.
1275 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1277 * config/aarch64/aarch64.md (movprfx): New attr.
1278 (length): Default movprfx to 8.
1279 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1280 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1281 (*<su>mul<SVE_I>3_highpart): Likewise.
1282 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1283 (*v<ASHIFT><SVE_I>3): Likewise.
1284 (*<su><MAXMIN><SVE_I>3): Likewise.
1285 (*<su><MAXMIN><SVE_F>3): Likewise.
1286 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1287 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1288 (*div<SVE_F>4): Likewise.
1290 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1292 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1295 2018-07-02 Richard Biener <rguenther@suse.de>
1297 PR tree-optimization/86363
1298 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1299 memset argument refers to a non-variable address.
1301 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1303 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1304 {PLUS,MINUS}_EXPR code to...
1305 (adjust_symbolic_bound): ...here,
1306 (combine_bound): ...here,
1307 (set_value_range_with_overflow): ...and here.
1309 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1311 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1313 (extract_range_from_abs_expr): ...here.
1315 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1317 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1318 -fno-omit-frame-pointer when not optimizing.
1320 2018-07-02 Martin Liska <mliska@suse.cz>
1323 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1324 (propagate_nothrow): Likewise.
1326 2018-07-02 Martin Liska <mliska@suse.cz>
1329 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1331 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1333 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1334 function in r262149, changing "loc" param from source_location to
1335 const dump_location_t &.
1336 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1337 declaration, as above.
1339 2018-07-01 Paul Koning <ni1d@arrl.net>
1341 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1342 -munit-asm, -mgnu-asm, -mdec-asm.
1343 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1344 (pdp11_output_labelref): New.
1345 (pdp11_output_def): New.
1346 (pdp11_output_addr_vec_elt): New.
1347 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1348 %# and %@ format codes.
1349 (pdp11_option_override): New.
1350 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1351 (pdp11_output_ident): New.
1352 (pdp11_asm_named_section): New.
1353 (pdp11_asm_init_sections): New.
1354 (pdp11_file_start): New.
1355 (pdp11_file_end): New.
1356 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1357 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1358 %o, like %c but octal.
1359 (pdp11_option_override): New.
1360 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1362 (DATA_SECTION_ASM_OP): Ditto.
1363 (READONLY_DATA_SECTION_ASM_OP): New.
1364 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1365 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1366 (ASM_OUTPUT_LABELREF): Ditto.
1367 (ASM_OUTPUT_DEF): Ditto.
1368 (ASM_OUTPUT_EXTERNAL): New.
1369 (ASM_OUTPUT_SOURCE_FILENAME): New.
1370 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1371 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1372 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1373 %# and %@ format codes.
1374 * config/pdp11/pdp11.opt (mgnu-asm): New.
1375 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1376 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1377 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1379 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1381 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1382 dereferencing path[] beyond its length.
1384 (debug_all_paths): New.
1385 (rewire_first_differing_edge): New.
1386 (adjust_paths_after_duplication): New.
1387 (duplicate_thread_path): Call adjust_paths_after_duplication.
1389 (thread_through_all_blocks): Add new argument to
1390 duplicate_thread_path.
1392 2018-06-30 Jim Wilson <jimw@sifive.com>
1394 * config/riscv/predicates.md (p2m1_shift_operand): New.
1395 (high_mask_shift_operand): New.
1396 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1397 pattern using p2m1_shift_operand.
1398 (lshsi3_zero_extend_3+2): New combiner pattern using
1399 high_mask_shift_operand.
1401 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1403 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1405 (vect_recog_rotate_pattern): ...here.
1406 (vect_convert_input): Try to insert casts of invariants in the
1408 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1409 preheader to be empty.
1411 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1413 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1414 vector type. If given, install it in the new statement's
1416 (vect_element_precision): New function.
1417 (vect_unpromoted_value): New struct.
1418 (vect_unpromoted_value::vect_unpromoted_value): New function.
1419 (vect_unpromoted_value::set_op): Likewise.
1420 (vect_look_through_possible_promotion): Likewise.
1421 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1422 (vect_widened_op_tree, vect_convert_input): Likewise.
1423 (vect_convert_inputs, vect_convert_output): Likewise.
1424 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1425 to handle the optional cast of the multiplication result and
1426 vect_widened_op_tree to detect the widened multiplication itself.
1427 Do not require the input and output of promotion casts to have
1428 the same sign, but base the signedness of the operation on the
1429 input rather than the result. If the pattern includes two
1430 promotions, check that those promotions have the same sign.
1431 Do not restrict the MULT_EXPR handling to a double-width result;
1432 handle quadruple-width results and wider. Use vect_convert_inputs
1433 to convert the inputs to the common type.
1434 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1435 to handle the optional cast of the ABS result. Also allow a sign
1436 change or a sign extension between the ABS and MINUS.
1437 Use vect_widened_op_tree to detect the widened subtraction and use
1438 vect_convert_inputs to convert the inputs to the common type.
1439 (vect_handle_widen_op_by_const): Delete.
1440 (vect_recog_widen_op_pattern): New function.
1441 (vect_recog_widen_mult_pattern): Use it.
1442 (vect_recog_widen_shift_pattern): Likewise.
1443 (vect_recog_widen_sum_pattern): Use
1444 vect_look_through_possible_promotion to handle the promoted
1447 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1449 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1450 the containing gimple_seq *.
1451 * gimple-iterator.h (gsi_for_stmt): Declare it.
1452 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1453 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1454 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1455 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1456 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1458 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1460 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1461 statement being replaced is part of an existing pattern
1462 definition sequence, inserting the new pattern statements before
1464 (vect_pattern_recog_1): Don't return a bool. If the statement
1465 is already part of a pattern, instead apply pattern matching
1466 to the pattern definition statements. Don't clear the
1467 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1468 (vect_pattern_recog): Don't break after the first match;
1469 continue processing the pattern definition statements instead.
1470 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1472 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1474 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1475 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1476 (vect_recog_widen_sum_pattern): Use it.
1478 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1480 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1481 phi is not a pattern statement and has not been replaced by
1482 a pattern statement.
1483 * tree-vect-patterns.c (type_conversion_p): Don't check
1484 STMT_VINFO_IN_PATTERN_P.
1485 (vect_recog_vector_vector_shift_pattern): Likewise.
1486 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1487 the pattern statement rather than the original statement; check
1488 directly for a WIDEN_MULT_EXPR here.
1489 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1490 vect_is_simple_use to return the pattern statement rather
1491 than the original statement; use is_pattern_stmt_p to check
1492 for such a pattern statement.
1493 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1494 to return the pattern statement rather than the original statement;
1495 don't do the same transformation here.
1496 (vect_is_simple_use): If the defining statement has been replaced
1497 by a pattern statement, return the pattern statement instead.
1498 Remove the corresponding (local) transformation from the vectype
1501 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1503 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1504 end and default to null.
1505 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1506 (vectorizable_reduction): Update calls accordingly, dropping the
1507 gimple ** argument if the passed-back statement isn't needed.
1508 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1509 (vect_recog_rotate_pattern): Likewise.
1510 (vect_recog_mask_conversion_pattern): Likewise.
1511 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1512 (vect_mask_constant_operand_p): Likewise.
1513 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1514 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1515 (get_group_load_store_type, get_load_store_type): Likewise.
1516 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1517 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1518 (vectorizable_conversion, vectorizable_assignment): Likewise.
1519 (vectorizable_shift, vectorizable_operation): Likewise.
1520 (vectorizable_store, vect_is_simple_cond): Likewise.
1521 (vectorizable_condition, vectorizable_comparison): Likewise.
1522 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1523 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1524 and move it to the end. Cope with null def_stmt_outs.
1526 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1528 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1530 2018-06-29 Jeff Law <law@redhat.com>
1532 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1533 displacements for TARGET_V850E2V3 and newer.
1534 (TARGET_LRA_P): Remove. Defaults to LRA now.
1535 * config/v850/v850.md (sign23byte_load): Remove.
1536 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1537 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1539 2018-06-29 Martin Liska <mliska@suse.cz>
1542 * coverage.c (coverage_init): Mangle full path name.
1543 * doc/invoke.texi: Document the change.
1544 * gcov-io.c (mangle_path): New.
1545 * gcov-io.h (mangle_path): Likewise.
1546 * gcov.c (mangle_name): Use mangle_path for path mangling.
1548 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1550 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1551 if starting source register is not even.
1553 2018-06-29 Martin Liska <mliska@suse.cz>
1555 PR tree-optimization/86263
1556 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1557 Make edge redirection.
1559 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1561 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1562 (dump_scope_depth): New variable.
1563 (get_dump_scope_depth): New function.
1564 (dump_begin_scope): New function.
1565 (dump_end_scope): New function.
1566 * dumpfile.h (get_dump_scope_depth): New declaration.
1567 (dump_begin_scope): New declaration.
1568 (dump_end_scope): New declaration.
1569 (class auto_dump_scope): New class.
1570 (AUTO_DUMP_SCOPE): New macro.
1571 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1574 2018-06-29 Richard Biener <rguenther@suse.de>
1576 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1577 compute_all_dependences succeeds.
1578 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1579 exceed --param loop-max-datarefs-for-datadeps.
1581 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1583 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1585 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1588 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1589 alternative 0 in preferred_for_speed attribute.
1591 2018-06-28 Paul Koning <ni1d@arrl.net>
1593 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1594 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1595 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1596 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1597 for shift insn patterns.
1599 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1601 * cgraph.c (cgraph_node::get_body): Replace assignments to
1602 "dump_file" with calls to set_dump_file.
1603 * dumpfile.c (alt_dump_file): Make static, and group with...
1604 (alt_flags): ...this definition.
1605 (dumps_are_enabled): New variable.
1606 (refresh_dumps_are_enabled): New function.
1607 (set_dump_file): New function.
1608 (set_alt_dump_file): New function.
1609 (gcc::dump_manager::dump_start): Replace assignments to
1610 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1612 (gcc::dump_manager::dump_finish): Likewise.
1613 * dumpfile.h (alt_dump_file): Delete decl.
1614 (dumps_are_enabled): New variable decl.
1615 (set_dump_file): New function decl.
1616 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1618 * tree-nested.c (lower_nested_functions): Replace assignments to
1619 "dump_file" with calls to set_dump_file.
1621 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1623 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1624 goto_locus of each outgoing edge of each basic block.
1626 2018-06-28 Richard Biener <rguenther@suse.de>
1628 * dwarf2out.c (decl_scope_table): Remove.
1629 (push_decl_scope): Likewise.
1630 (pop_decl_scope): Likewise.
1631 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1632 (gen_struct_or_union_type_die): Likewise.
1633 (gen_tagged_type_die): Likewise.
1634 (dwarf2out_init): Do not initialize decl_scope_table.
1635 (dwarf2out_c_finalize): Do not free it.
1637 2018-06-28 Richard Biener <rguenther@suse.de>
1639 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1640 deciding whether to not re-use a DIE.
1642 2018-06-28 Richard Biener <rguenther@suse.de>
1644 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1645 DW_AT_abstract_origin attribute.
1647 2018-06-28 Martin Liska <mliska@suse.cz>
1649 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1650 Use newly introduced constants.
1651 * tree-switch-conversion.h (struct jump_table_cluster):
1652 Define max_ratio_for_size and max_ratio_for_speed.
1654 2018-06-28 Martin Liska <mliska@suse.cz>
1656 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1657 Add new checking assert to catch invalid state.
1658 (jump_table_cluster::can_be_handled): Handle single case
1660 (jump_table_cluster::is_beneficial): Bail out for such case.
1661 (bit_test_cluster::find_bit_tests):
1662 Add new checking assert to catch invalid state.
1663 (bit_test_cluster::can_be_handled): Handle single case
1665 (bit_test_cluster::is_beneficial): Bail out for such case.
1666 (switch_decision_tree::analyze_switch_statement):
1669 2018-06-28 Martin Liska <mliska@suse.cz>
1671 * common.opt: Introduce -completion option.
1672 * gcc.c (driver_handle_option): Handle it.
1673 (driver::main): Print completions if completion
1675 * opt-suggestions.c (option_proposer::get_completions):
1677 (option_proposer::suggest_completion): Likewise.
1678 (option_proposer::find_param_completions): Likewise.
1679 (verify_autocompletions): Likewise.
1680 (test_completion_valid_options): Likewise.
1681 (test_completion_valid_params): Likewise.
1682 (in_completion_p): Likewise.
1683 (empty_completion_p): Likewise.
1684 (test_completion_partial_match): Likewise.
1685 (test_completion_garbage): Likewise.
1686 (opt_proposer_c_tests): Likewise.
1687 * opt-suggestions.h: Declare new functions.
1688 * opts.c (common_handle_option): Handle OPT__completion_.
1689 * selftest-run-tests.c (selftest::run_tests): Add
1690 opt_proposer_c_tests.
1691 * selftest.c (assert_str_startswith): New.
1692 * selftest.h (assert_str_startswith): Likewise.
1693 (opt_proposer_c_tests): New.
1694 (ASSERT_STR_STARTSWITH): Likewise.
1696 2018-06-28 Martin Liska <mliska@suse.cz>
1698 * Makefile.in: Add opt-suggestions.o.
1699 * gcc-main.c: Include opt-suggestions.h.
1700 * gcc.c (driver::driver): Likewise.
1701 (driver::~driver): Remove m_option_suggestions.
1702 (driver::build_option_suggestions): Moved to option_proposer.
1703 (driver::suggest_option): Likewise.
1704 (driver::handle_unrecognized_options): Use option_proposer.
1705 * gcc.h (class driver): Add new memver m_option_proposer.
1706 * opt-suggestions.c: New file.
1707 * opt-suggestions.h: New file.
1709 2018-06-28 Martin Liska <mliska@suse.cz>
1711 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1712 (auto_string_vec::~auto_string_vec): Likewise.
1714 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1716 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1717 prevent_decl_creation_for_types fields up and add reset_location field.
1718 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1719 statement if id->reset_location is true.
1720 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1721 id->reset_location is true.
1722 (copy_phis_for_bb): Force input_location on the arguments if
1723 id->reset_location is true.
1724 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1725 is set on the function to be inlined.
1727 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1729 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1731 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1733 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1734 registers for Pmode.
1735 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1736 hard registers for the clobbered pseudo.
1738 2018-06-27 Paul Koning <ni1d@arrl.net>
1740 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1741 mutually exclusive options.
1742 * config/pdp11/constraints.md (h): New constraint.
1743 (O): Update definition to match shift code generation.
1744 (D): New constraint.
1745 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1747 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1749 (output_jump): Change arguments.
1750 (pdp11_fixed_cc_regs): New function.
1751 (pdp11_cc_mode): Ditto.
1752 (pdp11_expand_shift): Ditto.
1753 (pdp11_assemble_shift): Ditto.
1754 (pdp11_small_shift): Ditto.
1755 (pdp11_branch_cost): Remove.
1756 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1758 (pdp11_register_move_cost): Update for CC registers.
1759 (pdp11_rtx_costs): Add case for LSHIFTRT.
1760 (pdp11_output_jump): Add CCNZ mode conditional branches.
1761 (notice_update_cc_on_set): Remove.
1762 (pdp11_cc_mode): New function.
1763 (simple_memory_operand): Correct pre/post decrement case.
1764 (no_side_effect_operand): New function.
1765 (pdp11_regno_reg_class): Add CC_REGS class.
1766 (pdp11_fixed_cc_regs): New function.
1767 (pdp11_small_shift): New function.
1768 (pdp11_expand_shift): New function to expand shift insns.
1769 (pdp11_assemble_shift): New function to output shifts.
1770 (pdp11_branch_cost): Remove.
1771 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1772 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1773 (WCHAR_TYPE): Ditto.
1774 (PTRDIFF_TYPE): Ditto.
1775 (ADJUST_INSN_LENGTH): New macro.
1776 (FIXED_REGISTERS): Add CC registers.
1777 (CALL_USED_REGISTERS): Ditto.
1779 (REG_CLASS_NAMES): Ditto.
1780 (REG_CLASS_CONTENTS): Ditto.
1781 (SELECT_CC_MODE): Use new function.
1782 (TARGET_FLAGS_REGNUM): New macro.
1783 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1784 (cc0_reg_rtx): Remove.
1785 (CC_STATUS_MDEP): Remove.
1786 (CC_STATUS_MDEFP_INIT): Remove.
1787 (CC_IN_FPU): Remove.
1788 (NOTICE_UPDATE_CC): Remove.
1789 (REGISTER_NAMES): Add CC registers.
1790 (BRANCH_COST): Change to constant 1.
1791 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1793 * config/pdp11/pdp11.opt (mbcopy): Remove.
1794 (mbcopy-builtin): Remove.
1795 (mbranch-cheap): Remove.
1796 (mbranch-expensive): Remove.
1797 * config/pdp11/predicates.md (expand_shift_operand): Update to
1798 match shift code generation.
1799 (ccnz_operator): New predicate.
1800 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1801 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1802 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1804 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1805 description of O constraint.
1807 2018-06-27 Jeff Law <law@redhat.com>
1808 Austin Law <austinklaw@gmail.com>
1810 * config/v850/v850.md (addsi3_set_flags): New pattern.
1811 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1812 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1813 (zero_extendhisi2_v850_set_flags): Likewise.
1814 (zero_extendqisi2_v850_set_flags): Likewise.
1815 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1816 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1817 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1819 * config/v850/v850-protos.h (notice_update_cc): Remove.
1820 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1821 (v850_print_operand): Handle 'D' and "d".
1822 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1823 Add handling of arithmetic/logical operations compared against zero.
1824 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1825 Do not look at v850_compare_op, instead get mode from last argument.
1826 (v850_gen_compare): Remove
1827 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1828 after reload for prologue insns.
1829 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1831 (construct_save_jarl): Likewise.
1832 (TARGET_FLAGS_REGNUM): Define.
1833 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1834 (NOTICE_UPDATE_CC): Remove.
1835 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1836 than cc0. Conditionalize on reload_completed.
1837 (cmpsi_insn, setfcc_insn): Likewise.
1838 (tst1 splitter): Turn into define_and_split which sets the flags
1840 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1841 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1842 (cstoresf4, cstoredf4): Clobber the flags.
1843 (cmpsi, cmpsf, cmpdf): Remove expanders.
1844 (setf_insn): Remove pattern.
1845 (addsi3): Turn into define_and_split which clobbers the flags after
1846 reload and a suitable pattern (addsi3_clobber_flags) for use after
1848 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1849 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1850 (ashrsi3, ashrsi3_v850e2): Likewise.
1851 (bins): Clobber the flags.
1852 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1853 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1854 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1855 (call_value_internal_short, call_value_internal_long): Likewise.
1856 (callt_save_interrupt, callt_return_interrupt): Likewise.
1857 (save_interrupt, return_interrupt): Likewise.
1858 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1859 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1860 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1861 (All FP comparisons): Only allow after reload has completed.
1863 (divh, divhu): Tweak output template.
1864 (branch_z_normal, branch_z_invert): Remove
1865 (branch_nz_normal, branch_nz_invert): Likewise.
1866 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1868 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1869 * config/v850/v850.c (notice_update_cc): Remove.
1870 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1871 (CC_NO_CARRY): Likewise.
1872 (NOTICE_UPDATE_CC): Define to nothing.
1873 * config/v850/v850.md: Remove block comment on cc0 handling
1874 Remove "cc" attribute from all patterns. Remove cc_status handling
1875 from all patterns. Minor formatting fixes.
1877 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1879 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1880 (cortex-a76.cortex-a55): Likewise.
1881 * config/aarch64/aarch64-tune.md: Regenerate.
1882 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1883 cortex-a76.cortex-a55.
1885 2018-06-27 Jeff Law <law@redhat.com>
1887 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1888 (MULTILIB_DIRNAMES): Similarly.
1890 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1892 * gimple.h (gimple_return_retbnd): Delete.
1893 (gimple_return_set_retbnd): Likewise.
1894 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1895 gimple_return_set_retbnd.
1896 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1897 gimple_return_retbnd and adjust.
1898 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1899 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1900 Explicitly return NULL in a couple more cases. Move assertion
1901 on debug statements and remove unreachable code.
1902 (reset_debug_binding): Do not test id->retbnd.
1903 (expand_call_inline): Do not set it.
1905 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1907 * configure.ac: Add --disable-gcov option.
1908 * configure: Regenerate.
1909 * Makefile.in: Honour @enable_gcov@.
1910 * doc/install.texi: Document --disable-gcov.
1912 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1914 * config/arm/arm-cpus.in (cortex-a76): New entry.
1915 (cortex-a76.cortex-a55): Likewise.
1916 * config/arm/arm-tables.opt: Regenerate.
1917 * config/arm/arm-tune.md: Likewise.
1918 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1919 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1920 cortex-a76.cortex-a55.
1922 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1925 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1927 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1929 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1931 (EPILOGUE_USES): Likewise.
1933 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1935 * tree-inline.c (remap_location): New function extracted from...
1936 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1937 (copy_phis_for_bb): ...here. Call remap_location.
1938 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1940 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1942 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1943 unaligned vsx for 16B memset.
1945 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1948 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1949 ieee128_float_type_node to long_double_type_node unless
1950 TARGET_LONG_DOUBLE_128 is set.
1952 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1954 * cfgloop.c (get_loop_location): Convert return type from
1955 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1956 by implicit construction from rtx_insn *, and using
1957 dump_user_location_t::from_function_decl for the fallback case.
1958 * cfgloop.h (get_loop_location): Convert return type from
1959 location_t to dump_user_location_t.
1960 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1961 dump_printf_loc to pass in a dump_location_t rather than a
1962 location_t, via the gimple stmt.
1963 * coverage.c (get_coverage_counts): Update calls to
1964 dump_printf_loc to pass in dump_location_t rather than a
1966 * doc/optinfo.texi (Dump types): Convert example of
1967 dump_printf_loc from taking "locus" to taking "insn". Update
1968 description of the "_loc" calls to cover dump_location_t.
1969 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1971 (dump_user_location_t::dump_user_location_t): New constructors,
1972 from gimple *stmt and rtx_insn *.
1973 (dump_user_location_t::from_function_decl): New function.
1974 (dump_loc): Make static.
1975 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1976 const dump_location_t &.
1977 (dump_generic_expr_loc): Delete.
1978 (dump_printf_loc): Convert param "loc" from location_t to
1979 const dump_location_t &.
1980 (selftest::test_impl_location): New function.
1981 (selftest::dumpfile_c_tests): New function.
1982 * dumpfile.h: Include "profile-count.h".
1983 (class dump_user_location_t): New class.
1984 (struct dump_impl_location_t): New struct.
1985 (class dump_location_t): New class.
1986 (dump_printf_loc): Convert 2nd param from source_location to
1987 const dump_location_t &.
1988 (dump_generic_expr_loc): Delete.
1989 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1990 const dump_location_t &.
1991 * gimple-fold.c (fold_gimple_assign): Update call to
1992 dump_printf_loc to pass in a dump_location_t rather than a
1993 location_t, via the gimple stmt.
1994 (gimple_fold_call): Likewise.
1995 * gimple-loop-interchange.cc
1996 (loop_cand::analyze_iloop_reduction_var): Update for change to
1997 check_reduction_path.
1998 (tree_loop_interchange::interchange): Update for change to
2000 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2001 change in return-type of find_loop_location.
2002 (graphite_regenerate_ast_isl): Likewise.
2003 * graphite-optimize-isl.c (optimize_isl): Likewise.
2004 * graphite.c (graphite_transform_loops): Likewise.
2005 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2006 pass in a dump_location_t rather than a location_t, via the
2008 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2009 * ipa.c (walk_polymorphic_call_targets): Likewise.
2010 * loop-unroll.c (report_unroll): Convert "locus" param from
2011 location_t to dump_location_t.
2012 (decide_unrolling): Update for change to get_loop_location's
2014 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2015 location_t to dump_user_location_t.
2016 (grid_find_single_omp_among_assignments_1): Updates calls to
2017 dump_printf_loc to pass in a dump_location_t rather than a
2018 location_t, via the gimple stmt.
2019 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2020 location_t to dump_location_t. Updates calls to dump_printf_loc
2021 to pass in a dump_location_t rather than a location_t, via the
2023 (grid_inner_loop_gridifiable_p): Likewise.
2024 (grid_dist_follows_simple_pattern): Likewise.
2025 (grid_gfor_follows_tiling_pattern): Likewise.
2026 (grid_target_follows_gridifiable_pattern): Likewise.
2027 (grid_attempt_target_gridification): Convert initialization
2028 of local "grid" from memset to zero-initialization; FIXME: does
2029 this require C++11? Update call to dump_printf_loc to pass in a
2030 optinfo_location rather than a location_t, via the gimple stmt.
2031 * profile.c (read_profile_edge_counts): Updates call to
2032 dump_printf_loc to pass in a dump_location_t rather than a
2034 (compute_branch_probabilities): Likewise.
2035 * selftest-run-tests.c (selftest::run_tests): Call
2037 * selftest.h (dumpfile_c_tests): New decl.
2038 * tree-loop-distribution.c (pass_loop_distribution::execute):
2039 Update for change in return type of find_loop_location.
2040 * tree-parloops.c (parallelize_loops): Likewise.
2041 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2042 "locus" from location_t to dump_user_location_t.
2043 (canonicalize_loop_induction_variables): Likewise.
2044 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2045 for change in return type of find_loop_location.
2046 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2047 to dump_printf_loc to pass in a dump_location_t rather than a
2048 location_t, via the stmt.
2049 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2051 * tree-vect-loop-manip.c (find_loop_location): Convert return
2052 type from source_location to dump_user_location_t.
2053 (vect_do_peeling): Update for above change.
2054 (vect_loop_versioning): Update for change in type of
2056 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2057 from location_t to dump_user_location_t.
2058 (vect_estimate_min_profitable_iters): Update for change in type
2060 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2061 location_t to dump_location_t.
2062 (vect_slp_bb): Update for change in type of vect_location.
2063 * tree-vectorizer.c (vect_location): Convert from source_location
2064 to dump_user_location_t.
2065 (try_vectorize_loop_1): Update for change in vect_location's type.
2066 (vectorize_loops): Likewise.
2067 (increase_alignment): Likewise.
2068 * tree-vectorizer.h (vect_location): Convert from source_location
2069 to dump_user_location_t.
2070 (find_loop_location): Convert return type from source_location to
2071 dump_user_location_t.
2072 (check_reduction_path): Convert 1st param from location_t to
2073 dump_user_location_t.
2074 * value-prof.c (check_counter): Update call to dump_printf_loc to
2075 pass in a dump_user_location_t rather than a location_t; update
2076 call to error_at for change in type of "locus".
2077 (check_ic_target): Update call to dump_printf_loc to
2078 pass in a dump_user_location_t rather than a location_t, via the
2081 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2083 * config/s390/s390.h (enum processor_flags): Do not use
2084 default tune parameter when -march was specified.
2086 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2089 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2090 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2093 2018-06-26 Richard Biener <rguenther@suse.de>
2095 PR tree-optimization/86287
2097 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2098 (vect_analyze_loop): Initialize n_stmts.
2100 2018-06-26 Richard Biener <rguenther@suse.de>
2103 * fold-const.c (fold_convertible_p): Pointer extension
2106 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2109 * dwarf2out.c (loc_list_has_views): Adjust comments.
2110 (dw_loc_list): Split single cross-partition range with
2113 2018-06-25 Jeff Law <law@redhat.com>
2115 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2116 on -mbig-switch by default.
2118 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2120 (const_float_0_operand): Remove unused predicate.
2121 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2122 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2123 (recipsf2): New expander. Original pattern now called
2125 (recipdf2, recipdf2_insn): Similarly.
2126 (rsqrtsf2, rsqrtsf2_insn): Similarly
2127 (rsqrtdf2, rsqrtdf2_insn): Similarly
2129 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2131 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2132 Simplify logic for FreeBSD (twice).
2134 2018-06-25 Martin Sebor <msebor@redhat.com>
2136 PR tree-optimization/86204
2137 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2138 a strnlen result if it's less than the length of the string.
2140 2018-06-25 Martin Sebor <msebor@redhat.com>
2142 PR tree-optimization/85700
2143 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2144 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2145 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2147 2018-06-25 Martin Sebor <msebor@redhat.com>
2149 * doc/extend.texi (Zero-length arrays): Update and clarify.
2151 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2153 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2154 added IEEE/IBM long double multilib support on PowerPC little
2155 endian Linux systems.
2156 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2157 (MULTILIB_DEFAULTS): Likewise.
2158 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2160 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2161 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2162 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2164 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2167 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2168 (REORDER_45): Likewise.
2170 2018-06-25 Jeff Law <law@redhat.com>
2172 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2173 dividend to 32 bits. Adjust length.
2174 (udivmodhi4): Cleanup output template. Fix length.
2176 2018-06-25 Carl Love <cel@us.ibm.com>
2178 * config/rs6000/vsx.md: Change word selector to prefered location.
2180 2018-06-25 Richard Biener <rguenther@suse.de>
2182 PR tree-optimization/86304
2183 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2184 epilogue-if-converted loops as well.
2186 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2188 * lto-section-out.c (lto_begin_section): Do not print section
2189 name for noaddr and unnumbered dumps.
2191 2018-06-25 Richard Biener <rguenther@suse.de>
2193 * tree-vectorizer.h (struct vec_info_shared): New structure
2194 with parts split out from struct vec_info and loop_nest from
2195 struct _loop_vec_info.
2196 (struct vec_info): Adjust accordingly.
2197 (struct _loop_vec_info): Likewise.
2198 (LOOP_VINFO_LOOP_NEST): Adjust.
2199 (LOOP_VINFO_DATAREFS): Likewise.
2200 (LOOP_VINFO_DDRS): Likewise.
2201 (struct _bb_vec_info): Likewise.
2202 (BB_VINFO_DATAREFS): Likewise.
2203 (BB_VINFO_DDRS): Likewise.
2204 (struct _stmt_vec_info): Add dr_aux member.
2205 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2206 (DR_MISALIGNMENT_UNINITIALIZED): New.
2207 (set_dr_misalignment): Adjust.
2208 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2209 (vect_analyze_loop): Adjust prototype.
2210 (vect_analyze_loop_form): Likewise.
2211 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2212 Compute dependences lazily.
2213 (vect_record_base_alignments): Use shared datarefs/ddrs.
2214 (vect_verify_datarefs_alignment): Likewise.
2215 (vect_analyze_data_refs_alignment): Likewise.
2216 (vect_analyze_data_ref_accesses): Likewise.
2217 (vect_analyze_data_refs): Likewise.
2218 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2219 constructor parameter for shared part.
2220 (vect_analyze_loop_form): Pass in shared part and adjust.
2221 (vect_analyze_loop_2): Pass in storage for the number of
2222 stmts. Move loop nest finding to the caller. Compute
2224 (vect_analyze_loop): Pass in shared part.
2225 (vect_transform_loop): Verify shared datarefs are unchanged.
2226 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2227 constructor parameter for shared part.
2228 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2229 (vect_slp_bb): Verify shared datarefs are unchanged before
2231 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2233 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2234 DR_MISALIGNMENT_UNINITIALIZED.
2235 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2236 parameter for shared part.
2237 (vec_info::~vec_info): Adjust.
2238 (vec_info_shared::vec_info_shared): New.
2239 (vec_info_shared::~vec_info_shared): Likewise.
2240 (vec_info_shared::save_datarefs): Likewise.
2241 (vec_info_shared::check_datarefs): Likewise.
2242 (try_vectorize_loop_1): Construct shared part live for analyses
2243 of a single loop for multiple vector sizes.
2244 * tree-parloops.c (gather_scalar_reductions): Adjust.
2246 2018-06-25 Richard Biener <rguenther@suse.de>
2248 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2249 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2250 (vect_analyze_data_refs): Remove similar code from here and
2251 simplify accordingly.
2253 2018-06-25 Richard Biener <rguenther@suse.de>
2255 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2256 for reverse storage order accesses rather than asserting
2257 they cannot happen here.
2259 2018-06-25 Tom de Vries <tdevries@suse.de>
2262 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2263 Use data16 instead of .byte for insn prefix.
2265 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2268 * parser.c (make_char_string_pack): Pass this literal chars
2269 through cpp_interpret_string.
2270 (cp_parser_userdef_numeric_literal): Check the result of
2271 make_char_string_pack.
2273 2018-06-24 Maya Rashish <coypu@sdf.org>
2275 * ginclude/stddef.h: Simplify conditions around avoiding
2276 re-definition of __size_t.
2278 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2280 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2281 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2283 2018-06-22 Maya Rashish <coypu@sdf.org>
2285 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2286 made to OpenBSD and NetBSD through the years.
2288 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2290 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2291 behavior of vec_pack (vector double, vector double) to match
2292 behavior of vec_float2 (vector double, vector double).
2294 2018-06-22 Olivier Hainque <hainque@adacore.com>
2296 * gimplify.c (gimplify_function_tree): Prevent creation
2297 of a trampoline for the address of the current function
2298 passed to entry/exit instrumentation hooks.
2300 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2303 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2306 2018-06-22 Martin Liska <mliska@suse.cz>
2308 PR tree-optimization/86263
2309 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2310 Bail out if is_enabled is false.
2311 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2313 (jump_table_cluster::is_enabled): New function.
2315 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2317 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2318 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2319 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2320 (lto_input_ts_binfo_tree_pointers): Likewise.
2321 * tree-streamer-out.c (streamer_write_tree_bitfields,
2322 write_ts_binfo_tree_pointers): Likewise.
2323 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2325 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2327 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2329 2018-06-22 Martin Liska <mliska@suse.cz>
2331 * symbol-summary.h (get): Make it pure and inline move
2332 functionality from ::get function.
2333 (get): Remove and inline into ::get and ::get_create.
2334 (get_create): Move code from ::get function.
2336 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2339 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2340 -x assembler-with-cpp.
2342 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2344 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2345 _FILE_OFFSET_BITS=64 for C++.
2347 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2349 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2350 conversion insn that shows up when pr85657-3.c is compiled using
2351 IEEE 128-bit long double.
2352 (neg<mode>2_internal): Use the correct mode to check whether the
2353 mode is IBM extended.
2354 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2355 multiply and divide external functions from being created more
2358 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2360 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2362 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2363 the edge can be forwarded.
2364 (cfg_layout_merge_blocks): Likewise.
2366 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2368 * except.c (finish_eh_generation): Commit edge insertions only after
2369 the EH edges have been redirected from post-landing to landing pads.
2371 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2373 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2374 create_tmp_var_for to create the FRAME decl.
2375 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2377 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2379 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2380 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2381 debug statement when resetting its value.
2382 (expand_call_inline): Copy the locus of the call onto the assignment
2383 of the return value, if any. Use local variable in more cases.
2385 2018-06-21 Martin Liska <mliska@suse.cz>
2387 * ipa-pure-const.c (propagate_nothrow): Use
2388 funct_state_summaries->get.
2389 (dump_malloc_lattice): Likewise.
2390 (propagate_malloc): Likewise.
2392 2018-06-21 Richard Biener <rguenther@suse.de>
2394 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2395 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2396 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2398 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2399 BLOCK_ABSTRACT_ORIGIN unconditionally.
2401 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2403 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2405 * ipa-reference.c (ipa_reference_c_finalize): Delete
2406 ipa_ref_opt_sum_summaries and set it to NULL.
2408 2018-06-21 Tom de Vries <tdevries@suse.de>
2410 PR tree-optimization/85859
2411 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2412 test with comment from bb_no_side_effects_p.
2414 2018-06-21 Richard Biener <rguenther@suse.de>
2416 PR tree-optimization/86232
2417 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2418 max for constant niter.
2420 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2422 * config/aarch64/aarch64-simd.md
2423 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2425 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2427 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2428 Make opernads of the unspec commutative.
2430 2018-06-21 Richard Biener <rguenther@suse.de>
2432 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2433 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2434 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2435 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2437 (vect_record_base_alignments): Do not record base alignment
2438 for gathers or scatters.
2439 (vect_compute_data_ref_alignment): Drop return value that is always
2440 true. Bail out early for gathers or scatters.
2441 (vect_enhance_data_refs_alignment): Bail out early for gathers
2443 (vect_find_same_alignment_drs): Likewise.
2444 (vect_analyze_data_refs_alignment): Remove dead code.
2445 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2446 (vect_analyze_data_refs): For possible gathers or scatters do
2447 not create an alternate DR, just check their possible validity
2448 and mark them. Adjust DECL_NONALIASED handling to not rely
2450 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2451 update inits of gathers or scatters.
2452 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2453 Also copy gather/scatter flag to pattern vinfo.
2455 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2457 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2458 behavior of vec_packsu (vector unsigned long long, vector unsigned
2459 long long) to match behavior of vec_packs with same signature.
2461 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2462 Thomas Schwinge <thomas@codesourcery.com>
2463 Cesar Philippidis <cesar@codesourcery.com>
2465 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2466 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2467 (gimplify_adjust_omp_clauses): Likewise.
2468 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2469 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2470 (gimplify_omp_target_update): Update handling of acc update and
2472 * omp-low.c (install_var_field): Remove unused parameter
2473 base_pointers_restrict.
2474 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2475 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2477 (omp_target_base_pointers_restrict_p): Delete.
2478 (scan_omp_target): Update call to scan_sharing_clauses.
2479 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2481 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2482 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2483 (convert_local_omp_clauses): Likewise.
2484 * tree-pretty-print.c (dump_omp_clause): Likewise.
2485 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2487 (omp_clause_code_name): Likewise.
2489 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2492 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2495 PR tree-optimization/86231
2496 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2497 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2499 2018-06-20 Tom de Vries <tdevries@suse.de>
2501 PR tree-optimization/86097
2502 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2503 iv type if signedness of iv type is not the same as that of *nit.
2505 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2507 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2508 EDGE_EH edges, verify they are all EDGE_EH.
2510 2018-06-20 Maya Rashish <coypu@sdf.org>
2512 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2514 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2516 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2517 * config/aarch64/aarch64.c (xgene1_tunings): Add
2518 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2519 (aarch64_mode_valid_for_sched_fusion_p):
2520 Allow 16-byte modes.
2521 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2522 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2524 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2526 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2527 * config/aarch64/iterators.md (VQ2): New mode iterator.
2529 2018-06-20 Martin Liska <mliska@suse.cz>
2531 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2532 Change default ratio from 10 to 8.
2534 2018-06-20 Martin Liska <mliska@suse.cz>
2536 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2538 (bit_test_cluster::find_bit_tests): Likewise.
2539 (switch_decision_tree::analyze_switch_statement): Find clusters.
2540 * tree-switch-conversion.h (struct jump_table_cluster): Document
2543 2018-06-20 Martin Liska <mliska@suse.cz>
2545 * tree-switch-conversion.c (switch_conversion::collect):
2546 Record m_uniq property.
2547 (switch_conversion::expand): Bail out for special conditions.
2548 (group_cluster::~group_cluster): New.
2549 (group_cluster::group_cluster): Likewise.
2550 (group_cluster::dump): Likewise.
2551 (jump_table_cluster::emit): New.
2552 (switch_decision_tree::fix_phi_operands_for_edges): New.
2553 (struct case_node): Remove struct.
2554 (jump_table_cluster::can_be_handled): New.
2555 (case_values_threshold): Moved to header.
2556 (reset_out_edges_aux): Likewise.
2557 (jump_table_cluster::is_beneficial): New.
2558 (bit_test_cluster::can_be_handled): Likewise.
2559 (add_case_node): Remove.
2560 (bit_test_cluster::is_beneficial): New.
2561 (case_bit_test::cmp): New.
2562 (bit_test_cluster::emit): New.
2563 (expand_switch_as_decision_tree_p): Remove.
2564 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2565 (fix_phi_operands_for_edge): Likewise.
2566 (switch_decision_tree::analyze_switch_statement): New.
2567 (compute_cases_per_edge): Move ...
2568 (switch_decision_tree::compute_cases_per_edge): ... here.
2569 (try_switch_expansion): Likewise.
2570 (switch_decision_tree::try_switch_expansion): Likewise.
2571 (record_phi_operand_mapping): Likewise.
2572 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2573 (emit_case_decision_tree): Likewise.
2574 (switch_decision_tree::emit): Likewise.
2575 (balance_case_nodes): Likewise.
2576 (switch_decision_tree::balance_case_nodes): Likewise.
2577 (dump_case_nodes): Likewise.
2578 (switch_decision_tree::dump_case_nodes): Likewise.
2579 (emit_jump): Likewise.
2580 (switch_decision_tree::emit_jump): Likewise.
2581 (emit_cmp_and_jump_insns): Likewise.
2582 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2583 (emit_case_nodes): Likewise.
2584 (switch_decision_tree::emit_case_nodes): Likewise.
2585 (conditional_probability): Remove.
2586 * tree-switch-conversion.h (enum cluster_type): New.
2588 (struct cluster): Likewise.
2589 (cluster::cluster): Likewise.
2590 (struct simple_cluster): Likewise.
2591 (simple_cluster::simple_cluster): Likewise.
2592 (struct group_cluster): Likewise.
2593 (struct jump_table_cluster): Likewise.
2594 (struct bit_test_cluster): Likewise.
2595 (struct min_cluster_item): Likewise.
2596 (struct case_tree_node): Likewise.
2597 (case_tree_node::case_tree_node): Likewise.
2598 (jump_table_cluster::case_values_threshold): Likewise.
2599 (struct case_bit_test): Likewise.
2600 (struct switch_decision_tree): Likewise.
2601 (struct switch_conversion): Likewise.
2602 (switch_decision_tree::reset_out_edges_aux): Likewise.
2604 2018-06-20 Martin Liska <mliska@suse.cz>
2606 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2607 (hoist_edge_and_branch_if_true): Likewise.
2608 (expand_switch_using_bit_tests_p): Likewise.
2609 (struct case_bit_test): Likewise.
2610 (case_bit_test_cmp): Likewise.
2611 (emit_case_bit_tests): Likewise.
2612 (switch_conversion::switch_conversion): New class.
2613 (struct switch_conv_info): Remove old struct.
2614 (collect_switch_conv_info): More to ...
2615 (switch_conversion::collect): ... this.
2616 (check_range): Likewise.
2617 (switch_conversion::check_range): Likewise.
2618 (check_all_empty_except_final): Likewise.
2619 (switch_conversion::check_all_empty_except_final): Likewise.
2620 (check_final_bb): Likewise.
2621 (switch_conversion::check_final_bb): Likewise.
2622 (create_temp_arrays): Likewise.
2623 (switch_conversion::create_temp_arrays): Likewise.
2624 (free_temp_arrays): Likewise.
2625 (gather_default_values): Likewise.
2626 (switch_conversion::gather_default_values): Likewise.
2627 (build_constructors): Likewise.
2628 (switch_conversion::build_constructors): Likewise.
2629 (constructor_contains_same_values_p): Likewise.
2630 (switch_conversion::contains_same_values_p): Likewise.
2631 (array_value_type): Likewise.
2632 (switch_conversion::array_value_type): Likewise.
2633 (build_one_array): Likewise.
2634 (switch_conversion::build_one_array): Likewise.
2635 (build_arrays): Likewise.
2636 (switch_conversion::build_arrays): Likewise.
2637 (gen_def_assigns): Likewise.
2638 (switch_conversion::gen_def_assigns): Likewise.
2639 (prune_bbs): Likewise.
2640 (switch_conversion::prune_bbs): Likewise.
2641 (fix_phi_nodes): Likewise.
2642 (switch_conversion::fix_phi_nodes): Likewise.
2643 (gen_inbound_check): Likewise.
2644 (switch_conversion::gen_inbound_check): Likewise.
2645 (process_switch): Use the newly created class.
2646 (switch_conversion::expand): New.
2647 (switch_conversion::~switch_conversion): New.
2648 * tree-switch-conversion.h: New file.
2650 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2652 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2653 tree-vect-patterns.c.
2654 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2655 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2656 (vect_recog_sad_pattern): Likewise.
2657 (vect_recog_widen_sum_pattern): Likewise.
2658 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2659 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2660 (vect_recog_rotate_pattern): Likewise.
2661 (vect_recog_mult_pattern): Likewise.
2662 (vect_recog_vector_vector_shift_pattern): Likewise.
2663 (vect_recog_divmod_pattern): Likewise.
2664 (vect_recog_mixed_size_cond_pattern): Likewise.
2665 (vect_recog_bool_pattern): Likewise.
2666 (vect_recog_mask_conversion_pattern): Likewise.
2667 (vect_try_gather_scatter_pattern): Likewise.
2668 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2669 (vect_recog_over_widening_pattern): Likewise.
2670 (vect_recog_gather_scatter_pattern): Likewise.
2671 (vect_recog_func_ptr): Move from tree-vectorizer.h
2672 (vect_vect_recog_func_ptrs): Move further down the file.
2673 (vect_recog_func): Likewise. Remove the third argument.
2674 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2675 (vect_pattern_recog_1): Expect the pattern function to do any
2676 necessary target tests. Also expect it to provide a vector type.
2677 Remove the type_in handling.
2679 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2681 * tree-vect-patterns.c (vect_pattern_detected): New function.
2682 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2683 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2684 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2685 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2686 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2687 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2688 (vect_recog_mask_conversion_pattern)
2689 (vect_try_gather_scatter_pattern): Likewise.
2691 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2693 * tree-vect-patterns.c (vect_get_internal_def): New function.
2694 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2695 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2696 (search_type_for_mask_1): Use it.
2698 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2700 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2701 redundant WIDEN_SUM_EXPR handling.
2702 (vect_recog_sad_pattern): Likewise.
2704 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2706 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2707 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2708 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2709 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2710 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2712 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2714 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2715 use the stmt_vec_info of the original bb statement for the
2716 new zero assignment, even if the call is part of a pattern.
2718 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2720 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2721 that the sequence is attached to the original statement rather
2722 than the pattern statement.
2723 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2724 PATTERN_DEF_SEQ from the original statement rather than
2725 the main pattern statement.
2726 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2727 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2728 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2730 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2732 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2733 definition statements before the early exit for statements that aren't
2735 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2737 (vect_transform_loop): ...here. Process pattern definition
2738 statements without first checking whether the main pattern
2739 statement is live or relevant.
2741 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2743 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2744 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2746 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2748 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2749 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2750 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2751 (expand_block_compare): Change select_block_compare_mode call.
2752 (expand_strncmp_align_check): Use new functions, fix comment.
2753 (emit_final_str_compare_gpr): New function.
2754 (expand_strn_compare): Refactor and clean up code.
2755 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2757 2018-06-19 Tony Reix <tony.reix@atos.com>
2758 Damien Bergamini <damien.bergamini@atos.com>
2759 David Edelsohn <dje.gcc@gmail.com>
2761 * collect2.c (static_obj): New variable.
2762 (static_libs): New variable.
2763 (is_in_list): Uncomment declaration.
2764 (main): Track AIX libraries linked statically.
2765 (is_in_list): Uncomment definition.
2766 (scan_prog_file): Don't add AIX shared libraries initializer
2767 to constructor list if linking statically.
2769 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2771 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2773 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2775 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2777 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2780 2018-06-19 Martin Liska <mliska@suse.cz>
2782 * config/i386/i386.c (ix86_can_inline_p): Do not use
2783 ipa_fn_summaries::get_create.
2784 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2786 (devirtualization_time_bonus): Likewise.
2787 (ipcp_propagate_stage): Likewise.
2788 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2789 (edge_set_predicate): Likewise.
2790 (evaluate_conditions_for_known_args): Likewise.
2791 (evaluate_properties_for_edge): Likewise.
2792 (ipa_call_summary::reset): Tranform to ...
2793 (ipa_call_summary::~ipa_call_summary): ... this.
2794 (ipa_fn_summary::reset): Transform to ...
2795 (ipa_fn_summary::~ipa_fn_summary): ... this.
2796 (ipa_fn_summary_t::remove): Rename to ...
2797 (ipa_fn_summary_t::remove_callees): ... this.
2798 (ipa_fn_summary_t::duplicate): Use placement new
2799 instead of memory copy.
2800 (ipa_call_summary_t::duplicate): Likewise.
2801 (ipa_call_summary_t::remove): Remove.
2802 (dump_ipa_call_summary): Change get_create to get.
2803 (ipa_dump_fn_summary): Dump only when summary exists.
2804 (analyze_function_body): Use symbol_summary::get instead
2806 (compute_fn_summary): Likewise.
2807 (estimate_edge_devirt_benefit): Likewise.
2808 (estimate_edge_size_and_time): Likewise.
2809 (inline_update_callee_summaries): Likewise.
2810 (remap_edge_change_prob): Likewise.
2811 (remap_edge_summaries): Likewise.
2812 (ipa_merge_fn_summary_after_inlining): Likewise.
2813 (write_ipa_call_summary): Likewise.
2814 (ipa_fn_summary_write): Likewise.
2815 (ipa_free_fn_summary): Likewise.
2816 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2817 (struct ipa_call_summary): Likewise.
2818 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2820 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2821 (estimate_size_after_inlining): Likewise.
2822 (estimate_growth): Likewise.
2823 (growth_likely_positive): Likewise.
2824 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2825 (inline_call): Likewise.
2826 * ipa-inline.c (caller_growth_limits): Likewise.
2827 (can_inline_edge_p): Likewise.
2828 (can_inline_edge_by_limits_p): Likewise.
2829 (compute_uninlined_call_time): Likewise.
2830 (compute_inlined_call_time): Likewise.
2831 (want_inline_small_function_p): Likewise.
2832 (edge_badness): Likewise.
2833 (update_caller_keys): Likewise.
2834 (update_callee_keys): Likewise.
2835 (inline_small_functions): Likewise.
2836 (inline_to_all_callers_1): Likewise.
2837 (dump_overall_stats): Likewise.
2838 (early_inline_small_functions): Likewise.
2839 (early_inliner): Likewise.
2840 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2841 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2842 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2843 * ipa-split.c (execute_split_functions): Likewise.
2844 * symbol-summary.h: Likewise.
2845 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2847 2018-06-19 Richard Biener <rguenther@suse.de>
2849 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2850 (vectorize_loops): ... here. Fix dbgcnt handling.
2851 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2853 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2856 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2857 ieee128 argument takes up only one (vector) register, not two (floating
2860 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2862 * gimplify.c (gimplify_init_constructor): Really never clear for an
2863 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2865 2018-06-19 Richard Biener <rguenther@suse.de>
2867 PR tree-optimization/86179
2868 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2869 after failed recognition.
2871 2018-06-18 Martin Sebor <msebor@redhat.com>
2874 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2875 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2876 Handle integer subtraction.
2877 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2878 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2880 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2882 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2883 param from rtx to rtx_insn *.
2884 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2886 (frv_ifcvt_modify_insn): Likwise.
2887 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2888 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2889 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2890 as_a <rtx_insn *> cast to local "unprotected_region" once
2891 it's been established that it's not NULL or pc_rtx.
2892 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2893 param "sethi" from rtx to rtx_insn *.
2894 (nds32_group_float_insns): Likewise for param "insn".
2895 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2897 (vax_output_int_subtract): Likewise.
2898 * config/vax/vax.c (vax_output_int_add): Likewise for param
2900 (vax_output_int_subtract): Likewise.
2901 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2902 (emit_pattern_after): Likewise for param "after".
2903 (emit_insn_after): Likewise.
2904 (emit_jump_insn_after): Likewise.
2905 (emit_call_insn_after): Likewise.
2906 (emit_debug_insn_after): Likewise.
2907 (emit_pattern_before): Likewise for param "before".
2908 (emit_insn_before): Likewise.
2909 (emit_jump_insn_before): Likewise.
2910 * final.c (get_insn_template): Likewise for param "insn", removing
2912 * output.h (get_insn_template): Likewise for 2nd param.
2913 * rtl.h (emit_insn_before): Likewise.
2914 (emit_jump_insn_before): Likewise.
2915 (emit_debug_insn_before_noloc): Likewise.
2916 (emit_insn_after): Likewise.
2917 (emit_jump_insn_after): Likewise.
2918 (emit_call_insn_after): Likewise.
2919 (emit_debug_insn_after): Likewise.
2920 (set_insn_deleted): Likewise for param.
2922 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2925 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2926 floating point modes, so that IFmode is numerically greater than
2927 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2928 to declare the ordering. This prevents IFmode from being
2929 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2930 machine. Include rs6000-modes.h to share the fractional values
2931 between genmodes* and the rest of the compiler.
2935 * config/rs6000/rs6000-modes.h: New file.
2936 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2937 meaning of rs6000_long_double_size so that 126..128 selects an
2938 appropriate 128-bit floating point type.
2939 (rs6000_option_override_internal): Likewise.
2940 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2941 (TARGET_LONG_DOUBLE_128): Change the meaning of
2942 rs6000_long_double_size so that 126..128 selects an appropriate
2943 128-bit floating point type.
2944 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2945 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2946 source and destination to match the standard usage.
2947 (truncifkf2): Likewise.
2948 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2949 ISA 2.07 to use an explicit clobber, instead of passing in a
2951 (copysign<mode>3_soft): Likewise.
2953 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2955 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2956 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2957 (vect_slp_analyze_instance_dependence): Likewise.
2958 (vect_enhance_data_refs_alignment): Likewise.
2959 (vect_analyze_data_refs_alignment): Likewise.
2960 (vect_slp_analyze_and_verify_instance_alignment
2961 (vect_analyze_data_ref_accesses): Likewise.
2962 (vect_prune_runtime_alias_test_list): Likewise.
2963 (vect_analyze_data_refs): Likewise.
2964 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2965 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2966 (vect_analyze_scalar_cycles_1): Likewise.
2967 (vect_get_loop_niters): Likewise.
2968 (vect_analyze_loop_form_1): Likewise.
2969 (vect_update_vf_for_slp): Likewise.
2970 (vect_analyze_loop_operations): Likewise.
2971 (vect_analyze_loop): Likewise.
2972 (vectorizable_induction): Likewise.
2973 (vect_transform_loop): Likewise.
2974 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2975 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2976 (vect_make_slp_decision): Likewise.
2977 (vect_detect_hybrid_slp): Likewise.
2978 (vect_slp_analyze_operations): Likewise.
2979 (vect_slp_bb): Likewise.
2980 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2981 (vectorizable_bswap): Likewise.
2982 (vectorizable_call): Likewise.
2983 (vectorizable_simd_clone_call): Likewise.
2984 (vectorizable_conversion): Likewise.
2985 (vectorizable_assignment): Likewise.
2986 (vectorizable_shift): Likewise.
2987 (vectorizable_operation): Likewise.
2988 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2990 2018-06-18 Martin Sebor <msebor@redhat.com>
2992 PR tree-optimization/81384
2993 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2994 * builtins.c (expand_builtin_strnlen): New function.
2995 (expand_builtin): Call it.
2996 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2997 * builtins.def (BUILT_IN_STRNLEN): New.
2998 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2999 Warn for bounds in excess of maximum object size.
3000 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3001 single-value ranges. Handle strnlen.
3002 (handle_builtin_strlen): Handle strnlen.
3003 (strlen_check_and_optimize_stmt): Same.
3004 * doc/extend.texi (Other Builtins): Document strnlen.
3006 2018-06-18 Maya Rashish <coypu@sdf.org>
3008 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3009 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3010 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3012 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3014 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3016 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3018 * tree.c (escaped_string::escape): Replace cast to char * by
3019 const_cast<char *> (unescaped).
3021 2018-06-18 Nick Clifton <nickc@redhat.com>
3024 * tree.c (escaped_string): New class. Converts an unescaped
3025 string into its escaped equivalent.
3026 (warn_deprecated_use): Use the new class to convert the
3027 deprecation message, if present.
3028 (test_escaped_strings): New self test.
3029 (test_c_tests): Add test_escaped_strings.
3030 * doc/extend.texi (deprecated): Add a note that the
3031 deprecation message is affected by the -fmessage-length
3032 option, and that control characters will be escaped.
3033 (#pragma GCC error): Document this pragma.
3034 (#pragma GCC warning): Likewise.
3035 * doc/invoke.texi (-fmessage-length): Document this option's
3036 effect on the #warning and #error preprocessor directives and
3037 the deprecated attribute.
3039 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3041 * tree.c (decl_value_expr_lookup): Revert latest change.
3042 (decl_value_expr_insert): Likewise.
3044 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3046 * gimplify.c (nonlocal_vlas): Delete.
3047 (nonlocal_vla_vars): Likewise.
3048 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3050 (gimplify_body): Do not create and destroy nonlocal_vlas.
3051 * tree-nested.c: Include diagnostic.h.
3052 (use_pointer_in_frame): Tweak.
3053 (lookup_field_for_decl): Add assertion and declare the transformation.
3054 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3055 internal error when the reference is in a wrong context. Do not
3056 create a debug decl by default.
3057 (note_nonlocal_block_vlas): Delete.
3058 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3059 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3060 create a debug decl by default.
3061 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3062 call is in a wrong context.
3063 (fixup_vla_decls): New function.
3064 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3065 debug variables were created.
3066 * tree.c (decl_value_expr_lookup): Add checking assertion.
3067 (decl_value_expr_insert): Likewise.
3069 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3072 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3073 * tree-scalar-evolution.c (interpret_expr): Likewise.
3074 (expression_expensive_p): Likewise.
3075 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3076 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3077 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3078 (ssa_defined_by_minus_one_stmt_p): New.
3080 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3083 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3084 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3085 * dojump.c (do_jump): Likewise.
3086 * expr.c (expand_expr_real_2): Check operand type's sign.
3087 * fold-const.c (const_unop): Handle ABSU_EXPR.
3088 (fold_abs_const): Likewise.
3089 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3090 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3091 (strip_sign_op_1): Likesise.
3092 * match.pd: Add new pattern to generate ABSU_EXPR.
3093 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3094 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3095 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3096 * tree-inline.c (estimate_operator_cost): Likewise.
3097 * tree-pretty-print.c (dump_generic_node): Likewise.
3098 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3099 * tree.def (ABSU_EXPR): New.
3101 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3104 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3105 documented as preserved for backward compatibility only.
3106 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3108 PR rtl-optimization/86108
3109 * bb-reorder.c (create_forwarder_block): Renamed to ...
3110 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3111 jump from new landing pad to the second part.
3112 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3115 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3118 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3119 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3120 Only call store_expr for halves if the mode is the same.
3123 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3124 Fix up comment formatting.
3126 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3128 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3129 (typed_splay_tree::closure,
3130 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3131 (typed_splay_tree::typed_splay_tree,
3132 typed_splay_tree::operator =): Declared private.
3133 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3134 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3135 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3136 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3137 typed_splay_tree::splay_tree_splay,
3138 typed_splay_tree::splay_tree_foreach_helper,
3139 typed_splay_tree::splay_tree_insert,
3140 typed_splay_tree::splay_tree_remove,
3141 typed_splay_tree::splay_tree_lookup,
3142 typed_splay_tree::splay_tree_predecessor,
3143 typed_splay_tree::splay_tree_successor,
3144 typed_splay_tree::splay_tree_min,
3145 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3146 (typed_splay_tree::root, typed_splay_tree::comp,
3147 typed_splay_tree::delete_key,
3148 typed_splay_tree::delete_value): New data members.
3149 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3150 typed_splay_tree::remove.
3152 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3154 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3155 -mginv and -mno-ginv to the assembler.
3156 * config/mips/mips.opt (-mcrc): New option.
3158 * doc/invoke.text (-mcrc): Document.
3161 2018-06-15 Nick Clifton <nickc@redhat.com>
3164 * tree.c (escaped_string): New class. Converts an unescaped
3165 string into its escaped equivalent.
3166 (warn_deprecated_use): Use the new class to convert the
3167 deprecation message, if present.
3168 (test_escaped_strings): New self test.
3169 (test_c_tests): Add test_escaped_strings.
3170 * doc/extend.texi (deprecated): Add a note that the
3171 deprecation message is affected by the -fmessage-length
3172 option, and that control characters will be escaped.
3173 (#pragma GCC error): Document this pragma.
3174 (#pragma GCC warning): Likewise.
3175 * doc/invoke.texi (-fmessage-length): Document this option's
3176 effect on the #warning and #error preprocessor directives and
3177 the deprecated attribute.
3179 2018-06-15 Richard Biener <rguenther@suse.de>
3181 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3182 here, also noting vector size used.
3183 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3184 size used in MSG_OPTIMIZED_LOCATIONS dump.
3185 (pass_slp_vectorize::execute): Adjust.
3187 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3190 * config/arc/arc.c (arc_return_address_register): Fix
3193 2018-06-15 Richard Biener <rguenther@suse.de>
3196 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3197 leave useless conversion stripping to force_gimple_operand_gsi.
3198 (gimplify_build2): Likewise.
3199 (gimplify_build1): Likewise.
3201 2018-06-15 Richard Biener <rguenther@suse.de>
3204 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3205 before adjusting their block.
3207 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3209 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3210 multilibs for *-*-rtems*.
3211 * config/riscv/t-rtems: New file.
3213 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3216 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3217 unsigned_type_for returns NULL.
3220 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3221 subregs of multi-word pseudos unless the float mode has word size.
3223 2018-06-14 Richard Biener <rguenther@suse.de>
3226 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3227 duplicate and harmful type_hash_canon.
3228 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3230 2018-06-14 Richard Biener <rguenther@suse.de>
3233 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3236 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3238 * config/rtems.h (STDINT_LONG32): Define.
3240 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3241 Prachi Godbole <prachi.godbole@imgtec.com>
3243 * config/mips/mips-cpus.def: Define P6600.
3244 * config/mips/mips-tables.opt: Regenerate.
3245 * config/mips/mips.c (mips_ucbranch_type): New enum.
3246 (mips_rtx_cost_data): Add support for P6600.
3247 (mips_issue_rate): Likewise.
3248 (mips_multipass_dfa_lookahead): Likewise.
3249 (mips_avoid_hazard): Likewise.
3250 (mips_reorg_process_insns): Likewise.
3251 (mips_classify_branch_p6600): New function.
3252 * config/mips/mips.h (TUNE_P6600): New define.
3253 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3254 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3255 * config/mips/mips.md: Include p6600.md.
3256 (processor): Add p6600.
3257 * config/mips/p6600.md: New file.
3258 * doc/invoke.texi: Add p6600 to supported architectures.
3260 2018-06-13 Martin Sebor <msebor@redhat.com>
3262 PR tree-optimization/86114
3263 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3265 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3267 2018-06-13 Richard Biener <rguenther@suse.de>
3269 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3270 Properly set vector type of the intermediate stmt.
3271 * tree-vect-stmts.c (vectorizable_operation): The destination
3272 var always has vectype_out type.
3274 2018-06-13 Jeff Law <law@redhat.com>
3276 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3277 integer 0 for argument to print_rtl_with_bb.
3278 (rl78_reorg): Likewise.
3280 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3282 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3283 from rtx to rtx_insn *.
3284 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3286 (add_sched_insns_for_speculation): Likewise for local "target",
3287 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3288 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3289 from rtx_insn ** to rtx_code_label **.
3290 (reorg_emit_nops): Likewise.
3291 (c6x_reorg): Likewise for local "call_labels".
3292 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3294 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3295 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3296 the loops over LABEL_REFS.
3297 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3299 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3300 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3301 (split_branches): Strengthen local "olabel" from rtx to
3302 rtx_insn *, adding a safe_as_a cast.
3303 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3305 (add_insn_after): Likewise for first two params.
3306 (add_insn_before): Likewise.
3307 (remove_insn): Likewise for param.
3308 (emit_pattern_before_noloc): Likewise for second and third params.
3309 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3310 (emit_call_insn_before_noloc): Likewise.
3311 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3313 (emit_barrier_before): Likewise.
3314 (emit_label_before): Strengthen "label" param from "rtx" to
3315 "rtx_code_label *". Strengthen "before" param from "rtx" to
3317 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3319 (emit_pattern_after_noloc): Likewise.
3320 (emit_insn_after_noloc): Likewise.
3321 (emit_jump_insn_after_noloc): Likewise.
3322 (emit_call_insn_after_noloc): Likewise.
3323 (emit_debug_insn_after_noloc): Likewise.
3324 (emit_barrier_after): Likewise.
3325 (emit_label_after): Likewise for both params.
3326 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3327 "loc" param from "int" to "location_t".
3328 (emit_insn_after_setloc): Likewise.
3329 (emit_jump_insn_after_setloc): Likewise.
3330 (emit_call_insn_after_setloc): Likewise.
3331 (emit_debug_insn_after_setloc): Likewise.
3332 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3333 "loc" param from "int" to "location_t".
3334 (emit_pattern_before): Convert NULL_RTX to NULL.
3335 (emit_insn_before_setloc): Convert "loc" param from "int" to
3337 (emit_jump_insn_before_setloc): Likewise.
3338 (emit_call_insn_before_setloc): Likewise.
3339 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3340 rtx_insn *. Convert "loc" param from "int" to "location_t".
3341 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3342 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3343 Convert 3rd param from "int" to "location_t".
3344 (emit_barrier_before, emit_barrier_after, next_real_insn):
3345 Strengthen param from rtx to rtx_insn *.
3346 (emit_label_before): Strengthen 1st param from "rtx" to
3347 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3349 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3350 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3351 Strengthen 2nd param from "rtx" to "rtx_insn *".
3352 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3353 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3354 Likewise. Convert 3rd param from "int" to "location_t".
3355 (emit_label_after): Strengthen 1st param from "rtx" to
3357 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3359 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3360 from "rtx" to "rtx_insn *".
3362 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3364 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3365 bodies streamed in with -Q.
3366 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3367 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3368 * gimple-streamer-out.c: Include gimple-pretty-print.h
3369 (output_bb): Dump stmts streamed.
3370 * lto-section-out.c: Include print-tree.h
3371 (lto_begin_section): Dump sections created.
3372 (lto_output_decl_index): Dump decl encoded.
3373 * lto-streamer-out.c: Include print-tree.h
3374 (create_output_block): Dump output block created.
3375 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3376 (output_function): Dump function output.
3377 (output_constructor): Dump constructor streamed.
3378 (write_global_stream): Output indexes encoded.
3379 (produce_asm_for_decls): Dump streams encoded.
3380 * lto-streamer.c (streamer_dump_file): New global var.
3381 * lto-streamer.h (streamer_dump_file): Declare.
3382 * passes.c (ipa_write_summaries): Initialize streamer dump.
3383 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3386 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3389 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3390 offsets for register save directives. Emit a second batch of save
3391 directives, if need be, when the function accesses prior frames.
3393 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3395 * config/arc/fpu.md (fmasf4): Force operand to register.
3396 (fnmasf4): Likewise.
3398 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3400 * config/arc/arc-protos.h (arc_pad_return): Remove.
3401 * config/arc/arc.c (machine_function): Remove force_short_suffix
3403 (arc_print_operand): Adjust printing of '&'.
3404 (arc_verify_short): Remove conditional printing of short suffix.
3405 (arc_final_prescan_insn): Remove reference to size_reason.
3406 (pad_return): New function.
3407 (arc_reorg): Call pad_return.
3408 (arc_pad_return): Remove.
3409 (arc_init_machine_status): Remove reference to force_short_suffix.
3410 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3411 (attr length): When attribute iscompact is true force to 2
3412 regardless; in the case of maybe check if we want to force the
3413 instruction to have 4 bytes length.
3414 (nopv): Change it to generate 4 byte long nop as well.
3415 (blockage): New pattern.
3416 (simple_return): Remove call to arc_pad_return.
3417 (p_return_i): Likewise.
3419 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3421 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3423 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3425 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3428 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3430 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3431 for ARC700 and ARCv2.
3433 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3436 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3437 operands[2] instead of operands[1].
3440 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3442 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3443 case, check whether the outer register overlaps an unallocatable
3444 register, not just whether it fits the required class.
3446 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3448 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3450 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3451 (memrefs_conflict_p): Likewise.
3452 (init_alias_analysis): Likewise.
3453 * cfgexpand.c (expand_debug_expr): Likewise.
3454 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3455 * cse.c (fold_rtx): Likewise.
3456 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3457 * expr.c (emit_block_move_hints): Likewise.
3458 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3459 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3460 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3461 (emit_group_store): Likewise.
3462 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3463 to read the PRE/POST_MODIFY increment.
3464 * calls.c (store_one_arg): Use strip_offset.
3465 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3467 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3469 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3470 (simplify_binary_operation_1): Extend CONST_INT handling to
3472 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3473 than a HOST_WIDE_INT.
3474 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3476 (adjust_mems, add_stores): Update accodingly.
3477 (vt_canonicalize_addr): Track polynomial offsets.
3478 (emit_note_insn_var_location): Likewise.
3479 (vt_add_function_parameter): Likewise.
3480 (vt_initialize): Likewise.
3482 2018-06-12 Jeff Law <law@redhat.com>
3484 * config.gcc (alpha*-*-freebsd*): Remove.
3485 * config/alpha/freebsd.h: Remove.
3487 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3490 * spellcheck-tree.c (levenshtein_distance): Rename to...
3491 (get_edit_distance): ...this, and update for underlying renaming.
3492 * spellcheck-tree.h (levenshtein_distance): Rename to...
3493 (get_edit_distance): ...this.
3494 * spellcheck.c (levenshtein_distance): Rename to...
3495 (get_edit_distance): ...this. Convert from Levenshtein distance
3496 to Damerau-Levenshtein distance by supporting transpositions of
3497 adjacent characters. Rename "v1" to "v_next" and "v0" to
3499 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3500 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3501 update for underlying renaming.
3502 (selftest::levenshtein_distance_unit_test): Rename to...
3503 (selftest::test_get_edit_distance_unit): ...this, and update for
3504 underlying renaming.
3505 (selftest::test_find_closest_string): Add example from PR 69968
3506 where transposition helps
3507 (selftest::test_metric_conditions): Update for renaming.
3508 (selftest::test_metric_conditions): Likewise.
3509 (selftest::spellcheck_c_tests): Likewise.
3510 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3511 (get_edit_distance): ...this.
3512 (best_match::consider): Update for renaming.
3514 2018-06-12 Martin Sebor <msebor@redhat.com>
3516 PR tree-optimization/85259
3517 * builtins.c (compute_objsize): Handle constant offsets.
3518 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3519 true iff a warning has been issued.
3520 * gimple.h (gimple_nonartificial_location): New function.
3521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3522 gimple_nonartificial_location and handle -Wno-system-headers.
3523 (handle_builtin_stxncpy): Same.
3525 2018-06-12 Martin Sebor <msebor@redhat.com>
3528 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3530 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3532 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3533 BUILTIN_VEC_XST entries for pointer to double and long long.
3535 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3538 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3540 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3543 2018-06-12 Martin Liska <mliska@suse.cz>
3545 * doc/options.texi: Document IntegerRange.
3547 2018-06-12 Martin Liska <mliska@suse.cz>
3549 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3550 * opt-functions.awk: Handle Deprecated flag.
3551 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3553 (read_cmdline_option): Report warning for a deprecated option.
3554 * opts.h (struct cl_option): Add new field cl_deprecated.
3555 (CL_ERR_DEPRECATED): New.
3557 2018-06-12 Martin Liska <mliska@suse.cz>
3559 * doc/options.texi: Document Deprecated option flag.
3561 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3563 * config/arc/arc-arch.h (arc_extras): New enum.
3564 (arc_cpu_t):Add field extra.
3565 (arc_cpu_types): Consider the extras.
3566 * config/arc/arc-cpus.def: Add extras info.
3567 * config/arc/arc-opts.h (processor_type): Consider extra field.
3568 * config/arc/arc.c (arc_override_options): Handle extra field.
3570 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3572 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3573 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3575 * config/arc/arc.c (arc_init): Update pic warning.
3576 (irq_range): Update irq range parsing warnings.
3577 (arc_override_options): Update various warning messages.
3578 (arc_handle_aux_attribute): Likewise.
3580 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3582 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3584 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3586 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3587 compiler messages directives.
3589 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3591 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3592 * config/mips/mips-tables.opt: Regenerate.
3593 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3595 * doc/invoke.texi: Document -march=i6500.
3597 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3599 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3600 (i6400_gpmul): Add cpu_unit.
3601 (i6400_gpdiv): Likewise.
3602 (i6400_msa_add_d): Update reservations.
3603 (i6400_msa_int_add) Likewise.
3604 (i6400_msa_short_logic3) Likewise.
3605 (i6400_msa_short_logic2) Likewise.
3606 (i6400_msa_short_logic) Likewise.
3607 (i6400_msa_move) Likewise.
3608 (i6400_msa_cmp) Likewise.
3609 (i6400_msa_short_float2) Likewise.
3610 (i6400_msa_div_d) Likewise.
3611 (i6400_msa_long_logic1) Likewise.
3612 (i6400_msa_long_logic2) Likewise.
3613 (i6400_msa_mult) Likewise.
3614 (i6400_msa_long_float2) Likewise.
3615 (i6400_msa_long_float4) Likewise.
3616 (i6400_msa_long_float5) Likewise.
3617 (i6400_msa_long_float8) Likewise.
3618 (i6400_fpu_fadd): Include frint type.
3619 (i6400_fpu_store): New define_insn_reservation.
3620 (i6400_fpu_load): Likewise.
3621 (i6400_fpu_move): Likewise.
3622 (i6400_fpu_fcmp): Likewise.
3623 (i6400_fpu_fmadd): Likewise.
3624 (i6400_int_mult): Include imul3nc type and update reservation.
3625 (i6400_int_div): Include idiv3 type and update reservation.
3626 (i6400_int_load): Update to check type not move_type.
3627 (i6400_int_store): Likewise.
3628 (i6400_int_prefetch): Set zero latency.
3630 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3632 * gcc.c: Document new %@{...} sequence.
3633 (LINK_COMMAND_SPEC): Use it for the -L switches.
3634 (cpp_unique_options): Use it for the -I switches.
3635 (at_file_argbuf): New global variable.
3636 (in_at_file): Likewise.
3637 (alloc_args): Create at_file_argbuf.
3638 (clear_args): Truncate at_file_argbuf.
3639 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3640 (open_at_file): New function.
3641 (close_at_file): Likewise.
3642 (create_at_file): Delete.
3643 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3646 (validate_switches_from_spec): Deal with %@{...} sequence.
3647 (validate_switches): Likewise.
3648 (driver::finalize): Call clear_args.
3650 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3652 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3654 2018-06-11 Martin Sebor <msebor@redhat.com>
3656 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3657 (Wno-class-memaccess): Add @opindex.
3658 (Wno-templates, Wno-multiple-inheritance): Same.
3659 (Wno-virtual-inheritance, Wno-namespaces): Same.
3660 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3661 (Wno-format-overflow, Wno-format-truncation): Same.
3662 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3663 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3664 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3665 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3666 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3667 (Wno-misspelled-isr): Same.
3669 2018-06-11 Martin Sebor <msebor@redhat.com>
3671 * PR tree-optimization/86083
3672 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3674 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3676 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3678 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3681 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3682 on the correct operand.
3683 (*movdi_internal64): Ditto.
3685 2018-06-11 Martin Liska <mliska@suse.cz>
3687 PR tree-optimization/86089
3688 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3690 2018-06-11 Julia Koval <julia.koval@intel.com>
3692 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3693 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3694 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3696 2018-06-11 Olivier Hainque <hainque@adacore.com>
3698 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3699 for Ada with strict dwarf2.
3701 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3704 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3707 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3709 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3711 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3713 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3716 2018-06-08 Martin Liska <mliska@suse.cz>
3718 * tree-cfg.h (debug_function): Fix argument type to match
3721 2018-06-08 Martin Liska <mliska@suse.cz>
3723 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3724 Remove usage of MPX-related (and removed) fields.
3725 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3727 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3729 * cfg.c (debug): Use TDF_NONE rather than 0.
3730 * cfghooks.c (debug): Likewise.
3731 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3732 (struct dump_option_value_info): Convert to...
3733 (struct kv_pair): ...this template type.
3734 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3736 (optinfo_verbosity_options): Likewise.
3737 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3739 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3740 than int for "optgroup_flags" param.
3741 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3743 (dump_dec): Likewise.
3744 (dump_finish): Use TDF_NONE rather than 0.
3745 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3746 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3747 than 0. Update for change to option_ptr.
3748 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3749 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3750 0. Update for changes to optinfo_verbosity_options and
3752 (opt_info_switch_p): Convert optgroup_flags from int to
3754 (dump_basic_block): Use dump_flags_t rather than int
3755 for "dump_kind" param.
3756 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3757 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3758 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3759 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3760 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3761 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3762 TDF_NONE): Convert from macros to...
3763 (enum dump_flag): ...this new enum.
3764 (dump_flags_t): Update to use enum.
3765 (operator|, operator&, operator~, operator|=, operator&=):
3766 Implement for dump_flags_t.
3767 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3768 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3769 Convert from macros to...
3770 (enum optgroup_flag): ...this new enum.
3771 (optgroup_flags_t): New typedef.
3772 (operator|, operator|=): Implement for optgroup_flags_t.
3773 (struct dump_file_info): Convert field "alt_flags" to
3774 dump_flags_t. Convert field "optgroup_flags" to
3776 (dump_basic_block): Use dump_flags_t rather than int for param.
3777 (dump_generic_expr_loc): Likewise.
3778 (dump_dec): Likewise.
3779 (dump_register): Convert param "optgroup_flags" to
3781 (opt_info_enable_passes): Likewise.
3782 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3784 * gimple-pretty-print.c (debug): Likewise.
3785 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3786 (merged_store_group::apply_stores): Likewise.
3787 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3788 * gimple.c (verify_gimple_pp): Likewise.
3789 * graphite-poly.c (print_pbb_body): Likewise.
3790 * passes.c (pass_manager::register_one_dump_file): Convert
3791 local "optgroup_flags" to optgroup_flags_t.
3792 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3794 (debug_body): Likewise.
3795 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3796 to optgroup_flags_t.
3797 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3799 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3800 (convert_mult_to_fma): Likewise.
3801 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3802 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3803 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3804 "dump_kind" to dump_flags_t.
3806 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3808 * config/rs6000/rs6000.c (min, max): Delete.
3810 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3812 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3813 -mabi=spe and -mabi=no-spe.
3815 2018-06-08 Martin Liska <mliska@suse.cz>
3817 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3818 where we expect an existing summary.
3820 2018-06-08 Martin Liska <mliska@suse.cz>
3822 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3823 * ipa-inline.h (estimate_edge_growth): Likewise.
3825 2018-06-08 Martin Liska <mliska@suse.cz>
3827 * cgraph.c (function_version_hasher::hash): Use
3828 cgraph_node::get_uid ().
3829 (function_version_hasher::equal):
3830 * cgraph.h (cgraph_node::get_uid): New method.
3831 * ipa-inline.c (update_caller_keys): Use
3832 cgraph_node::get_uid ().
3833 (update_callee_keys): Likewise.
3834 * ipa-utils.c (searchc): Likewise.
3835 (ipa_reduced_postorder): Likewise.
3836 * lto-cgraph.c (input_node): Likewise.
3837 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3838 * symbol-summary.h (symtab_insertion): Likewise.
3839 (symtab_removal): Likewise.
3840 (symtab_duplication): Likewise.
3841 * tree-pretty-print.c (dump_function_header): Likewise.
3842 * tree-sra.c (convert_callers_for_node): Likewise.
3844 2018-06-08 Martin Liska <mliska@suse.cz>
3846 * cgraph.c (symbol_table::create_edge): Always assign a new
3848 (symbol_table::free_edge): Do not recycle numbers.
3849 * cgraph.h (cgraph_edge::get): New method.
3850 * symbol-summary.h (symtab_removal): Use it.
3851 (symtab_duplication): Likewise.
3852 (call_summary::hashable_uid): Remove.
3854 2018-06-08 Martin Liska <mliska@suse.cz>
3856 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3857 (initialize_growth_caches): Remove.
3858 (free_growth_caches): Likewise.
3859 (do_estimate_edge_time): Use edge_growth_cache.
3860 (do_estimate_edge_size): Likewise.
3861 (do_estimate_edge_hints): Likewise.
3862 * ipa-inline.c (reset_edge_caches): Likewise.
3863 (recursive_inlining): Likewise.
3864 (inline_small_functions): Likewise.
3865 * ipa-inline.h (initialize_growth_caches): Remove.
3866 (estimate_edge_size): Likewise.
3867 (estimate_edge_time): Likewise.
3868 (estimate_edge_hints): Likewise.
3869 (reset_edge_growth_cache): Likewise.
3870 * symbol-summary.h (call_summary::remove): New method.
3872 2018-06-08 Martin Liska <mliska@suse.cz>
3874 * ipa-cp.c (class edge_clone_summary): New summary.
3875 (grow_edge_clone_vectors): Remove.
3876 (ipcp_edge_duplication_hook): Remove.
3877 (class edge_clone_summary_t): New call_summary class.
3878 (ipcp_edge_removal_hook): Remove.
3879 (edge_clone_summary_t::duplicate): New function.
3880 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3881 (create_specialized_node): Likewise.
3882 (ipcp_driver): Initialize edge_clone_summaries and do not
3885 2018-06-08 Martin Liska <mliska@suse.cz>
3887 * symbol-summary.h (get): New function.
3888 (call_summary::m_initialize_when_cloning): New class member.
3890 2018-06-08 Martin Liska <mliska@suse.cz>
3892 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3893 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3894 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3895 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3896 (remove_cgraph_node_from_order): Use the removed_nodes set.
3897 (do_per_function_toporder): Likwise.
3898 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3899 instead of summary_uid.
3900 (symtab_removal): Likewise.
3901 (symtab_duplication): Likewise.
3903 2018-06-08 Martin Liska <mliska@suse.cz>
3905 * ipa-cp.c (ipcp_store_bits_results): Use
3906 ipcp_transformation_sum.
3907 (ipcp_store_vr_results): Likewise.
3908 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3910 (ipcp_transformation_initialize): ... this.
3911 (ipa_set_node_agg_value_chain):
3912 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3913 (write_ipcp_transformation_info): Likewise.
3914 (read_ipcp_transformation_info): Likewise.
3915 (ipcp_update_bits): Likewise.
3916 (ipcp_update_vr): Likewise.
3917 (ipcp_transform_function): Likewise.
3918 * ipa-prop.h: Rename ipcp_transformation_summary to
3919 ipcp_transformation.
3920 (class ipcp_transformation_t): New function summary.
3921 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3922 (ipa_get_agg_replacements_for_node): Likewise.
3924 2018-06-08 Martin Liska <mliska@suse.cz>
3926 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3928 (class funct_state_summary_t): New function_summary class.
3929 (has_function_state): Remove.
3930 (get_function_state): Likewise.
3931 (set_function_state): Likewise.
3932 (add_new_function): Likewise.
3933 (funct_state_summary_t::insert): New function.
3934 (duplicate_node_data): Remove.
3935 (remove_node_data): Remove.
3936 (funct_state_summary_t::duplicate): New function.
3937 (register_hooks): Create new funct_state_summaries.
3938 (pure_const_generate_summary): Use it.
3939 (pure_const_write_summary): Likewise.
3940 (pure_const_read_summary): Likewise.
3941 (propagate_pure_const): Likewise.
3942 (propagate_nothrow): Likewise.
3943 (dump_malloc_lattice): Likewise.
3944 (propagate_malloc): Likewise.
3945 (execute): Do not register hooks, just remove summary
3947 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3950 2018-06-08 Martin Liska <mliska@suse.cz>
3952 * ipa-reference.c (remove_node_data): Remove.
3953 (duplicate_node_data): Likewise.
3954 (class ipa_ref_var_info_summary_t): New class.
3955 (class ipa_ref_opt_summary_t): Likewise.
3956 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3957 (get_reference_optimization_summary): Use
3958 ipa_ref_opt_sum_summaries.
3959 (set_reference_vars_info): Remove.
3960 (set_reference_optimization_summary): Likewise.
3961 (ipa_init): Create summaries.
3962 (init_function_info): Use function summary.
3963 (ipa_ref_opt_summary_t::duplicate): New function.
3964 (ipa_ref_opt_summary_t::remove): New function.
3965 (get_read_write_all_from_node): Fix GNU coding style.
3966 (propagate): Use function summary.
3967 (write_node_summary_p): Fix GNU coding style.
3968 (stream_out_bitmap): Likewise.
3969 (ipa_reference_read_optimization_summary): Use function summary.
3970 (ipa_reference_c_finalize): Do not release hooks.
3972 2018-06-08 Martin Liska <mliska@suse.cz>
3974 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3975 (analyze_function_body): Extract multiple calls of get_create.
3976 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3977 * ipa-inline.c (recursive_inlining): Use ::get method.
3978 * ipa-inline.h (estimate_edge_growth): Likewise.
3980 2018-06-08 Martin Liska <mliska@suse.cz>
3982 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3984 (hsa_function_summary::hsa_function_summary): Use the new enum
3986 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3987 * hsa-gen.c (hsa_get_host_function): Likewise.
3988 (get_brig_function_name): Likewise.
3989 * ipa-hsa.c (process_hsa_functions): Likewise.
3990 (ipa_hsa_write_summary): Likewise.
3991 * symbol-summary.h (symtab_duplication): Use ::get function/
3992 (get): New function.
3994 2018-06-08 Martin Liska <mliska@suse.cz>
3996 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3998 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3999 (hsa_register_kernel): Likewise.
4000 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4001 * hsa-gen.c (hsa_get_host_function): Likewise.
4002 (get_brig_function_name): Likewise.
4003 (generate_hsa): Likewise.
4004 (pass_gen_hsail::execute): Likewise.
4005 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4006 (devirtualization_time_bonus): Likewise.
4007 (ipcp_propagate_stage): Likewise.
4008 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4009 (edge_set_predicate): Likewise.
4010 (evaluate_conditions_for_known_args): Likewise.
4011 (evaluate_properties_for_edge): Likewise.
4012 (ipa_fn_summary::reset): Likewise.
4013 (ipa_fn_summary_t::duplicate): Likewise.
4014 (dump_ipa_call_summary): Likewise.
4015 (ipa_dump_fn_summary): Likewise.
4016 (analyze_function_body): Likewise.
4017 (compute_fn_summary): Likewise.
4018 (estimate_edge_devirt_benefit): Likewise.
4019 (estimate_edge_size_and_time): Likewise.
4020 (estimate_calls_size_and_time): Likewise.
4021 (estimate_node_size_and_time): Likewise.
4022 (inline_update_callee_summaries): Likewise.
4023 (remap_edge_change_prob): Likewise.
4024 (remap_edge_summaries): Likewise.
4025 (ipa_merge_fn_summary_after_inlining): Likewise.
4026 (ipa_update_overall_fn_summary): Likewise.
4027 (read_ipa_call_summary): Likewise.
4028 (inline_read_section): Likewise.
4029 (write_ipa_call_summary): Likewise.
4030 (ipa_fn_summary_write): Likewise.
4031 (ipa_free_fn_summary): Likewise.
4032 * ipa-hsa.c (process_hsa_functions): Likewise.
4033 (ipa_hsa_write_summary): Likewise.
4034 (ipa_hsa_read_section): Likewise.
4035 * ipa-icf.c (sem_function::merge): Likewise.
4036 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4037 (do_estimate_edge_time): Likewise.
4038 (estimate_size_after_inlining): Likewise.
4039 (estimate_growth): Likewise.
4040 (growth_likely_positive): Likewise.
4041 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4042 (inline_call): Likewise.
4043 * ipa-inline.c (caller_growth_limits): Likewise.
4044 (can_inline_edge_p): Likewise.
4045 (can_inline_edge_by_limits_p): Likewise.
4046 (compute_uninlined_call_time): Likewise.
4047 (compute_inlined_call_time): Likewise.
4048 (want_inline_small_function_p): Likewise.
4049 (edge_badness): Likewise.
4050 (update_caller_keys): Likewise.
4051 (update_callee_keys): Likewise.
4052 (recursive_inlining): Likewise.
4053 (inline_small_functions): Likewise.
4054 (inline_to_all_callers_1): Likewise.
4055 (dump_overall_stats): Likewise.
4056 (early_inline_small_functions): Likewise.
4057 (early_inliner): Likewise.
4058 * ipa-inline.h (estimate_edge_growth): Likewise.
4059 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4060 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4061 * ipa-prop.h (IPA_NODE_REF): Likewise.
4062 (IPA_EDGE_REF): Likewise.
4063 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4064 (propagate_malloc): Likewise.
4065 * ipa-split.c (execute_split_functions): Likewise.
4066 * symbol-summary.h: Rename get to get_create.
4068 (get_create): Likewise.
4069 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4071 2018-06-08 Martin Liska <mliska@suse.cz>
4073 * symbol-summary.h (release): Move definition out of class
4075 (symtab_removal): Likewise.
4076 (symtab_duplication): Likewise.
4078 2018-06-08 Martin Liska <mliska@suse.cz>
4080 * symbol-summary.h (function_summary): Move constructor
4081 implementation out of class declaration.
4082 (release): Likewise.
4083 (symtab_insertion): Likewise.
4084 (symtab_removal): Likewise.
4085 (symtab_duplication): Likewise.
4088 2018-06-08 Martin Liska <mliska@suse.cz>
4090 * Makefile.in: Remove support for MPX (macros, related functions,
4091 fields in cgraph_node, ...).
4092 * builtin-types.def (BT_BND): Likewise.
4093 (BT_FN_BND_CONST_PTR): Likewise.
4094 (BT_FN_CONST_PTR_BND): Likewise.
4095 (BT_FN_VOID_PTR_BND): Likewise.
4096 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4097 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4098 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4099 (expand_builtin_mempcpy_with_bounds): Likewise.
4100 (expand_builtin_memset_with_bounds): Likewise.
4101 (expand_builtin_memset_args): Likewise.
4102 (std_expand_builtin_va_start): Likewise.
4103 (expand_builtin): Likewise.
4104 (expand_builtin_with_bounds): Likewise.
4105 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4106 (DEF_LIB_BUILTIN_CHKP): Likewise.
4107 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4108 (DEF_CHKP_BUILTIN): Likewise.
4109 (BUILT_IN_MEMCPY): Likewise.
4110 (BUILT_IN_MEMMOVE): Likewise.
4111 (BUILT_IN_MEMPCPY): Likewise.
4112 (BUILT_IN_MEMSET): Likewise.
4113 (BUILT_IN_STPCPY): Likewise.
4114 (BUILT_IN_STRCAT): Likewise.
4115 (BUILT_IN_STRCHR): Likewise.
4116 (BUILT_IN_STRCPY): Likewise.
4117 (BUILT_IN_STRLEN): Likewise.
4118 (BUILT_IN_MEMCPY_CHK): Likewise.
4119 (BUILT_IN_MEMMOVE_CHK): Likewise.
4120 (BUILT_IN_MEMPCPY_CHK): Likewise.
4121 (BUILT_IN_MEMSET_CHK): Likewise.
4122 (BUILT_IN_STPCPY_CHK): Likewise.
4123 (BUILT_IN_STRCAT_CHK): Likewise.
4124 (BUILT_IN_STRCPY_CHK): Likewise.
4125 * calls.c (store_bounds): Likewise.
4126 (emit_call_1): Likewise.
4127 (special_function_p): Likewise.
4128 (maybe_warn_nonstring_arg): Likewise.
4129 (initialize_argument_information): Likewise.
4130 (finalize_must_preallocate): Likewise.
4131 (compute_argument_addresses): Likewise.
4132 (expand_call): Likewise.
4133 * cfgexpand.c (expand_call_stmt): Likewise.
4134 (expand_return): Likewise.
4135 (expand_gimple_stmt_1): Likewise.
4136 (pass_expand::execute): Likewise.
4137 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4138 (cgraph_node::remove): Likewise.
4139 (cgraph_node::dump): Likewise.
4140 (cgraph_node::verify_node): Likewise.
4141 * cgraph.h (chkp_function_instrumented_p): Likewise.
4142 (symtab_node::get_alias_target): Likewise.
4143 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4144 (cgraph_local_p): Likewise.
4145 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4146 (cgraph_edge::rebuild_references): Likewise.
4147 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4148 (walk_polymorphic_call_targets): Likewise.
4149 (cgraph_node::expand_thunk): Likewise.
4150 (symbol_table::output_weakrefs): Likewise.
4151 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4152 (ix86_handle_option): Likewise.
4153 * config/i386/constraints.md: Likewise.
4154 * config/i386/i386-builtin-types.def (BND): Likewise.
4158 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4159 (BDESC_FIRST): Likewise.
4161 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4162 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4163 * config/i386/i386.c (enum reg_class): Likewise.
4164 (ix86_target_string): Likewise.
4165 (ix86_option_override_internal): Likewise.
4166 (ix86_conditional_register_usage): Likewise.
4167 (ix86_valid_target_attribute_inner_p): Likewise.
4168 (ix86_set_indirect_branch_type): Likewise.
4169 (ix86_set_current_function): Likewise.
4170 (ix86_function_arg_regno_p): Likewise.
4171 (init_cumulative_args): Likewise.
4172 (ix86_function_arg_advance): Likewise.
4173 (ix86_function_arg): Likewise.
4174 (ix86_pass_by_reference): Likewise.
4175 (ix86_function_value_regno_p): Likewise.
4176 (ix86_function_value_1): Likewise.
4177 (ix86_function_value_bounds): Likewise.
4178 (ix86_return_in_memory): Likewise.
4179 (ix86_setup_incoming_vararg_bounds): Likewise.
4180 (ix86_va_start): Likewise.
4181 (indirect_thunk_need_prefix): Likewise.
4182 (print_reg): Likewise.
4183 (ix86_print_operand): Likewise.
4184 (ix86_expand_call): Likewise.
4185 (ix86_output_function_return): Likewise.
4186 (reg_encoded_number): Likewise.
4187 (BDESC_VERIFYS): Likewise.
4188 (ix86_init_mpx_builtins): Likewise.
4189 (ix86_init_builtins): Likewise.
4190 (ix86_emit_cmove): Likewise.
4191 (ix86_emit_move_max): Likewise.
4192 (ix86_expand_builtin): Likewise.
4193 (ix86_builtin_mpx_function): Likewise.
4194 (ix86_get_arg_address_for_bt): Likewise.
4195 (ix86_load_bounds): Likewise.
4196 (ix86_store_bounds): Likewise.
4197 (ix86_load_returned_bounds): Likewise.
4198 (ix86_store_returned_bounds): Likewise.
4199 (ix86_class_likely_spilled_p): Likewise.
4200 (ix86_hard_regno_mode_ok): Likewise.
4201 (x86_order_regs_for_local_alloc): Likewise.
4202 (ix86_mitigate_rop): Likewise.
4203 (ix86_bnd_prefixed_insn_p): Likewise.
4204 (ix86_mpx_bound_mode): Likewise.
4205 (ix86_make_bounds_constant): Likewise.
4206 (ix86_initialize_bounds): Likewise.
4207 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4208 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4209 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4210 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4211 (TARGET_CHKP_BOUND_MODE): Likewise.
4212 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4213 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4214 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4215 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4216 * config/i386/i386.h (TARGET_MPX): Likewise.
4217 (TARGET_MPX_P): Likewise.
4218 (VALID_BND_REG_MODE): Likewise.
4219 (FIRST_BND_REG): Likewise.
4220 (LAST_BND_REG): Likewise.
4221 (enum reg_class): Likewise.
4222 (BND_REG_P): Likewise.
4223 (BND_REGNO_P): Likewise.
4224 (BNDmode): Likewise.
4225 (ADJUST_INSN_LENGTH): Likewise.
4226 * config/i386/i386.md: Likewise.
4227 * config/i386/i386.opt: Likewise.
4228 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4229 (defined): Likewise.
4230 (LINK_MPX): Likewise.
4231 (MPX_SPEC): Likewise.
4232 (LIBMPX_SPEC): Likewise.
4233 (LIBMPXWRAPPERS_SPEC): Likewise.
4234 (CHKP_SPEC): Likewise.
4235 * config/i386/predicates.md: Likewise.
4236 * dbxout.c (dbxout_type): Likewise.
4237 * doc/extend.texi: Likewise.
4238 * doc/invoke.texi: Likewise.
4239 * doc/md.texi: Likewise.
4240 * doc/tm.texi: Likewise.
4241 * doc/tm.texi.in: Likewise.
4242 * dwarf2out.c (is_base_type): Likewise.
4243 (gen_formal_types_die): Likewise.
4244 (gen_subprogram_die): Likewise.
4245 (gen_type_die_with_usage): Likewise.
4246 (gen_decl_die): Likewise.
4247 (dwarf2out_late_global_decl): Likewise.
4248 * expr.c (expand_assignment): Likewise.
4249 (emit_storent_insn): Likewise.
4250 (store_expr_with_bounds): Likewise.
4251 (store_expr): Likewise.
4252 (expand_expr_real_1): Likewise.
4253 * expr.h (store_expr_with_bounds): Likewise.
4254 * function.c (use_register_for_decl): Likewise.
4255 (struct bounds_parm_data): Likewise.
4256 (assign_parms_augmented_arg_list): Likewise.
4257 (assign_parm_find_entry_rtl): Likewise.
4258 (assign_parm_is_stack_parm): Likewise.
4259 (assign_parm_load_bounds): Likewise.
4260 (assign_bounds): Likewise.
4261 (assign_parms): Likewise.
4262 (expand_function_start): Likewise.
4263 * gcc.c (CHKP_SPEC): Likewise.
4264 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4265 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4266 (wrestrict_dom_walker::check_call): Likewise.
4267 * gimple.c (gimple_build_call_from_tree): Likewise.
4268 * gimple.h (enum gf_mask): Likewise.
4269 (gimple_call_with_bounds_p): Likewise.
4270 (gimple_call_set_with_bounds): Likewise.
4271 * gimplify.c (gimplify_init_constructor): Likewise.
4272 * ipa-cp.c (initialize_node_lattices): Likewise.
4273 (propagate_constants_across_call): Likewise.
4274 (find_more_scalar_values_for_callers_subset): Likewise.
4275 * ipa-hsa.c (process_hsa_functions): Likewise.
4276 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4277 * ipa-icf.c (sem_function::merge): Likewise.
4278 * ipa-inline.c (early_inliner): Likewise.
4279 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4280 (warn_function_cold): Likewise.
4281 (propagate_pure_const): Likewise.
4282 * ipa-ref.h (enum GTY): Likewise.
4283 * ipa-split.c (find_retbnd): Likewise.
4284 (consider_split): Likewise.
4285 (split_function): Likewise.
4286 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4287 * ipa.c (walk_polymorphic_call_targets): Likewise.
4288 (symbol_table::remove_unreachable_nodes): Likewise.
4289 (process_references): Likewise.
4290 (cgraph_build_static_cdtor_1): Likewise.
4291 * lto-cgraph.c (lto_output_node): Likewise.
4292 (output_refs): Likewise.
4293 (compute_ltrans_boundary): Likewise.
4294 (input_overwrite_node): Likewise.
4295 (input_node): Likewise.
4296 (input_cgraph_1): Likewise.
4297 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4298 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4299 (class pass_chkp_instrumentation_passes): Likewise.
4300 (make_pass_chkp_instrumentation_passes): Likewise.
4301 * passes.def: Likewise.
4302 * rtl.h (struct GTY): Likewise.
4303 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4304 * stor-layout.c (layout_type): Likewise.
4305 * symtab.c: Likewise.
4306 * target.def: Likewise.
4307 * targhooks.c (default_chkp_bound_type): Likewise.
4308 (default_chkp_bound_mode): Likewise.
4309 (default_builtin_chkp_function): Likewise.
4310 (default_chkp_function_value_bounds): Likewise.
4311 (default_chkp_make_bounds_constant): Likewise.
4312 (default_chkp_initialize_bounds): Likewise.
4313 * targhooks.h (default_chkp_bound_type): Likewise.
4314 (default_chkp_bound_mode): Likewise.
4315 (default_builtin_chkp_function): Likewise.
4316 (default_chkp_function_value_bounds): Likewise.
4317 (default_chkp_make_bounds_constant): Likewise.
4318 (default_chkp_initialize_bounds): Likewise.
4319 * toplev.c (compile_file): Likewise.
4320 (process_options): Likewise.
4321 * tree-core.h (DEF_BUILTIN): Likewise.
4322 (DEF_BUILTIN_CHKP): Likewise.
4323 * tree-inline.c (declare_return_variable): Likewise.
4324 (remap_gimple_stmt): Likewise.
4325 (copy_bb): Likewise.
4326 (initialize_inlined_parameters): Likewise.
4327 (expand_call_inline): Likewise.
4328 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4329 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4330 (make_pass_ipa_chkp_produce_thunks): Likewise.
4331 (make_pass_chkp): Likewise.
4332 (make_pass_chkp_opt): Likewise.
4333 (make_pass_chkp_instrumentation_passes): Likewise.
4334 * tree-pretty-print.c (dump_generic_node): Likewise.
4335 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4336 * tree-ssa-dce.c (propagate_necessity): Likewise.
4337 (eliminate_unnecessary_stmts): Likewise.
4338 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4339 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4340 * tree-ssa-sccvn.h: Likewise.
4341 * tree-ssa-strlen.c (get_string_length): Likewise.
4342 (valid_builtin_call): Likewise.
4343 (adjust_last_stmt): Likewise.
4344 (handle_builtin_strchr): Likewise.
4345 (handle_builtin_strcpy): Likewise.
4346 (handle_builtin_stxncpy): Likewise.
4347 (handle_builtin_memcpy): Likewise.
4348 (handle_builtin_strcat): Likewise.
4349 (strlen_check_and_optimize_stmt): Likewise.
4350 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4351 * tree-streamer-in.c: Likewise.
4352 * tree-streamer.c (record_common_node): Likewise.
4353 * tree.c (tree_code_size): Likewise.
4354 (wide_int_to_tree_1): Likewise.
4355 (type_contains_placeholder_1): Likewise.
4356 (build_common_tree_nodes): Likewise.
4357 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4358 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4359 (POINTER_BOUNDS_P): Likewise.
4360 (BOUNDED_TYPE_P): Likewise.
4361 (BOUNDED_P): Likewise.
4362 (CALL_WITH_BOUNDS_P): Likewise.
4363 (pointer_bounds_type_node): Likewise.
4364 * value-prof.c (gimple_ic): Likewise.
4365 * var-tracking.c (vt_add_function_parameters): Likewise.
4366 * varasm.c (make_decl_rtl): Likewise.
4367 (assemble_start_function): Likewise.
4368 (output_constant): Likewise.
4369 (maybe_assemble_visibility): Likewise.
4370 * varpool.c (ctor_for_folding): Likewise.
4371 * chkp-builtins.def: Remove.
4372 * ipa-chkp.c: Remove.
4373 * ipa-chkp.h: Remove.
4374 * rtl-chkp.c: Remove.
4375 * rtl-chkp.h: Remove.
4376 * tree-chkp-opt.c: Remove.
4377 * tree-chkp.c: Remove.
4378 * tree-chkp.h: Remove.
4380 2018-06-07 Carl Love <cel@us.ibm.com>
4382 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4383 vextract_fp_from_shortl): Add BE support.
4385 2018-06-07 Paul Koning <ni1d@arrl.net>
4387 * compare-elim.c (try_merge_compare): Don't merge compare if
4388 address contains a side effect.
4389 (try_eliminate_compare): Likewise.
4391 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4393 * config.gcc: Support "tremont".
4394 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4395 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4397 * config/i386/i386.c (m_TREMONT): Define.
4398 (processor_target_table): Add "tremont".
4399 (PTA_TREMONT): Define.
4400 (ix86_lea_outperforms): Add TARGET_TREMONT.
4401 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4402 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4403 and M_INTEL_GOLDMONT_PLUS.
4404 (fold_builtin_cpu): Add "tremont".
4405 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4406 (ix86_option_override_internal): Add "tremont".
4407 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4408 (processor_type): Add PROCESSOR_TREMONT.
4409 * config/i386/x86-tune.def: Add m_TREMONT.
4410 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4412 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4414 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4415 symbol defined for msp430i* devices to be lower case.
4417 gcc/testsuite/gcc.target/msp430/
4418 * msp430i-device-symbol.c: New test.
4419 * msp430f-device-symbol.c: New test.
4420 * msp430.h: New test header file.
4422 2018-06-07 Richard Biener <rguenther@suse.de>
4424 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4425 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4426 Properly wrap signed arithmetic if overflow wraps.
4428 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4430 PR tree-optimization/69615
4431 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4432 of a cast from a same precision integral SSA_NAME in a bb dominated
4433 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4434 cast to utype if rhs2 has already a compatible type.
4436 2018-06-07 Richard Biener <rguenther@suse.de>
4438 PR tree-optimization/85935
4439 * graphite-scop-detection.c (find_params_in_bb): Analyze
4440 condition operands with respect to the correct loop. Assert
4441 the analysis doesn't fail.
4443 2018-06-04 Carl Love <cel@us.ibm.com>
4445 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4446 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4447 as it is slightly cheaper.
4448 (first_match_or_eos_index_<mode>):
4449 Calculate index using natural element order.
4450 (first_match_index_<mode>):
4451 Calculate index using natural element order.
4452 (first_match_or_eos_index_<mode>):
4453 Calculate index using natural order.
4454 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4455 for BE and LE modes.
4456 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4457 P9V_BUILTIN_VCLZLSBB_V16QI.
4458 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4461 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4463 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4464 indentation and line wrap for many prototypes. Add missing
4465 @smallexample directives around block of prototypes for vec_xl and
4468 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4470 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4471 track if we pass or return IEEE 128-bit floating point.
4472 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4473 C++ mangling that is compatible with GCC 8.1.
4474 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4475 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4476 floating point types.
4477 (rs6000_function_arg_advance_1): Likewise.
4478 (rs6000_mangle_type): Optionally generate mangled names that match
4479 what GCC 8.1 generated for IEEE 128-bit floating point types.
4480 (rs6000_globalize_decl_name): If we have an external function that
4481 passes or returns IEEE 128-bit types, generate a weak reference
4482 from the mangled name used in GCC 8.1 to the current mangled
4484 (rs6000_init_builtins): Make __ibm128 use the long double type if
4485 long double is IBM extended double. Make __float128 use the long
4486 double type if long double is IEEE 128-bit.
4489 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4490 macro for __ibm128 built-in functions.
4491 (PACK_IF): Add __ibm128 pack/unpack functions.
4492 (UNPACK_IF): Likewise.
4493 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4494 enable long double built-in functions if long double is IEEE
4495 128-bit floating point.
4496 (rs6000_invalid_builtin): Update long double built-in function
4498 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4499 functions, adjust the built-in function to use the long double
4500 built-in function if __ibm128 and long double are the same type.
4501 * doc/extend.texi (PowerPC builtins): Update documention for
4502 __builtin_{,un}pack_longdouble. Add documentation for
4503 __builtin_{,un}pack_ibm128.
4505 2018-06-06 Jim Wilson <jimw@sifive.com>
4507 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4508 (struct machine_function): New field interrupt_mode.
4509 (riscv_handle_type_attribute): New function. Add forward declaration.
4510 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4511 (riscv_expand_epilogue): Check interrupt_mode field.
4512 (riscv_set_current_function): Check interrupt attribute args and
4513 set interrupt_mode field.
4514 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4515 (riscv_sret, riscv_uret): New.
4516 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4517 new arguments to interrupt attribute.
4519 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4522 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4523 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4525 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4527 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4528 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4529 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4530 several redundant entries.
4532 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4534 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4535 type from "rtx" to "rtx_insn *".
4536 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4537 for local "call_insn", removing cast.
4538 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4540 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4542 PR tree-optimization/86066
4543 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4544 for BIT_INSERT_EXPR stores.
4546 2018-06-06 Richard Biener <rguenther@suse.de>
4548 PR tree-optimization/86062
4549 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4550 component refs ontop
4551 of to be offsetted base.
4553 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4555 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4556 to be static and remove check on interrupt attribute name.
4558 gcc/testsuite/gcc.target/msp430/
4559 * function-attributes-4.c: New test.
4560 * static-interrupts.c: New test.
4562 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4564 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4565 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4567 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4570 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4572 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4573 Remove second argument, change how error is called.
4574 (aarch64_layout_arg): Remove second argument from
4575 aarch64_err_no_fpadvsimd call.
4576 (aarch64_init_cumulative_args): Ditto.
4577 (aarch64_gimplify_va_arg_expr): Ditto.
4578 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4580 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4582 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4583 (*simple_return_indirect_internal<mode>): Rename from
4584 simple_return_indirect_internal. Use W mode iterator.
4585 (rstorssp): New expander.
4586 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4587 (clrssbsy): New expander.
4588 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4590 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4592 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4594 (cmse_check_pointed_object): Likewise.
4596 2018-06-05 Martin Liska <mliska@suse.cz>
4598 PR gcov-profile/47618
4599 * doc/invoke.texi: Document how -fprofile-dir format
4602 2018-06-05 Richard Biener <rguenther@suse.de>
4604 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4605 removal pretend DOM info isn't available so we do not update
4606 it and only remove edges, not dominated blocks. Actually free
4607 DOM info in case we removed something. Remove unreachable blocks.
4608 (mfb_keep_latches): Work with either DOM info or marked backedges.
4609 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4610 first. Mark backedges if DOM info isn't available.
4611 (Re-)compute DOM info after cleanup_control_flow_pre.
4613 2018-06-05 Richard Biener <rguenther@suse.de>
4615 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4616 (locus_discrim_hasher::hash): Adjust.
4617 (locus_discrim_hasher::equal): Likewise.
4618 (next_discriminator_for_locus): Work on line directly.
4619 (same_line_p): Pass in expanded locus1 as well.
4620 (assign_discriminators): Avoid redundant location expansions.
4622 2018-06-05 Richard Biener <rguenther@suse.de>
4624 PR tree-optimization/86046
4625 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4626 if required after clearing TREE_ADDRESSABLE.
4628 2018-06-05 Richard Biener <rguenther@suse.de>
4630 PR tree-optimization/86047
4631 * tree-ssa-loop.c (for_each_index): Glob handling of all
4632 decls and constants and really handle all of them.
4634 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4637 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4638 qualifier_void_pointer and qualifier_const_void_pointer.
4639 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4640 (arm_init_builtins): Handle the above.
4641 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4642 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4643 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4646 2018-06-05 Martin Liska <mliska@suse.cz>
4648 * auto-profile.c (read_autofdo_file): Do not use
4649 gcov_ctr_summary struct.
4650 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4651 * coverage.c (struct counts_entry): Likewise.
4652 (read_counts_file): Read just single summary entry.
4653 (get_coverage_counts): Use gcov_summary struct.
4654 * coverage.h (get_coverage_counts): Likewise.
4655 * gcov-dump.c (dump_working_sets): Likewise.
4656 (tag_summary): Dump just single summary.
4657 * gcov-io.c (gcov_write_summary): Write just histogram
4659 (gcov_read_summary): Read just single summary.
4660 (compute_working_sets): Use gcov_summary struct.
4661 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4662 of GCOV_COUNTERS_SUMMABLE.
4663 (GCOV_COUNTERS_SUMMABLE): Remove.
4664 (GCOV_FIRST_VALUE_COUNTER): Replace with
4665 GCOV_COUNTER_V_INTERVAL.
4666 (struct gcov_ctr_summary): Remove.
4667 (struct gcov_summary): Directly use fields of former
4669 (compute_working_sets): Use gcov_summary struct.
4670 * gcov.c (read_count_file): Do not use ctrs fields.
4671 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4673 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4675 * profile.c: Likewise.
4676 * profile.h: Likewise.
4678 2018-06-05 Martin Liska <mliska@suse.cz>
4680 PR gcov-profile/84846
4681 * gcov.c (output_lines): Print working directory only
4682 in intermediate format.
4684 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4686 * config/s390/s390-builtin-types.def: Add void function type.
4687 * config/s390/s390-builtins.def: Use the function type for the
4690 2018-06-04 Jim Wilson <jimw@sifive.com>
4692 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4694 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4695 and maybe_eh_return. Change regno to unsigned int. Use new args to
4696 handle EH_RETURN_DATA_REGNO registers properly.
4697 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4698 (riscv_expand_epilogue): Update comment. Change argument name and
4699 type. Update code to use new name and type. Pass new args to
4700 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4702 * config/riscv/riscv.md (NORMAL_RETURN): New.
4703 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4704 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4705 (eh_return): Call gen_eh_return_internal and emit barrier.
4706 (eh_return_internal): Call riscv_expand_epilogue.
4708 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4710 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4711 bit_insertion field and declare can_be_merged_into method.
4712 (merged_store_group::can_be_merged_into): New method.
4713 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4714 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4715 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4717 2018-06-04 Richard Biener <rguenther@suse.de>
4719 PR tree-optimization/85955
4720 * builtins.c (fold_builtin_sincos): Convert pointers to
4721 destination to appropriate type before dereferencing.
4723 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4725 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4727 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4729 * expr.c (expand_expr_real_1): Force the operand into memory if
4730 its TYPE_MODE is BLKmode and if there is no integer mode for
4731 the number of bits being extracted.
4733 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4737 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4738 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4740 2018-06-04 Richard Biener <rguenther@suse.de>
4742 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4743 (cleanup_tree_cfg_noloop): ... single caller. Do
4744 start_recording_case_labels later.
4746 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4748 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4749 to _IMMINTRIN_H_INCLUDED.
4750 * config/i386/pconfigintrin.h: Ditto.
4751 * config/i386/waitpkgintrin.h: Ditto.
4752 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4753 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4754 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4755 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4756 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4757 waitpkgintrin.h and cldemoteintrin.h.
4759 2018-06-04 Richard Biener <rguenther@suse.de>
4761 PR tree-optimization/86038
4762 * tracer.c (find_best_successor): Check probability for
4763 being initialized, bail out if not.
4765 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4768 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4769 of bits to ignore when comparing architectures.
4771 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4773 PR tree-optimization/69615
4774 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4775 maximum or minimum of the type, try to merge it also as if
4776 range1 is + [-, x - 1] or + [x + 1, -].
4779 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4781 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4783 PR tree-optimization/86034
4784 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4785 the unsigned bitfield type in a bit insertion sequence if it does not
4786 have a larger precision than the bitfield size.
4787 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4789 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4791 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4793 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4795 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4796 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4797 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4798 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4800 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4802 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4803 Disable -fdelete-null-pointer-checks for ELF toolchain.
4805 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4806 Kito Cheng <kito.cheng@gmail.com>
4808 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4809 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4810 (nds32*-*-*): Add glibc and uclibc conditions.
4811 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4812 (TARGET_EXCEPT_UNWIND_INFO): Define.
4813 * config/nds32/elf.h: New file.
4814 * config/nds32/linux.h: New file.
4815 * config/nds32/nds32-elf.opt: New file.
4816 * config/nds32/nds32-linux.opt: New file.
4817 * config/nds32/nds32-fp-as-gp.c
4818 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4819 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4821 (nds32_asm_file_end): Ditto.
4822 (nds32_print_operand): Ditto.
4823 (nds32_insert_attributes): Ditto.
4824 (nds32_init_libfuncs): New function.
4825 (TARGET_HAVE_TLS): Define.
4826 (TARGET_INIT_LIBFUNCS): Define.
4827 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4829 (TARGET_ELF): Apply different mcmodel setting.
4830 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4831 been migrated into elf.h and linux.h files.
4832 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4833 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4834 (mcmodel): The content has been migrated into nds32-elf.opt and
4835 nds32-linux.opt files.
4836 * config/nds32/t-elf: New file.
4837 * config/nds32/t-linux: New file.
4839 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4840 Shiva Chen <shiva0217@gmail.com>
4842 * config/nds32/constants.md (unspec_volatile_element): Add
4843 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4844 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4846 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4847 (make_pass_nds32_fp_as_gp): Declare.
4848 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4850 (nds32_asm_function_end_prologue): Remove unused asm output.
4851 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4852 (nds32_asm_file_start): Output necessary fp_as_gp information.
4853 (nds32_option_override): Adjust register usage.
4854 (nds32_expand_prologue): Consider fp_as_gp situation.
4855 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4856 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4859 (simple_return): Ditto.
4860 (omit_fp_begin): Output special directive for fp_as_gp.
4861 (omit_fp_end): Output special directive for fp_as_gp.
4862 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4863 mforbid-fp-as-gp): New options.
4865 2018-06-01 Mark Wielaard <mark@klomp.org>
4867 * dwarf2out.c (dwarf2out_finish): Remove generation of
4868 DW_AT_loclists_base.
4870 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4872 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4873 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4874 (struct merged_store_group): Add bit_insertion field.
4875 (dump_char_array): Use standard hexadecimal format.
4876 (merged_store_group::merged_store_group): Set bit_insertion to false.
4877 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4878 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4879 also print the mask in the dump file.
4880 (pass_store_merging::gate): Minor tweak.
4881 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4882 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4883 stores with INTEGER_CST stores.
4884 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4885 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4886 and use it throughout. Generate bit insertion sequences if need be.
4887 (pass_store_merging::process_store): Remove redundant condition.
4888 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4890 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4892 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4893 the 128-bit floating point types. Fix function comment.
4895 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4897 * config/aarch64/aarch64-simd.md
4898 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4900 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4903 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4905 PR tree-optimization/85989
4906 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4908 (backprop::intersect_uses): Check it when deciding whether this
4909 is a backedge reference.
4910 (backprop::process_block): Add each phi to m_visited_phis
4911 after visiting it, then clear it at the end.
4913 2018-06-01 Richard Biener <rguenther@suse.de>
4915 * tree-vectorizer.h (vect_dr_stmt): New function.
4916 (vect_get_load_cost): Adjust.
4917 (vect_get_store_cost): Likewise.
4918 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4919 Use vect_dr_stmt instead of DR_SMTT.
4920 (vect_record_base_alignments): Likewise.
4921 (vect_calculate_target_alignment): Likewise.
4922 (vect_compute_data_ref_alignment): Likewise and make static.
4923 (vect_update_misalignment_for_peel): Likewise.
4924 (vect_verify_datarefs_alignment): Likewise.
4925 (vector_alignment_reachable_p): Likewise.
4926 (vect_get_data_access_cost): Likewise. Pass down
4927 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4928 (vect_get_peeling_costs_all_drs): Likewise.
4929 (vect_peeling_hash_get_lowest_cost): Likewise.
4930 (vect_enhance_data_refs_alignment): Likewise.
4931 (vect_find_same_alignment_drs): Likewise.
4932 (vect_analyze_data_refs_alignment): Likewise.
4933 (vect_analyze_group_access_1): Likewise.
4934 (vect_analyze_group_access): Likewise.
4935 (vect_analyze_data_ref_access): Likewise.
4936 (vect_analyze_data_ref_accesses): Likewise.
4937 (vect_vfa_segment_size): Likewise.
4938 (vect_small_gap_p): Likewise.
4939 (vectorizable_with_step_bound_p): Likewise.
4940 (vect_prune_runtime_alias_test_list): Likewise.
4941 (vect_analyze_data_refs): Likewise.
4942 (vect_supportable_dr_alignment): Likewise.
4943 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4944 (vect_gen_prolog_loop_niters): Likewise.
4945 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4946 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4948 (vect_recog_mask_conversion_pattern): Likewise.
4949 (vect_try_gather_scatter_pattern): Likewise.
4950 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4951 to vect_get_store_cost.
4952 (vect_get_store_cost): Get stmt_info instead of DR.
4953 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4954 (vect_get_load_cost): Get stmt_info instead of DR.
4956 2018-06-01 Richard Biener <rguenther@suse.de>
4959 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4960 vars wrapped in MEM_REFs.
4962 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4964 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4965 Fix subreg tests so that we only return a choice between
4966 GENERAL_REGS and FP_REGS if the original classes included both.
4968 2018-06-01 Richard Biener <rguenther@suse.de>
4971 * tree-ssa-structalias.c (get_function_part_constraint):
4972 Handle NULL fi->decl.
4973 (find_func_aliases_for_call): Properly handle indirect
4974 fi from direct call.
4975 (find_func_clobbers): Likewise.
4976 (ipa_pta_execute): Likewise.
4977 (create_variable_info_for): For functions that are ifunc_resolver
4978 resolve to a varinfo that contains the result of the resolver call.
4979 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4982 2018-05-31 Michael Collison <michael.collison@arm.com>
4984 * config/aarch64/aarch64.md:
4985 (*fix_to_zero_extenddfdi2): New pattern.
4986 * gcc.target/aarch64/fix_extend1.c: New testcase.
4988 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4992 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4993 and BUILT_IN_STRNCMP_EQ.
4994 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4995 BUILT_IN_STRNCMP_EQ.
4996 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4997 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4998 (gimple_fold_builtin): Likewise.
4999 * tree-ssa-strlen.c (compute_string_length): New function.
5000 (determine_min_obsize): New function.
5001 (handle_builtin_string_cmp): New function to handle calls to
5002 string compare functions.
5003 (strlen_optimize_stmt): Add handling to builtin string compare
5005 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5006 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5007 * tree.c (build_common_builtin_nodes): Add new defines of
5008 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5010 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5013 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5014 functions with naked attribute.
5016 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5018 * config/i386/sse.md (avx_vec_concat<mode>):
5019 Substitute concat_tg_mode mode attribute with xtg_mode.
5020 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5021 (concat_tg_mode): Remove mode attribute.
5023 2018-05-31 Martin Sebor <msebor@redhat.com>
5026 * calls.c (alloc_max_size): Correct a logic error/typo.
5027 Treat excessive arguments as infinite. Warn for invalid arguments.
5028 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5030 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5033 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5034 and movx for Haswell.
5036 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5037 Cesar Philippidis <cesar@codesourcery.com>
5040 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5041 when emitting error on private/firstprivate reductions.
5042 * omp-low.c (lower_omp_target): Avoid reference-type processing
5043 on pointers for firstprivate clause.
5045 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5047 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5050 * config/aarch64/aarch64-simd.md
5051 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5052 (aarch64_ld1_x3_<mode>): Likewise
5053 (aarch64_st1x2<VALLDIF:mode>): Likewise
5054 (aarch64_st1_x2_<mode>): Likewise
5055 (aarch64_st1x3<VALLDIF:mode>): Likewise
5056 (aarch64_st1_x3_<mode>): Likewise
5057 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5058 (vld1_s8_x3): Likewise.
5059 (vld1_u16_x3): Likewise.
5060 (vld1_s16_x3): Likewise.
5061 (vld1_u32_x3): Likewise.
5062 (vld1_s32_x3): Likewise.
5063 (vld1_u64_x3): Likewise.
5064 (vld1_s64_x3): Likewise.
5065 (vld1_f16_x3): Likewise.
5066 (vld1_f32_x3): Likewise.
5067 (vld1_f64_x3): Likewise.
5068 (vld1_p8_x3): Likewise.
5069 (vld1_p16_x3): Likewise.
5070 (vld1_p64_x3): Likewise.
5071 (vld1q_u8_x3): Likewise.
5072 (vld1q_s8_x3): Likewise.
5073 (vld1q_u16_x3): Likewise.
5074 (vld1q_s16_x3): Likewise.
5075 (vld1q_u32_x3): Likewise.
5076 (vld1q_s32_x3): Likewise.
5077 (vld1q_u64_x3): Likewise.
5078 (vld1q_s64_x3): Likewise.
5079 (vld1q_f16_x3): Likewise.
5080 (vld1q_f32_x3): Likewise.
5081 (vld1q_f64_x3): Likewise.
5082 (vld1q_p8_x3): Likewise.
5083 (vld1q_p16_x3): Likewise.
5084 (vld1q_p64_x3): Likewise.
5085 (vst1_s64_x2): Likewise.
5086 (vst1_u64_x2): Likewise.
5087 (vst1_f64_x2): Likewise.
5088 (vst1_s8_x2): Likewise.
5089 (vst1_p8_x2): Likewise.
5090 (vst1_s16_x2): Likewise.
5091 (vst1_p16_x2): Likewise.
5092 (vst1_s32_x2): Likewise.
5093 (vst1_u8_x2): Likewise.
5094 (vst1_u16_x2): Likewise.
5095 (vst1_u32_x2): Likewise.
5096 (vst1_f16_x2): Likewise.
5097 (vst1_f32_x2): Likewise.
5098 (vst1_p64_x2): Likewise.
5099 (vst1q_s8_x2): Likewise.
5100 (vst1q_p8_x2): Likewise.
5101 (vst1q_s16_x2): Likewise.
5102 (vst1q_p16_x2): Likewise.
5103 (vst1q_s32_x2): Likewise.
5104 (vst1q_s64_x2): Likewise.
5105 (vst1q_u8_x2): Likewise.
5106 (vst1q_u16_x2): Likewise.
5107 (vst1q_u32_x2): Likewise.
5108 (vst1q_u64_x2): Likewise.
5109 (vst1q_f16_x2): Likewise.
5110 (vst1q_f32_x2): Likewise.
5111 (vst1q_f64_x2): Likewise.
5112 (vst1q_p64_x2): Likewise.
5113 (vst1_s64_x3): Likewise.
5114 (vst1_u64_x3): Likewise.
5115 (vst1_f64_x3): Likewise.
5116 (vst1_s8_x3): Likewise.
5117 (vst1_p8_x3): Likewise.
5118 (vst1_s16_x3): Likewise.
5119 (vst1_p16_x3): Likewise.
5120 (vst1_s32_x3): Likewise.
5121 (vst1_u8_x3): Likewise.
5122 (vst1_u16_x3): Likewise.
5123 (vst1_u32_x3): Likewise.
5124 (vst1_f16_x3): Likewise.
5125 (vst1_f32_x3): Likewise.
5126 (vst1_p64_x3): Likewise.
5127 (vst1q_s8_x3): Likewise.
5128 (vst1q_p8_x3): Likewise.
5129 (vst1q_s16_x3): Likewise.
5130 (vst1q_p16_x3): Likewise.
5131 (vst1q_s32_x3): Likewise.
5132 (vst1q_s64_x3): Likewise.
5133 (vst1q_u8_x3): Likewise.
5134 (vst1q_u16_x3): Likewise.
5135 (vst1q_u32_x3): Likewise.
5136 (vst1q_u64_x3): Likewise.
5137 (vst1q_f16_x3): Likewise.
5138 (vst1q_f32_x3): Likewise.
5139 (vst1q_f64_x3): Likewise.
5140 (vst1q_p64_x3): Likewise.
5142 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5144 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5145 user_label_prefix to name.
5147 * tree-core.h: Update comment about the format of NAME string
5148 passed to handler in attribute_spec.
5150 * config/msp430/msp430.md: Remove erroneous subreg expression from
5151 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5152 zero_extend{q,h}isi2.
5154 2018-05-30 Borislav Petkov <bp@suse.de>
5156 * doc/extend.texi: Document some architecture specific
5157 constraints and sort entries.
5159 2018-05-30 Martin Sebor <msebor@redhat.com>
5162 * builtins.c (expand_builtin_stpcpy_1): New function.
5163 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5164 only if the former succeeds.
5166 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5168 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5171 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5173 * doc/invoke.texi (-flinker-output): Document
5175 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5177 * passes.c (ipa_write_summaries): Only modify statements if body
5179 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5180 incrementally linking.
5181 (ipa_passes): Likewise.
5182 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5183 pass down resolution info.
5184 * common.opt (flag_incremental_link): Update info.
5185 * gcc.c (plugin specs): Turn flinker-output=* to
5186 -plugin-opt=-linker-output-known
5187 * toplev.c (compile_file): Also cut compilation when doing incremental
5189 * flag-types. (enum lto_partition_model): Add
5190 LTO_LINKER_OUTPUT_NOLTOREL.
5191 (invoke.texi): Add -flinker-output docs.
5192 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5193 link same way as WPA; do not stream in dead initializers.
5195 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5196 darf2out_register_external_decl): Support incremental link.
5198 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5200 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5202 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5204 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5205 it down to simple_object_copy_lto_debug_sections.
5206 (run_gcc): Determine incremental LTO link time and configure
5207 lto1 into non-wpa mode, disable renaming of debug sections.
5209 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5211 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5212 descriptions of various incorrectly documented functions.
5214 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5217 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5218 address check not strict.
5220 2018-05-30 Richard Biener <rguenther@suse.de>
5222 PR tree-optimization/85964
5223 * tracer.c (better_p): Drop initialized count check, we only
5224 call the function with initialized counts now.
5225 (find_best_successor): Do find a best edge if one
5226 has uninitialized count.
5227 (find_best_predecessor): Likewise. Do BB frequency check only
5228 if count is initialized.
5230 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5232 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5233 (aarch64_ldrstr_offset_compare): New.
5234 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5235 load/store orderings.
5236 (aarch64_gen_adjusted_ldpstp): Likewise.
5238 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5240 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5241 Check for subset of GENERAL_REGS and FP_REGS.
5242 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5245 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5247 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5248 and wi::to_poly_offset. Add the current offset and then check
5249 whether the sum fits, rather than using an unchecked addition of
5250 a checked term. Check for a shwi rather than a uhwi.
5251 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5252 (store_constructor): Use poly_int_tree_p.
5253 (expand_expr_real_1): Likewise.
5254 * function.c (assign_temp): Likewise.
5255 * fold-const.c (const_binop): Use poly_int_tree_p and
5257 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5259 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5260 to_poly_offset for MEM offsets.
5261 * ipa-icf.c (sem_variable::equals): Likewise.
5262 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5263 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5264 wi::to_poly_offset for BIT_FIELD_REF offsets.
5265 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5267 * var-tracking.c (emit_note_insn_var_location): Use
5268 tree_to_poly_uint64.
5270 2018-05-29 Jim Wilson <jimw@sifive.com>
5272 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5274 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5277 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5278 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5280 (sse4_1_round<mode>2): Use nonimmediate_operand
5281 for operand 1 predicate.
5283 2018-05-29 Martin Sebor <msebor@redhat.com>
5284 Richard Biener <rguenther@suse.de>
5287 * calls.c (get_size_range): Call determine_value_range instead
5288 of get_value_range..
5289 * tree-vrp.h (determine_value_range): Declared new function.
5290 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5292 2018-05-29 Richard Biener <rguenther@suse.de>
5294 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5295 sure to use non-pattern stmts for get_earlier_stmt arguments.
5296 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5297 called on pattern stmts.
5298 (get_later_stmt): Likewise.
5300 2018-05-29 Martin Liska <mliska@suse.cz>
5302 PR gcov-profile/85759
5303 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5306 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5308 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5310 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5314 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5315 VEC_PACK_FLOAT_EXPR): New tree codes.
5316 * tree-pretty-print.c (op_code_prio): Handle
5317 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5318 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5319 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5320 * tree-inline.c (estimate_operator_cost): Likewise.
5321 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5322 * fold-const.c (const_binop): Likewise.
5323 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5324 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5325 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5326 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5327 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5328 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5329 * expr.c (expand_expr_real_2): Likewise.
5330 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5331 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5332 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5334 * optabs.c (expand_widen_pattern_expr): For
5335 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5336 sign from result type rather than operand's type.
5337 (expand_binop_directly): For vec_packu_float_optab and
5338 vec_packs_float_optab allow result type to be different from operand's
5340 * optabs-tree.c (optab_for_tree_code): Handle
5341 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5342 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5343 * tree-vect-generic.c (expand_vector_operations_1): Handle
5344 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5345 VEC_PACK_FLOAT_EXPR.
5346 * tree-vect-stmts.c (supportable_widening_operation): Handle
5348 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5349 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5350 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5351 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5352 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5354 (vec_pack<floatprefix>_float_<mode>): New expander.
5355 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5357 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5358 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5359 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5360 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5361 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5363 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5364 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5365 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5366 VEC_PACK_FLOAT_EXPR): Document.
5368 2018-05-29 Richard Biener <rguenther@suse.de>
5370 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5372 (stmt_vec_info_vec): Make pointer.
5373 (init_stmt_vec_info_vec): Remove.
5374 (free_stmt_vec_info_vec): Likewise.
5375 (set_stmt_vec_info_vec): New function.
5376 (free_stmt_vec_infos): Likewise.
5377 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5378 (set_vinfo_for_stmt): Likewise.
5379 (get_earlier_stmt): Likewise.
5380 (get_later_stmt): Likewise.
5381 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5382 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5383 (vec_info::~vec_info): Free stmt_vec_infos.
5384 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5385 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5386 (pass_slp_vectorize::execute): Likewise.
5387 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5388 (free_stmt_vec_info_vec): Likewise.
5389 (set_stmt_vec_info_vec): New function.
5390 (free_stmt_vec_infos): Likewise.
5391 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5392 the global stmt_vec_info_vec.
5393 * tree-parloops.c (gather_scalar_reductions): Use
5394 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5397 2018-05-29 Richard Biener <rguenther@suse.de>
5399 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5401 2018-05-29 Martin Liska <mliska@suse.cz>
5402 David Malcolm <dmalcolm@redhat.com>
5404 * vec.c (test_reverse): New.
5405 (vec_c_tests): Add new test.
5406 * vec.h (vl_ptr>::reverse): New function.
5408 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5410 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5412 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5415 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5417 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5419 2018-05-28 Richard Biener <rguenther@suse.de>
5421 PR tree-optimization/85933
5422 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5423 look at stmts marked as vectorizable.
5425 2018-05-28 Richard Biener <rguenther@suse.de>
5427 PR tree-optimization/85934
5428 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5429 vector boolean check before scalar optimization.
5431 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5433 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5436 2018-05-28 Mark Wielaard <mark@klomp.org>
5438 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5439 if it is an expression containing a minus sign.
5441 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5443 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5445 2018-05-27 Paul Koning <ni1d@arrl.net>
5447 * config/pdp11/pdp11.md (truncsihi2): Remove.
5449 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5450 Chung-Ju Wu <jasonwucj@gmail.com>
5452 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5454 (unaligned_store_dw): Ditto.
5455 * config/nds32/nds32-memory-manipulation.c
5456 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5457 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5458 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5459 (emit_setmem_word_loop): Rename to ...
5460 (emit_setmem_doubleword_loop): ... this.
5461 (nds32_gen_dup_4_byte_to_word_value): New function.
5462 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5463 (nds32_expand_setmem_loop): Refine implementation.
5464 (nds32_expand_setmem_loop_v3m): Ditto.
5465 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5468 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5470 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5472 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5474 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5475 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5476 machine->attr_no_prologue_p.
5477 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5478 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5479 (nds32_expand_epilogue): Consider attr_naked_p.
5480 (nds32_expand_epilogue_v3pop): Likewise.
5481 (nds32_can_use_return_insn): Likewise.
5482 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5483 attr_no_prologue_p fields.
5484 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5486 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5489 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5491 * config/i386/sse.md
5492 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5494 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5496 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5498 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5500 (*<floatsuffix>floatv2div2sf2): Rename to ...
5501 (*float<floatunssuffix>v2div2sf2): ... this.
5502 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5503 (float<floatunssuffix>v2div2sf2_mask): ... this.
5504 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5505 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5506 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5508 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5510 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5512 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5514 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5516 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5518 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5519 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5520 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5521 gen_ufix_truncv8dfv8si2.
5522 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5523 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5524 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5525 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5526 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5527 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5528 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5529 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5531 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5535 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5537 2018-05-25 Jim Wilson <jimw@sifive.com>
5539 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5540 * config/riscv/riscv.c (struct machine_function): Add
5541 interrupt_handler_p and attribute_checked_p fields.
5542 (riscv_attribute_table): Add interrupt.
5543 (riscv_interrupt_type_p): New.
5544 (riscv_save_reg_p): Save extra regs for interrupt handler.
5545 (riscv_use_save_libcall): Return false for interrupt handler.
5546 (riscv_first_stack_step): Add forward declaration.
5547 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5548 for interrupt handler with large frame. Use it for saved reg list.
5549 (riscv_expand_prologue): Move flag_stack_usage_info support to
5550 eliminate duplication.
5551 (riscv_expand_epilogue): Generate mret for interrupt handler.
5552 (riscv_epilogue_uses): New.
5553 (riscv_can_use_return_insn): Return false for interrupt handler.
5554 (riscv_function_ok_for_sibcall): Likewise.
5555 (riscv_set_current_function): Add interrupt handler support.
5556 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5557 * config/riscv/riscv.md (UNSPECV_MRET): New.
5559 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5561 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5563 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5565 PR tree-optimization/85712
5566 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5567 this candidate has already been replaced in-situ by a copy.
5569 2018-05-25 Jason Merrill <jason@redhat.com>
5571 PR c++/80485 - inline function non-zero address.
5572 * symtab.c (nonzero_address): Check DECL_COMDAT.
5574 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5577 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5578 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5579 extension of SImode operation. Use const123_operand predicate.
5580 (*saddsi_1): Remove.
5581 (*saddl_se_1): Ditto.
5583 (*ssubl_se_1): Ditto.
5584 * config/alpha/predicates.md (const123_operand): New predicate.
5585 * config/alpha/constraints.md (P): Use IN_RANGE.
5587 2018-05-25 Richard Biener <rguenther@suse.de>
5589 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5591 (ref_maybe_used_by_stmt_p): Likewise.
5592 (stmt_may_clobber_ref_p): Likewise.
5593 (stmt_may_clobber_ref_p_1): Likewise.
5594 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5596 (ref_maybe_used_by_stmt_p): Likewise.
5597 (stmt_may_clobber_ref_p): Likewise.
5598 (stmt_may_clobber_ref_p_1): Likewise.
5599 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5600 the alias oracle to disambiguate DRs with stmts DR analysis
5602 (vect_analyze_data_refs): Do not give up on not analyzable
5603 DRs for BB vectorization. Remove code truncating the dataref
5606 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5609 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5610 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5611 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5613 2018-05-25 Richard Biener <rguenther@suse.de>
5615 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5616 function, combining stmt data ref gathering and fatal analysis
5618 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5619 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5620 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5621 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5622 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5624 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5626 PR tree-optimization/85720
5627 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5628 SCC if all partitions are builtins.
5629 (version_loop_by_alias_check): New parameter. Generate cancelable
5630 runtime alias check if all partitions are builtins.
5631 (distribute_loop): Update call to above function.
5633 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5635 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5636 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5637 (parm_default_def_partition_arg): Ditto.
5638 (set_parm_default_def_partition): Ditto.
5639 (get_parm_default_def_partitions): Ditto and make it static.
5640 (get_undefined_value_partitions): Ditto and make it static.
5641 (remove_ssa_form): Refactor call to init_var_map here.
5642 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5643 computation for loop region.
5644 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5645 (register_default_def): Delete.
5646 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5647 (parm_default_def_partition_arg): Ditto.
5648 (set_parm_default_def_partition): Ditto.
5649 (get_parm_default_def_partitions): Ditto and make it static.
5650 (get_undefined_value_partitions): Ditto and make it static.
5651 (coalesce_with_default, coalesce_with_default): Update comment.
5652 (create_coalesce_list_for_region): New func factored out from
5653 create_outofssa_var_map.
5654 (populate_coalesce_list_for_outofssa): New func factored out from
5655 create_outofssa_var_map and coalesce_ssa_name.
5656 (create_outofssa_var_map): Delete.
5657 (coalesce_ssa_name): Refactor to support live range computation.
5658 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5659 (get_parm_default_def_partitions): Delete.
5660 (get_undefined_value_partitions): Ditto.
5661 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5662 computation for loop region.
5663 (new_tree_live_info, loe_visit_block): Ditto.
5664 (live_worklist, set_var_live_on_entry): Ditto.
5665 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5666 * tree-ssa-live.h (struct _var_map): New fields.
5667 (init_var_map): Change decl.
5668 (region_contains_p): New.
5670 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5672 * tree-ssa-live.h (live_merge_and_clear): Delete.
5674 2018-05-25 Richard Biener <rguenther@suse.de>
5677 * tree-dump.c (dequeue_and_dump): Remove access to removed
5678 operand 2 of a SWITCH_EXPR.
5680 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5682 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5683 multiplication and division.
5684 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5685 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5686 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5687 (cond_udiv_optab, cond_umod_optab): New optabs.
5688 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5689 (IFN_COND_RDIV): New internal functions.
5690 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5691 TRUNC_MOD_EXPR and RDIV_EXPR.
5692 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5693 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5695 (SVE_INT_BINARY): Include mult.
5696 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5697 (optab, sve_int_op): Handle mult.
5698 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5700 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5701 for SVE_INT_BINARY_SD.
5703 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5705 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5706 (optab, sve_int_op): Handle div and udiv.
5707 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5708 for SVE_INT_BINARY_SD.
5709 (*<optab><mode>3): New insn for the same.
5711 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5713 * tree-vect-patterns.c: Include predict.h.
5714 (vect_recog_divmod_pattern): Restrict check for division support
5715 to when optimizing for size.
5717 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5719 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5720 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5721 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5722 (gimple_match_op::set_op): Likewise.
5723 (gimple_resimplify4): Declare.
5724 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5725 (expr::gen_transform): Likewise.
5726 (decision_tree::gen): Generate a simplification routine for 4 operands.
5727 * gimple-match-head.c (gimple_simplify): Add an overload for
5728 4 operands. In the top-level function, handle up to 4 call
5729 arguments and call gimple_resimplify4.
5730 (gimple_resimplify4): New function.
5731 (build_call_internal): Pass a fourth operand.
5732 (maybe_push_to_seq): Likewise.
5733 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5734 Fold VEC_COND_EXPRs of an operation and a default value into
5735 an IFN_COND_* function if possible.
5736 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5738 (SVE_COND_FP_BINARY): Include them.
5739 (optab, sve_fp_op): Handle them.
5740 (SVE_INT_BINARY_REV): New code iterator.
5741 (SVE_COND_FP_BINARY_REV): New int iterator.
5742 (commutative): New int attribute.
5743 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5745 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5747 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5748 (*cond_<optab><mode>): New patterns for reversed operands.
5750 2018-05-25 Richard Biener <rguenther@suse.de>
5752 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5753 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5754 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5755 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5756 * tree-vect-data-refs.c (everywhere): Adjust users.
5757 * tree-vect-loop.c (everywhere): Likewise.
5758 * tree-vect-slp.c (everywhere): Likewise.
5759 * tree-vect-stmts.c (everywhere): Likewise.
5760 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5762 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5764 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5766 (gcc_cv_as_section_exclude): ... this.
5767 Try Solaris as #exclude syntax.
5768 * configure: Regenerate.
5769 * config.in: Regenerate.
5770 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5772 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5773 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5775 * varasm.c (default_elf_asm_named_section): Don't check if
5776 HAVE_GAS_SECTION_EXCLUDE is defined.
5778 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5780 * doc/md.texi: Update the documentation of the cond_* optabs
5781 to mention the new final operand. Fix GET_MODE_NUNITS call.
5782 Describe the scalar case too.
5783 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5784 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5786 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5787 (get_conditional_internal_fn): Update comment.
5788 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5789 accumulator value as a final argument to conditional functions.
5790 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5791 a define_expand and add an "else" operand. Assert for now that
5792 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5793 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5794 (*cond_<optab><mode>): New patterns.
5795 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5796 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5797 (UNSPEC_COND_EOR): Delete.
5798 (optab): Remove associated mappings.
5799 (SVE_INT_BINARY): New code iterator.
5800 (sve_int_op): Remove int attribute and add "minus" to the code
5802 (SVE_COND_INT_OP): Delete.
5803 (SVE_COND_FP_OP): Rename to...
5804 (SVE_COND_FP_BINARY): ...this.
5806 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5808 * optabs.c (can_reuse_operands_p): New function.
5809 (maybe_legitimize_operands): Try to reuse the results for
5812 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5814 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5815 Add {q} suffix to insn mnemonic.
5817 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5819 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5820 (msp430_warn_func_return): New.
5822 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5824 * fold-const.c (tree_nonzero_bits): New function.
5825 * fold-const.h (tree_nonzero_bits): Likewise.
5826 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5827 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5829 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5833 * varasm.c (assemble_alias): Check ifunc_resolver only on
5836 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5839 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5840 when memory input operand is handled.
5842 2018-05-24 Luis Machado <luis.machado@linaro.org>
5844 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5846 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5848 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5850 * match.pd: Delay FMA folds until after vectorization.
5852 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5855 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5856 address check not strict.
5858 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5860 * gimple-match.h (gimple_match_op): New class.
5861 (mprts_hook): Replace parameters with a gimple_match_op *.
5862 (maybe_build_generic_op): Likewise.
5863 (gimple_simplified_result_is_gimple_val): Replace parameters with
5864 a const gimple_match_op *.
5865 (gimple_simplify): Replace code_helper * and tree * parameters with
5866 a gimple_match_op * parameter.
5867 (gimple_resimplify1): Replace code_helper *, tree and tree *
5868 parameters with a gimple_match_op * parameter.
5869 (gimple_resimplify2): Likewise.
5870 (gimple_resimplify3): Likewise.
5871 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5872 parameters with a gimple_match_op * parameter.
5873 * gimple-match-head.c (gimple_simplify): Change prototypes of
5874 auto-generated functions to take a gimple_match_op * instead of
5875 separate code_helper * and tree * parameters. Make the same
5876 change in the top-level overload and update calls to the
5877 gimple_resimplify routines. Update calls to the auto-generated
5878 functions and to maybe_push_res_to_seq in the publicly-facing
5879 operation-specific gimple_simplify overloads.
5880 (gimple_match_op::MAX_NUM_OPS): Define.
5881 (gimple_resimplify1): Replace rcode and ops with a single res_op
5882 parameter. Update call to gimple_simplify.
5883 (gimple_resimplify2): Likewise.
5884 (gimple_resimplify3): Likewise.
5885 (mprts_hook): Replace parameters with a gimple_match_op *.
5886 (maybe_build_generic_op): Likewise.
5887 (build_call_internal): Replace type, nargs and ops with
5888 a gimple_match_op *.
5889 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5890 with a single gimple_match_op *. Update calls to mprts_hook,
5891 build_call_internal and gimple_simplified_result_is_gimple_val.
5892 Factor out code that is common to the tree_code and combined_fn cases.
5893 * genmatch.c (expr::gen_transform): Replace tem_code and
5894 tem_ops with a gimple_match_op called tem_op. Update calls
5895 to the gimple_resimplify functions and maybe_push_res_to_seq.
5896 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5897 res_ops. Update call to the gimple_resimplify functions.
5898 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5899 (decision_tree::gen): Make the functions take a gimple_match_op *
5900 called res_op instead of separate res_code and res_ops parameters.
5901 Update call accordingly.
5902 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5903 and ops with a single res_op parameter. Update calls to
5904 maybe_build_generic_op and maybe_push_res_to_seq.
5905 (fold_stmt_1): Update calls to gimple_simplify and
5906 replace_stmt_with_simplification.
5907 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5908 and gimple_simplified_result_is_gimple_val.
5909 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5911 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5912 with a gimple_match_op *.
5913 (vn_nary_build_or_lookup): Likewise. Update call to
5914 vn_nary_build_or_lookup_1.
5915 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5916 gimple_match_op *. Update calls to the gimple_resimplify routines
5917 and to gimple_simplified_result_is_gimple_val.
5918 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5919 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5920 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5921 (visit_nary_op): Likewise.
5922 (visit_reference_op_load): Likewise.
5924 2018-05-23 Luis Machado <luis.machado@linaro.org>
5926 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5927 modifier for printing the step amount.
5929 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5932 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5935 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5937 * doc/sourcebuild.texi (Endianness): New subsubsection.
5939 2018-05-23 Luis Machado <luis.machado@linaro.org>
5941 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5942 <prefetch_dynamic_strides>: New const bool field.
5943 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5944 prefetch_dynamic_strides.
5945 (exynosm1_prefetch_tune): Likewise.
5946 (thunderxt88_prefetch_tune): Likewise.
5947 (thunderx_prefetch_tune): Likewise.
5948 (thunderx2t99_prefetch_tune): Likewise.
5949 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5951 (aarch64_override_options_internal): Update to set
5952 PARAM_PREFETCH_DYNAMIC_STRIDES.
5953 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5954 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5955 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5956 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5957 prefetch-dynamic-strides setting.
5959 2018-05-23 Luis Machado <luis.machado@linaro.org>
5961 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5962 <minimum_stride>: New const int field.
5963 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5964 minimum_stride field defaulting to -1.
5965 (exynosm1_prefetch_tune): Likewise.
5966 (thunderxt88_prefetch_tune): Likewise.
5967 (thunderx_prefetch_tune): Likewise.
5968 (thunderx2t99_prefetch_tune): Likewise.
5969 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5970 <default_opt_level>: Set to 3.
5971 (aarch64_override_options_internal): Update to set
5972 PARAM_PREFETCH_MINIMUM_STRIDE.
5973 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5974 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5975 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5976 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5977 stride is constant and is below the minimum stride threshold.
5979 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5981 * config/arm/arm-cpus.in (mode26): Delete.
5982 (armv4): Delete mode26 reference.
5983 * config/arm/arm.c (arm_configure_build_target): Delete use of
5986 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5988 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5990 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5991 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5993 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5994 (floatunsdidf2): Ditto.
5996 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5998 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5999 (fixuns_trunc<mode>si2_avx512f): Ditto.
6000 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6001 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6002 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6004 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6006 PR rtl-optimization/79985
6007 * df-scan.c (df_insn_refs_collect): Remove special case for
6008 global registers and asm statements.
6010 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6012 * extend.texi (Global Register Variables): Rewrite the bullet list.
6013 Note that the register is available for allocation. Note that access
6014 via inline asm must use constraints. Add note about async-signal
6015 handlers. Remove paragraph about automagic register selection.
6017 2018-05-23 Richard Biener <rguenther@suse.de>
6019 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6020 of fixed offset from memset VN.
6022 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6024 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6026 (alloc_cand_and_find_basis): Initialize first_interp field.
6027 (slsr_process_mul): Modify first_interp field.
6028 (slsr_process_add): Likewise.
6029 (slsr_process_cast): Modify first_interp field for each new
6031 (slsr_process_copy): Likewise.
6032 (dump_candidate): Dump first_interp field.
6033 (replace_mult_candidate): Process all interpretations, not just
6035 (replace_rhs_if_not_dup): Likewise.
6036 (replace_one_candidate): Likewise.
6038 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6040 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6042 (aarch64_needs_frame_chain): New function.
6043 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6045 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6048 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6049 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6050 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6051 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6052 as true for strict-align.
6053 (aarch64_can_inline_p): Perform checks even when callee has no
6054 attributes to check for strict alignment.
6055 * doc/extend.texi (AArch64 Function Attributes): Document
6057 * doc/invoke.texi: (AArch64 Options): Likewise.
6059 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6061 PR tree-optimization/85853
6062 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6063 the handling of the root of the node to...
6064 (vect_slp_analyze_node_operations_1): ...this new function,
6065 and run the whole thing with the child nodes' def types
6066 set according to their SLP node's def type.
6068 2018-05-23 Richard Biener <rguenther@suse.de>
6071 * tree-data-ref.c (create_runtime_alias_checks): Defer
6072 and ignore overflow warnings.
6074 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6076 PR tree-optimization/85822
6077 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6080 2018-05-23 Richard Biener <rguenther@suse.de>
6082 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6083 memset constants via native_interpret_expr.
6085 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6088 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6090 (cgraph_node::create_alias): Likewise.
6091 (cgraph_node::get_availability): Check ifunc_resolver instead
6092 of looking up ifunc attribute.
6093 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6094 * varasm.c (do_assemble_alias): Likewise.
6095 (assemble_alias): Likewise.
6096 (default_binds_local_p_3): Likewise.
6097 * cgraph.h (cgraph_node): Add ifunc_resolver.
6098 (cgraph_node::only_called_directly_or_aliased_p): Return false
6100 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6102 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6103 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6104 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6105 instead of looking up ifunc attribute.
6107 2018-05-22 Luis Machado <luis.machado@linaro.org>
6109 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6111 2018-05-22 Martin Sebor <msebor@redhat.com>
6114 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6115 only when expasion succeeds.
6116 (expand_builtin_strcmp): Same.
6117 (expand_builtin_strncmp): Same.
6119 2018-05-22 Martin Sebor <msebor@redhat.com>
6121 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6123 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6124 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6126 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6127 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6129 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6130 Add check for legitimate_address.
6131 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6132 (aarch64_swap_ldrstr_operands): New.
6133 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6136 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6137 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6139 * config/aarch64/aarch64.md: New patterns to generate stp
6141 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6142 single words and double words.
6143 (load_pair_sw, load_pair_dw): Likewise.
6144 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6146 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6148 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6149 for different mode ldpstp and add peephole for merged zero stores.
6151 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6153 (aarch64_gen_store_pair): Rename calls to match new patterns.
6154 (aarch64_gen_load_pair): Rename calls to match new patterns.
6155 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6156 (load_pair<DREG:mode><DREG2:mode>): ... This.
6157 (store_pair<mode>): Rename to...
6158 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6159 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6161 (V_INT_EQUIV): Handle SImode.
6162 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6165 2018-05-22 Martin Sebor <msebor@redhat.com>
6168 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6169 or ajust the presumed bound on an operation to avoid unnecessary
6172 2018-05-22 Martin Sebor <msebor@redhat.com>
6174 PR tree-optimization/85826
6175 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6176 assuming that a DECL necesarily has a constant size.
6178 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6181 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6183 2018-05-22 Richard Biener <rguenther@suse.de>
6185 PR tree-optimization/85834
6186 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6187 non-constant and non-zero memset arguments.
6189 2018-05-22 Martin Liska <mliska@suse.cz>
6192 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6194 2018-05-22 Richard Biener <rguenther@suse.de>
6196 PR tree-optimization/85863
6197 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6198 comparisons when vectype is specified.
6199 (vectorizable_condition): Do not specify vectype for
6200 vect_is_simple_cond when SLP vectorizing.
6202 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6205 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6206 define __ibm128 as long double.
6207 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6208 as a distinct type when IEEE 128-bit support is enabled.
6209 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6210 128-bit types to use the correct functions.
6211 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6212 convert between 128-bit floating point types that have different
6213 modes but the same representation, instead of using gen_lowpart to
6215 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6217 (IFKF_reg): New attributes to give the register constraints for
6219 (extend<mode>tf2_internal): New insns to mark an explicit
6220 conversion between 128-bit floating point types that have a
6221 different mode but share the same representation.
6223 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6225 PR tree-optimization/85814
6226 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6227 a null return from get_strinfo when unsharing the next
6228 strinfo in the chain.
6230 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6233 * varasm.c (weak_finish): Clean up weak_decls.
6235 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6237 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6238 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6239 UNSPEC_UADALP values.
6240 * config/aarch64/iterators.md (ABAL): New int iterator.
6243 (sur): Add mappings for the above.
6244 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6246 (aarch64_<sur>abal<mode>_4): Likewise.
6247 (aarch64_<sur>adalp<mode>_3): Likewise.
6248 (<sur>sadv16qi): New define_expand.
6250 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6252 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6253 (*movdf_internal): Ditto.
6254 (*rcpsf2_sse): Ditto.
6255 (*rsqrtsf2_sse): Ditto.
6256 (*sqrt<mode>2_sse): Ditto.
6258 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6260 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6262 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6263 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6264 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6265 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6267 * config/aarch64/arm_neon.h: Likewise.
6268 * config/aarch64/iterators.md (VQ_I): New.
6270 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6272 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6274 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6277 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6279 * config/nds32/constraints.md (S): New constraint.
6280 * config/nds32/nds32.md (call_internal): Use constraint S.
6281 (call_value_internal): Likewise.
6282 (sibcall_internal): Likewise.
6283 (sibcall_value_internal): Likewise.
6285 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6286 Chung-Ju Wu <jasonwucj@gmail.com>
6288 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6291 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6292 Chung-Ju Wu <jasonwucj@gmail.com>
6294 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6295 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6296 (nds32_rtx_costs_impl): Simplify.
6297 (nds32_address_cost_impl): Simplify.
6298 (nds32_init_rtx_costs): New function.
6299 (nds32_rtx_costs_speed_prefer): Likewise.
6300 (nds32_rtx_costs_size_prefer): Likewise.
6301 (nds32_address_cost_speed_prefer): Likewise.
6302 (nds32_address_cost_speed_fwprop): Likewise.
6303 (nds32_address_cost_size_prefer): Likewise.
6304 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6305 * config/nds32/nds32.c (nds32_option_override): Use
6306 nds32_init_rtx_costs function.
6308 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6310 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6311 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6312 (TARGET_PIPELINE_N8): Likewise.
6313 (TARGET_PIPELINE_N10): Likewise.
6314 (TARGET_PIPELINE_N13): Likewise.
6315 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6317 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6319 * config/nds32/nds32-fpu.md: Update copyright year.
6321 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6323 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6325 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6327 * config/nds32/nds32.c
6328 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6329 * config/nds32/nds32.opt (minline-asm-r15): New option.
6331 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6333 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6335 * config/nds32/nds32.md (abssi2): New pattern.
6337 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6339 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6340 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6341 Merge insn pattern from sse_cvtsi2ss<round_name> and
6342 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6343 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6344 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6345 using SWI48 mode iterator.
6346 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6347 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6348 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6349 pattern from sse_cvttss2si<round_saeonly_name>
6350 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6351 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6352 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6353 using SWI48 mode iterator.
6354 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6355 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6356 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6357 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6358 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6359 using SWI48 mode iterator.
6360 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6361 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6362 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6363 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6364 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6365 SWI48 mode iterator.
6366 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6367 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6368 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6369 pattern from sse_cvttsd2si<round_saeonly_name>
6370 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6372 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6374 * config/nds32/nds32-md-auxiliary.c
6375 (nds32_valid_smw_lwm_base_p): Refine.
6376 (nds32_output_smw_single_word): Refine.
6377 (nds32_output_smw_double_word): New.
6378 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6380 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6382 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6383 (nds32_output_stack_pop): Refine.
6384 (nds32_expand_unaligned_load): Refine.
6385 (nds32_expand_unaligned_store): Refine.
6387 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6388 Chung-Ju Wu <jasonwucj@gmail.com>
6390 * config/nds32/constants.md: Add TP_REGNUM constant.
6391 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6392 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6394 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6395 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6396 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6397 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6398 and PIC code generation.
6399 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6401 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6403 * config/nds32/nds32.md: Support TLS and PIC.
6404 * config/nds32/nds32.c: Support TLS and PIC.
6405 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6406 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6409 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6411 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6412 mode with E_ prefix.
6414 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6415 Chung-Ju Wu <jasonwucj@gmail.com>
6417 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6418 * config/nds32/nds32-md-auxiliary.c
6419 (symbolic_reference_mentioned_p): New.
6420 (nds32_legitimize_ict_address): New.
6421 (nds32_expand_ict_move): New.
6422 (nds32_indirect_call_referenced_p): New.
6423 (nds32_symbol_binds_local_p): Delete.
6424 (nds32_long_call_p): Modify.
6425 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6426 * config/nds32/nds32-protos.h
6427 (symbolic_reference_mentioned_p): Declare.
6428 (nds32_legitimize_ict_address): Declare.
6429 (nds32_expand_ict_move): Declare.
6430 (nds32_indirect_call_referenced_p): Declare.
6431 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6432 (nds32_relax_group): Use nds32_ict_const_p as condition.
6433 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6434 (nds32_asm_file_start): Output ict_model directive in asm code.
6435 (nds32_legitimate_address_p): Consider indirect call.
6436 (nds32_print_operand): Consider indirect call.
6437 (nds32_print_operand_address): Consider indirect call.
6438 (nds32_insert_attributes): Handle "indirect_call" attribute.
6439 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6440 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6441 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6442 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6443 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6444 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6445 (TARGET_ICT_MODEL_SMALL): Define.
6446 (TARGET_ICT_MODEL_LARGE): Define.
6447 * config/nds32/nds32.md (movsi): Consider ict model.
6448 (call, call_value): Consider ict model.
6449 (sibcall, sibcall_value): Consider ict model.
6450 * config/nds32/nds32.opt (mict-model): New option.
6451 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6454 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6455 Monk Chiang <sh.chiang04@gmail.com>
6456 Jim Wilson <jimw@sifive.com>
6458 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6459 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6460 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6461 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6462 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6463 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6464 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6465 compute save_libcall_adjustment properly.
6466 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6467 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6468 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6469 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6470 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6471 (ABI_SPEC): Handle mabi=ilp32e.
6472 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6473 (RVE): Add RVE mask.
6474 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6475 <-march>: Add rv32e as an example.
6477 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6480 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6481 (intra_create_variable_infos): Handle C++ constructors.
6483 2018-05-18 Martin Liska <mliska@suse.cz>
6485 * passes.def: Remove a redundant pass.
6487 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6490 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6492 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6494 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6496 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6498 (armv2, armv2a, armv3, armv3m): Delete architectures.
6499 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6500 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6501 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6503 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6504 (*mulsidi3adddi): Likewise.
6505 (mulsidi3): Likewise.
6506 (*mulsidi3_nov6): Likewise.
6507 (umulsidi3): Likewise.
6508 (umulsidi3_nov6): Likewise.
6509 (umaddsidi4): Likewise.
6510 (*umulsidi3adddi): Likewise.
6511 (smulsi3_highpart): Likewise.
6512 (*smulsi3_highpart_nov6): Likewise.
6513 (umulsi3_highpart): Likewise.
6514 (*umulsi3_highpart_nov6): Likewise.
6515 * config/arm/arm.h (arm_arch3m): Delete.
6516 * config/arm/arm.c (arm_arch3m): Delete.
6517 (arm_option_override_internal): Update armv3-related comment.
6518 (arm_configure_build_target): Delete use of isa_bit_mode32.
6519 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6520 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6521 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6523 (mulusa3): Likewise.
6524 * config/arm/arm-protos.h (arm_arch3m): Delete.
6525 * config/arm/arm-tables.opt: Regenerate.
6526 * config/arm/arm-tune.md: Likewise.
6527 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6528 deleted architectures.
6530 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6532 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6533 (armv5t, armv5te): New features.
6534 (ARMv5, ARMv5e): Delete fgroups.
6535 (ARMv5t, ARMv5te): Adjust for above changes.
6537 (armv5, armv5e): Delete arches.
6538 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6540 (*call_reg_arm): Likewise.
6541 (*call_value_reg_armv5): Likewise.
6542 (*call_value_reg_arm): Likewise.
6543 (*call_symbol): Likewise.
6544 (*call_value_symbol): Likewise.
6545 (*sibcall_insn): Likewise.
6546 (*sibcall_value_insn): Likewise.
6548 (prefetch): Likewise.
6549 (define_split and define_peephole2 dependent on arm_arch5):
6551 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6553 (TARGET_ARM_QBIT): Likewise.
6554 (TARGET_DSP_MULTIPLY): Likewise.
6555 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6556 (arm_arch5, arm_arch5e): Delete.
6557 (arm_arch5t, arm_arch5te): Declare.
6558 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6559 (arm_arch5t): Declare.
6560 (arm_option_reconfigure_globals): Update for the above.
6561 (arm_options_perform_arch_sanity_checks): Update comment, replace
6562 use of arm_arch5 with arm_arch5t.
6563 (use_return_insn): Likewise.
6564 (arm_emit_call_insn): Likewise.
6565 (output_return_instruction): Likewise.
6566 (arm_final_prescan_insn): Likewise.
6567 (arm_coproc_builtin_available): Likewise.
6568 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6569 arm_arch5e with arm_arch5t and arm_arch5te.
6570 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6571 (arm_arch5t, arm_arch5te): Declare.
6572 * config/arm/arm-tables.opt: Regenerate.
6573 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6574 * config/arm/t-multilib: Likewise.
6575 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6576 instead of arm_arch5.
6577 (*call_reg_thumb1): Likewise.
6578 (*call_value_reg_thumb1_v5): Likewise.
6579 (*call_value_reg_thumb1): Likewise.
6580 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6582 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6584 2018-05-18 Martin Liska <mliska@suse.cz>
6586 PR gcov-profile/84846
6587 * doc/gcov.texi: Document -t option of gcov tool.
6589 2018-05-18 Martin Liska <mliska@suse.cz>
6591 PR gcov-profile/84846
6592 * gcov.c (print_usage): Add new -t option.
6593 (process_args): Handle the option.
6594 (generate_results): Use stdout as output when requested by
6597 2018-05-18 Martin Liska <mliska@suse.cz>
6599 PR gcov-profile/84846
6600 * coverage.c (coverage_init): Write PWD to .gcno file.
6601 * doc/gcov.texi: Document how working directory is printed.
6602 * gcov-dump.c (dump_gcov_file): Print PWD.
6603 * gcov.c (output_intermediate_file): Likewise.
6604 (read_graph_file): Read PWD string.
6605 (output_lines): Print PWD.
6607 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6610 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6611 for retval and return false if all args to phi are zero.
6613 2018-05-18 Richard Biener <rguenther@suse.de>
6615 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6617 (evrp_dom_walker::before_dom_children): Call it.
6619 2018-05-18 Richard Biener <rguenther@suse.de>
6621 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6622 results when processing array refs with variable index.
6624 2018-05-18 Toon Moene <toon@moene.org>
6626 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6627 directly after that of -floop-interchange. Indicate that both
6628 options are enabled by default when specifying -O3.
6630 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6632 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6633 iterator. Delete separate integer-mode vec_set<mode> expander.
6634 (aarch64_simd_vec_setv2di): Delete.
6635 (vec_setv2di): Delete.
6636 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6637 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6638 the "w, r" alternative.
6640 2018-05-18 Martin Liska <mliska@suse.cz>
6642 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6643 * tree-pass.h (make_pass_lower_switch_O0): New function.
6644 * tree-switch-conversion.c (node_has_low_bound): Remove.
6645 (node_has_high_bound): Likewise.
6646 (node_is_bounded): Likewise.
6647 (class pass_lower_switch): Make it a template type and create
6649 (pass_lower_switch::execute): Add template argument.
6650 (make_pass_lower_switch): New function.
6651 (make_pass_lower_switch_O0): New function.
6652 (do_jump_if_equal): Remove.
6653 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6654 all the hard work to tree optimization passes.
6656 2018-05-18 Martin Liska <mliska@suse.cz>
6658 * dbgcnt.c (limit_low): Renamed from limit.
6659 (limit_high): New variable.
6660 (dbg_cnt_is_enabled): Check for upper limit.
6661 (dbg_cnt): Adjust dumping.
6662 (dbg_cnt_set_limit_by_index): Add new argument for high
6664 (dbg_cnt_set_limit_by_name): Likewise.
6665 (dbg_cnt_process_single_pair): Parse new format.
6666 (dbg_cnt_process_opt): Use strtok.
6667 (dbg_cnt_list_all_counters): Remove 'value' and add
6669 * doc/invoke.texi: Document changes.
6671 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6673 * doc/sourcebuild.texi (scalar_all_fma): Document.
6674 * tree.def (FMA_EXPR): Delete.
6675 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6676 * internal-fn.c (ternary_direct): New macro.
6677 (expand_ternary_optab_fn): Likewise.
6678 (direct_ternary_optab_supported_p): Likewise.
6679 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6680 * builtins.c (fold_builtin_fma): Delete.
6681 (fold_builtin_3): Don't call it.
6682 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6683 * expr.c (expand_expr_real_2): Likewise.
6684 * fold-const.c (operand_equal_p): Likewise.
6685 (fold_ternary_loc): Likewise.
6686 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6687 * gimple.c (DEFTREECODE): Likewise.
6688 * gimplify.c (gimplify_expr): Likewise.
6689 * optabs-tree.c (optab_for_tree_code): Likewise.
6690 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6691 * tree-eh.c (operation_could_trap_p): Likewise.
6692 (stmt_could_throw_1_p): Likewise.
6693 * tree-inline.c (estimate_operator_cost): Likewise.
6694 * tree-pretty-print.c (dump_generic_node): Likewise.
6695 (op_code_prio): Likewise.
6696 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6697 * tree-ssa-operands.c (get_expr_operands): Likewise.
6698 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6699 * fold-const-call.h (fold_fma): Delete.
6700 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6701 CFN_FNMA and CFN_FNMS.
6703 * genmatch.c (combined_fn): New enum.
6704 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6705 (commutative_op): New function.
6706 (commutate): Use it. Handle more than 2 operands.
6707 (dt_operand::gen_gimple_expr): Use commutative_op.
6708 (parser::parse_expr): Allow :c to be used with non-binary
6709 operators if the commutative operand is known.
6710 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6711 CFN_FMS, CFN_FNMA and CFN_FNMS.
6712 (backprop::process_assign_use): Remove FMA_EXPR handling.
6713 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6714 (gen_hsa_fma): New function.
6715 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6716 IFN_FNMA and IFN_FNMS.
6717 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6718 * gimple-fold.h (follow_all_ssa_edges): Declare.
6719 * gimple-fold.c (follow_all_ssa_edges): New function.
6720 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6721 gimple_build interface and use follow_all_ssa_edges to fold the result.
6722 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6723 instead of checking for optabs directly.
6724 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6725 rather than FMA_EXPRs.
6726 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6727 call to IFN_FMA instead of an FMA_EXPR.
6729 2018-05-17 Jim Wilson <jimw@sifive.com>
6731 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6732 sign extended promoted subreg, and the range does not have the sign bit
6733 set, then do a sign extend.
6735 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6736 test, check for sign extended subreg and/or constant operands, and
6737 do a sign extend in that case.
6739 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6741 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6742 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6744 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6745 Change logics_shift_reg to logics_shift_imm.
6746 (thunderx2t99_fp_loadpair_basic): Delete.
6747 (thunderx2t99_fp_storepair_basic): Delete.
6748 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6749 (thunderx2t99_asimd_polynomial): Delete.
6750 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6751 and neon_fp_mul_d_scalar_q.
6752 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6753 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6754 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6755 (thunderx2t99_asimd_lut): Add missing tbl types.
6756 (thunderx2t99_asimd_ext): Delete.
6757 (thunderx2t99_asimd_load1_1_mult): Delete.
6758 (thunderx2t99_asimd_load1_2_mult): Delete.
6759 (thunderx2t99_asimd_load1_ldp): New.
6760 (thunderx2t99_asimd_load1): New.
6761 (thunderx2t99_asimd_load2): Add missing *load2* types.
6762 (thunderx2t99_asimd_load3): New.
6763 (thunderx2t99_asimd_load4): New.
6764 (thunderx2t99_asimd_store1_1_mult): Delete.
6765 (thunderx2t99_asimd_store1_2_mult): Delete.
6766 (thunderx2t99_asimd_store2_mult): Delete.
6767 (thunderx2t99_asimd_store2_onelane): Delete.
6768 (thunderx2t99_asimd_store_stp): New.
6769 (thunderx2t99_asimd_store1): New.
6770 (thunderx2t99_asimd_store2): New.
6771 (thunderx2t99_asimd_store3): New.
6772 (thunderx2t99_asimd_store4): New.
6774 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6776 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6777 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6779 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6780 Segher Boessenkool <segher@kernel.crashing.org>
6783 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6786 2018-05-17 Richard Biener <rguenther@suse.de>
6788 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6789 for pruning loop and prune defs feeding only already visited PHIs.
6791 2018-05-17 Richard Biener <rguenther@suse.de>
6793 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6795 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6796 Richard Biener <rguenther@suse.de>
6798 PR tree-optimization/85793
6799 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6800 for VMAT_ELEMENTWISE.
6802 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6804 * internal-fn.h (lookup_internal_fn): Declare
6805 * internal-fn.c (lookup_internal_fn): New function.
6806 * gimple.c (gimple_build_call_from_tree): Handle calls to
6808 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6809 internal function names.
6810 * tree-pretty-print.c (dump_generic_node): Likewise.
6811 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6813 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6815 * gimple-fold.h (gimple_build): Make the function forms take
6816 combined_fn rather than built_in_function.
6817 (gimple_simplify): Likewise.
6818 * gimple-match-head.c (gimple_simplify): Likewise.
6819 * gimple-fold.c (gimple_build): Likewise.
6820 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6821 rather than gimple_build_call_internal.
6822 (get_initial_defs_for_reduction): Likewise.
6823 (vect_create_epilog_for_reduction): Likewise.
6824 (vectorizable_live_operation): Likewise.
6826 2018-05-17 Martin Liska <mliska@suse.cz>
6828 * gimple-ssa-sprintf.c (format_directive): Do not use
6829 space in between 'G_' and '('.
6831 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6834 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6835 even if the mask is not all ones.
6838 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6840 (ix86_gimple_fold_builtin): Likewise.
6843 * config/i386/i386.c: Include tree-vector-builder.h.
6844 (ix86_vector_shift_count): New function.
6845 (ix86_fold_builtin): Fold shift builtins by scalar count.
6846 (ix86_gimple_fold_builtin): Likewise.
6848 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6849 _mm512_setzero): New intrinsics.
6851 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6852 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6854 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6855 code generation for cases where splatting a value is not useful.
6856 * simplify-rtx.c (simplify_ternary_operation): Simplify
6857 vec_merge across a vec_duplicate and a paradoxical subreg forming
6858 a vector mode to a vec_concat.
6860 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6862 * config.gcc: Support "goldmont-plus".
6863 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6865 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6866 PROCESSOR_GOLDMONT_PLUS.
6867 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6868 (processor_target_table): Add "goldmont-plus".
6869 (PTA_GOLDMONT_PLUS): Define.
6870 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6871 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6872 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6873 (fold_builtin_cpu): Add "goldmont-plus".
6874 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6875 (ix86_option_override_internal): Add "goldmont-plus".
6876 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6877 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6878 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6879 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6881 2018-05-17 Richard Biener <rguenther@suse.de>
6883 PR tree-optimization/85757
6884 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6885 remove defs that only feed that PHI from further processing.
6887 2018-05-16 Jim Wilson <jimw@sifive.com>
6889 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6891 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6893 2018-05-16 Mark Wielaard <mark@klomp.org>
6895 * dwarf2out.c (count_index_strings): New function.
6896 (output_indirect_strings): Call count_index_strings and generate
6897 header for dwarf_version >= 5.
6899 2018-05-16 Mark Wielaard <mark@klomp.org>
6901 * dwarf2out.c (dwarf_FORM): New function.
6902 (set_indirect_string): Use dwarf_FORM.
6903 (reset_indirect_string): Likewise.
6904 (size_of_die): Likewise.
6905 (value_format): Likewise.
6906 (output_die): Likewise.
6907 (add_skeleton_AT_string): Likewise.
6908 (output_macinfo_op): Likewise.
6909 (index_string): Likewise.
6910 (output_index_string_offset): Likewise.
6911 (output_index_string): Likewise.
6912 (count_index_strings): Likewise.
6914 2018-05-16 Carl Love <cel@us.ibm.com>
6916 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6917 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6919 2018-05-16 Martin Jambor <mjambor@suse.cz>
6921 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6922 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6924 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6926 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6927 (fnma<mode>4): Likewise.
6928 (fms<mode>4): Likewise.
6929 (fnms<mode>4): Likewise.
6930 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6931 (aarch64_fnma<mode>4): Likewise.
6932 (aarch64_fms<mode>4): Likewise.
6933 (aarch64_fnms<mode>4): Likewise.
6934 (aarch64_fnmadd<mode>4): Likewise.
6936 2018-05-16 Jason Merrill <jason@redhat.com>
6938 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6940 2018-05-16 Richard Biener <rguenther@suse.de>
6942 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6943 (dump_stmt_cost): Declare.
6944 (add_stmt_cost): Dump cost we add.
6945 (add_stmt_costs): New function.
6946 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6948 (vect_analyze_stmt): Adjust prototype.
6949 (vectorizable_condition): Likewise.
6950 (vectorizable_live_operation): Likewise.
6951 (vectorizable_reduction): Likewise.
6952 (vectorizable_induction): Likewise.
6953 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6954 cost vector to pass to vectorizable_ and record afterwards.
6955 (vect_model_reduction_cost): Take cost vector argument and adjust.
6956 (vect_model_induction_cost): Likewise.
6957 (vectorizable_reduction): Likewise.
6958 (vectorizable_induction): Likewise.
6959 (vectorizable_live_operation): Likewise.
6960 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6961 SLP_TREE_NUMBER_OF_VEC_STMTS.
6962 (vect_analyze_slp_cost_1): Remove.
6963 (vect_analyze_slp_cost): Likewise.
6964 (vect_slp_analyze_node_operations): Take visited args and
6965 a target cost vector. Avoid processing already visited stmt sets.
6966 (vect_slp_analyze_operations): Use a local cost vector to gather
6967 costs and register those of non-discarded instances.
6968 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6969 (vect_schedule_slp_instance): Remove copying of
6970 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6972 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6973 adding cost. Record cost entry location.
6974 (vect_prologue_cost_for_slp_op): Function to compute cost of
6975 a constant or invariant generated for SLP vect in the prologue,
6976 split out from vect_analyze_slp_cost_1.
6977 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6978 (vect_model_promotion_demotion_cost): Likewise.
6979 (vect_model_store_cost): Likewise, make static.
6980 (vect_model_load_cost): Likewise.
6981 (vectorizable_bswap): Add cost vector arg and adjust.
6982 (vectorizable_call): Likewise.
6983 (vectorizable_simd_clone_call): Likewise.
6984 (vectorizable_conversion): Likewise.
6985 (vectorizable_assignment): Likewise.
6986 (vectorizable_shift): Likewise.
6987 (vectorizable_operation): Likewise.
6988 (vectorizable_store): Likewise.
6989 (vectorizable_load): Likewise.
6990 (vectorizable_condition): Likewise.
6991 (vectorizable_comparison): Likewise.
6992 (can_vectorize_live_stmts): Likewise.
6993 (vect_analyze_stmt): Likewise.
6994 (vect_transform_stmt): Adjust calls to vectorizable_*.
6995 * tree-vectorizer.c: Include gimple-pretty-print.h.
6996 (dump_stmt_cost): New function.
6998 2018-05-16 Richard Biener <rguenther@suse.de>
7000 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7001 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7002 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7003 (check_name): New callback.
7004 (dse_classify_store): Track cycles via a visited bitmap of PHI
7005 defs and simplify handling of in-loop and across loop dead stores
7006 and properly fail for loop-variant refs. Handle byte-tracking with
7007 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7010 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7012 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7013 (vect_get_mask_type_for_stmt): Likewise.
7014 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7016 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7017 to determine the statement's vector type and the vector type that
7018 should be used for calculating nunits. Deal with cases in which
7019 the type has to be deferred.
7020 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7021 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7022 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7023 (vect_determine_vf_for_stmt): New functions, split out from...
7024 (vect_determine_vectorization_factor): ...here.
7025 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7026 (vect_get_mask_type_for_stmt): New functions, split out from
7027 vect_determine_vectorization_factor.
7029 2018-05-16 Richard Biener <rguenther@suse.de>
7031 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7032 verify the [VEC_]COND_EXPR embedded comparison.
7034 2018-05-15 Martin Sebor <msebor@redhat.com>
7036 PR tree-optimization/85753
7037 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7038 RECORD_TYPE in addition to ARRAY_TYPE.
7040 2018-05-15 Martin Sebor <msebor@redhat.com>
7043 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7045 2018-05-15 Richard Biener <rguenther@suse.de>
7047 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7048 add by_clobber_p one. Change algorithm to collect all defs
7049 representing uses we need to walk and try reducing them to
7050 a single one before failing.
7051 (dse_dom_walker::dse_optimize_stmt): Adjust.
7053 2018-05-13 Mark Wielaard <mark@klomp.org>
7055 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7056 (size_of_loc_descr): Likewise.
7057 (output_loc_operands): Likewise.
7058 (output_loc_operands_raw): Likewise.
7059 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7060 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7061 (hash_loc_operands): Likewise.
7062 (compare_loc_operands): Likewise.
7064 2018-05-14 Mark Wielaard <mark@klomp.org>
7066 * dwarf2out.c (count_index_addrs): New function.
7067 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7069 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7071 PR tree-optimization/83648
7072 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7073 return value as malloc candidate.
7075 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7078 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7079 param as true in call to suggest_attribute.
7081 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7083 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7084 -mreadonly-in-sdata.
7086 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7088 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7090 (aarch64_crypto_aesd_fused): Likewise.
7092 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7094 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7095 (movsi_aarch64): Likewise.
7096 (load_pairsi): Likewise.
7097 (load_pairdi): Likewise.
7098 (store_pairsi): Likewise.
7099 (store_pairdi): Likewise.
7100 (load_pairsf): Likewise.
7101 (load_pairdf): Likewise.
7102 (store_pairsf): Likewise.
7103 (store_pairdf): Likewise.
7104 (zero_extend): Likewise.
7105 (trunc): Swap alternatives.
7106 (fcvt_target): Add '?' to prefer w over r.
7108 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7111 * config/i386/i386.md: Disallow non-commutative arithmetics in
7112 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7113 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7114 in the peephole2 before it.
7116 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7118 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7119 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7120 (ix86_handle_option): Handle -mcldemote.
7121 * config.gcc: New header.
7122 * config/i386/cldemoteintrin.h: New file.
7123 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7124 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7126 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7127 OPTION_MASK_ISA_CLDEMOTE.
7128 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7129 (ix86_valid_target_attribute_inner_p): Ditto.
7130 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7131 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7132 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7133 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7134 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7136 * config/i386/i386.opt: Add -mcldemote.
7137 * config/i386/x86intrin.h: New header.
7138 * doc/invoke.texi: Add -mcldemote.
7140 2018-05-14 Richard Biener <rguenther@suse.de>
7142 * doc/match-and-simplify.texi: Adjust :s documentation.
7144 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7146 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7147 intended memcpy size.
7148 (REORDER_45): Likewise.
7150 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7152 * sort.cc: New file.
7153 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7154 * vec.c (qsort_chk): Use gcc_qsort.
7155 * Makefile.in (OBJS-libcommon): Add sort.o.
7156 (build/sort.o): New target. Use it...
7157 (BUILD_RTL): ... here, and...
7158 (build/gencfn-macros): ... here, and...
7159 (build/genmatch): ... here.
7161 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7162 Chung-Ju Wu <jasonwucj@gmail.com>
7164 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7165 * config/nds32/nds32-graywolf.md: New file.
7166 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7167 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7169 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7170 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7171 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7172 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7173 * config/nds32/pipelines.md: Include n15 settings.
7175 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7176 Chung-Ju Wu <jasonwucj@gmail.com>
7178 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7179 * config/nds32/nds32-n13.md: New file.
7180 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7181 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7183 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7184 * config/nds32/nds32.md (pipeline_model): Add n13.
7185 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7186 * config/nds32/pipelines.md: Include n13 settings.
7188 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7189 Chung-Ju Wu <jasonwucj@gmail.com>
7191 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7192 * config/nds32/nds32-n10.md: New file.
7193 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7194 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7196 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7197 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7198 * config/nds32/nds32.md (pipeline_model): Add n10.
7199 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7200 * config/nds32/pipelines.md: Include n10 settings.
7202 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7203 Kito Cheng <kito.cheng@gmail.com>
7204 Chung-Ju Wu <jasonwucj@gmail.com>
7206 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7207 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7208 Add enum values for DSP extension instructions.
7209 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7211 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7212 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7214 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7215 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7216 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7217 * config/nds32/nds32-intrinsic.md: Likewise.
7218 * config/nds32/nds32_intrinsic.h: Likewise.
7219 * config/nds32/nds32-md-auxiliary.c: Likewise.
7220 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7221 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7222 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7223 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7224 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7225 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7226 TYPE_DMAC in switch statement.
7227 * config/nds32/nds32.c: New checking and implementation for DSP
7228 extension instructions.
7229 * config/nds32/nds32.h: Likewise.
7230 * config/nds32/nds32.md: Likewise.
7231 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7232 * config/nds32/predicates.md: Implement new predicates for DSP
7235 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7237 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7238 Reformat alternatives and attributes so it is easier to identify
7239 which constraints/attributes go with which instruction.
7240 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7241 (mov<mode>_softfloat32, FMOVE64): Likewise.
7242 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7243 (mov<mode>_softfloat64, FMOVE64): Likewise.
7245 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7247 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7249 (Basic PowerPC Built-in Functions): The new name of the
7250 subsection previously known as "PowerPC Built-in Functions".
7251 (Basic PowerPC Built-in Functions Available on all Configurations):
7253 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7254 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7255 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7256 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7258 2018-05-11 Martin Jambor <mjambor@suse.cz>
7261 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7264 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7267 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7269 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7271 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7272 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7273 (ix86_handle_option): Handle -mwaitpkg.
7274 * config.gcc: New header.
7275 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7276 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7277 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7279 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7280 OPTION_MASK_ISA_WAITPKG.
7281 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7282 (ix86_option_override_internal): Add PTA_WAITPKG.
7283 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7284 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7285 IX86_BUILTIN_TPAUSE.
7286 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7287 __builtin_ia32_umwait and __builtin_ia32_tpause.
7288 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7289 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7290 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7291 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7292 UNSPECV_TPAUSE): New.
7293 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7294 * config/i386/i386.opt: Add -mwaitpkg.
7295 * config/i386/waitpkgintrin.h: New file.
7296 * config/i386/x86intrin.h: New header.
7297 * doc/invoke.texi: Add -mwaitpkg.
7299 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7302 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7304 (cortex-m0): Use armv6s-m isa.
7305 (cortex-m0plus): Likewise.
7306 (cortex-m1): Likewise.
7307 (cortex-m0.small-multiply): Likewise.
7308 (cortex-m0plus.small-multiply): Likewise.
7309 (cortex-m1.small-multiply): Likewise.
7311 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7312 Jakub Jelinek <jakub@redhat.com>
7314 PR tree-optimization/85692
7315 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7316 source permute as well.
7318 2018-05-11 Martin Liska <mliska@suse.cz>
7321 * doc/extend.texi: Document LLVM style format for no_sanitize
7324 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7326 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7327 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7328 (mode_supports_vsx_dform_quad): Likewise.
7329 (mode_supports_vmx_dform): Move these functions to be next to the
7330 other mode_supports functions.
7331 (mode_supports_dq_form): Likewise.
7332 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7333 mode_supports_dq_form.
7334 (reg_offset_addressing_ok_p): Likewise.
7335 (offsettable_ok_by_alignment): Likewise.
7336 (rs6000_legitimate_offset_address_p): Likewise.
7337 (legitimate_lo_sum_address_p): Likewise.
7338 (rs6000_legitimize_address): Likewise.
7339 (rs6000_legitimize_reload_address): Likewise.
7340 (rs6000_secondary_reload_inner): Likewise.
7341 (rs6000_preferred_reload_class): Likewise.
7342 (rs6000_output_move_128bit): Likewise.
7344 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7346 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7347 Generate SImode target register for null target.
7348 <case IX86_BUILTIN_XGETBV>: Ditto.
7349 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7350 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7352 2018-05-10 Carl Love <cel@us.ibm.com>
7354 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7355 dcbtt and dcbtstt if operands[2] is 0.
7357 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7360 * config/i386/sse.md (usadv64qi): New expander.
7362 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7364 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7365 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7366 -maltivec=be support.
7367 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7368 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7369 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7370 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7371 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7372 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7373 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7374 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7375 altivec_vsumsws): Adjust.
7376 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7377 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7378 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7380 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7381 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7382 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7383 (altivec_lve<VI_char>x): Delete expand.
7384 (*altivec_lve<VI_char>x_internal): Rename to...
7385 (altivec_lve<VI_char>x): ... this.
7386 (altivec_lvxl_<mode>): Delete expand.
7387 (*altivec_lvxl_<mode>_internal): Rename to ...
7388 (altivec_lvxl_<mode>): ... this.
7389 (altivec_stvxl_<mode>): Delete expand.
7390 (*altivec_stvxl_<mode>_internal): Rename to ...
7391 (altivec_stvxl_<mode>): ... this.
7392 (altivec_stve<VI_char>x): Delete expand.
7393 (*altivec_stve<VI_char>x_internal): Rename to ...
7394 (altivec_stve<VI_char>x): ... this.
7395 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7396 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7397 reduc_plus_scal_<mode>): Adjust.
7398 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7400 (rs6000_cpu_cpp_builtins): Adjust.
7401 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7402 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7403 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7404 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7405 -maltivec=be support.
7406 (rs6000_split_vec_extract_var): Adjust.
7407 (rs6000_split_v4si_init): Adjust.
7408 (swap_selector_for_mode): Delete.
7409 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7410 altivec_expand_stvex_be): Delete.
7411 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7412 -maltivec=be support.
7413 (rs6000_gimple_fold_builtin): Ditto.
7414 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7416 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7417 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7418 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7419 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7420 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7421 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7422 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7423 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7424 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7425 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7426 anonymous split): Adjust.
7427 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7428 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7430 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7432 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7433 when --with-gxx-include-dir is also specified.
7434 * configure: Regenerate.
7436 2018-05-09 Jim Wilson <jimw@sifive.com>
7439 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7440 * config/riscv/t-withmultilib: New.
7441 * config/riscv/withmultilib.h: New.
7442 * doc/install.texi: Document RISC-V --with-multilib-list support.
7444 2018-05-09 Richard Biener <rguenther@suse.de>
7446 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7448 (vect_bb_vectorization_profitable_p): Adjust. Compute
7449 actual scalar cost using the cost vector and the add_stmt_cost
7452 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7454 PR rtl-optimization/85645
7455 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7456 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7458 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7460 PR rtl-optimization/85645
7461 * shrink-wrap.c (spread_components): Return a boolean saying if
7462 anything was changed.
7463 (try_shrink_wrapping_separate): Iterate spread_components until
7464 nothing changes anymore.
7466 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7468 PR rtl-optimization/85645
7469 * regrename.c (build_def_use): Also kill the chains that include the
7470 destination of a REG_CFA_REGISTER note.
7472 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7474 PR rtl-optimization/85645
7475 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7476 insn that has a REG_CFA_REGISTER note.
7478 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7480 * cfgexpand.c (expand_clobber): New function.
7481 (expand_gimple_stmt_1): Use it.
7482 * tree-vect-stmts.c (vect_clobber_variable): New function,
7484 (vectorizable_simd_clone_call): ...here.
7485 (vectorizable_store): Emit a clobber either side of an
7486 IFN_STORE_LANES sequence.
7487 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7489 2018-05-09 Tom de Vries <tom@codesourcery.com>
7492 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7493 (define_insn "trap_if_false"): Add exit after trap.
7495 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7497 PR rtl-optimization/85638
7498 * bb-reorder.c: Include common/common-target.h.
7499 (create_forwarder_block): New function extracted from...
7500 (fix_up_crossing_landing_pad): ...here. Rename into...
7501 (dw2_fix_up_crossing_landing_pad): ...this.
7502 (sjlj_fix_up_crossing_landing_pad): New function.
7503 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7504 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7505 from both partitions and exit the loop after one iteration.
7507 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7510 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7512 (Basic PowerPC Built-in Functions): The new name of the
7513 subsection previously known as "PowerPC Built-in Functions".
7514 (Basic PowerPC Built-in Functions Available on all Configurations):
7516 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7518 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7520 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7522 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7525 2018-05-08 Jim Wilson <jimw@sifive.com>
7527 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7528 (LD_EMUL_SUFFIX): New.
7529 (LINK_SPEC): Use it.
7531 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7533 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7535 (Basic PowerPC Built-in Functions): The new name of the
7536 subsection previously known as "PowerPC Built-in Functions".
7537 (Basic PowerPC Built-in Functions Available on all Configurations):
7539 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7541 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7543 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7545 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7548 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7551 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7552 after cmpelim optimization.
7554 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7556 * config.gcc: Support "goldmont".
7557 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7558 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7560 * config/i386/i386.c (m_GOLDMONT): Define.
7561 (processor_target_table): Add "goldmont".
7562 (PTA_GOLDMONT): Define.
7563 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7564 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7565 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7566 (fold_builtin_cpu): Add "goldmont".
7567 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7568 (ix86_option_override_internal): Add "goldmont".
7569 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7570 (processor_type): Add PROCESSOR_GOLDMONT.
7571 * config/i386/i386.md: Add CPU "glm".
7572 * config/i386/glm.md: New file.
7573 * config/i386/x86-tune.def: Add m_GOLDMONT.
7574 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7576 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7579 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7581 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7582 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7583 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7586 * config/i386/i386.c (ix86_fold_builtin): Handle
7587 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7590 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7591 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7593 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7596 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7597 (check_arch): Likewise.
7598 (check_fpu): Return the result rather than printing it.
7599 (end arch): Fix operator precedence.
7600 (end cpu): Likewise.
7601 (END): Print the result from check_fpu.
7603 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7604 Alan Hayward <alan.hayward@arm.com>
7605 David Sherwood <david.sherwood@arm.com>
7607 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7608 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7609 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7610 (*fcmuo<mode>_and): New patterns.
7612 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7614 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7615 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7616 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7617 (cmp_op, sve_imm_con): New code attributes.
7618 (SVE_COND_INT_CMP, imm_con): Delete.
7619 (cmp_op): Remove above unspecs from int attribute.
7620 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7622 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7623 comparison-specific unspecs.
7624 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7625 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7626 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7627 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7628 (*vec_fcm<cmp_op><mode>): Rename to...
7629 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7630 (*vec_fcmuo<mode>): Rename to...
7631 (*fcmuo<mode>): ...this and adjust likewise.
7632 (*pred_fcm<cmp_op><mode>): New pattern.
7633 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7634 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7636 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7638 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7639 (aarch64_emit_sve_predicated_cond): New function.
7640 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7641 (aarch64_emit_unspec_cond_or): Replace with...
7642 (aarch64_emit_sve_or_conds): ...this new function. Use
7643 aarch64_emit_sve_ptrue_op for the individual comparisons and
7644 aarch64_emit_binop to OR them together.
7645 (aarch64_emit_inverted_unspec_cond): Replace with...
7646 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7647 aarch64_emit_sve_ptrue_op for the comparison and
7648 aarch64_emit_unop to invert the result.
7649 (aarch64_expand_sve_vec_cmp_float): Update after the above
7650 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7652 2018-05-07 Nathan Sidwell <nathan@acm.org>
7654 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7655 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7656 (Backwards Compatibility): Likewise.
7658 2018-05-07 Luis Machado <luis.machado@linaro.org>
7662 2018-05-07 Luis Machado <luis.machado@linaro.org>
7664 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7665 <prefetch_dynamic_strides>: New const bool field.
7666 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7667 prefetch_dynamic_strides.
7668 (exynosm1_prefetch_tune): Likewise.
7669 (thunderxt88_prefetch_tune): Likewise.
7670 (thunderx_prefetch_tune): Likewise.
7671 (thunderx2t99_prefetch_tune): Likewise.
7672 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7674 (aarch64_override_options_internal): Update to set
7675 PARAM_PREFETCH_DYNAMIC_STRIDES.
7676 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7677 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7678 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7679 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7680 prefetch-dynamic-strides setting.
7682 2018-05-07 Luis Machado <luis.machado@linaro.org>
7684 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7685 <minimum_stride>: New const int field.
7686 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7687 minimum_stride field.
7688 (exynosm1_prefetch_tune): Likewise.
7689 (thunderxt88_prefetch_tune): Likewise.
7690 (thunderx_prefetch_tune): Likewise.
7691 (thunderx2t99_prefetch_tune): Likewise.
7692 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7693 (aarch64_override_options_internal): Update to set
7694 PARAM_PREFETCH_MINIMUM_STRIDE.
7695 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7696 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7697 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7698 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7699 stride is constant and is below the minimum stride threshold.
7701 2018-05-07 Luis Machado <luis.machado@linaro.org>
7703 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7706 2018-05-07 Luis Machado <luis.machado@linaro.org>
7708 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7709 <prefetch_dynamic_strides>: New const bool field.
7710 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7711 prefetch_dynamic_strides.
7712 (exynosm1_prefetch_tune): Likewise.
7713 (thunderxt88_prefetch_tune): Likewise.
7714 (thunderx_prefetch_tune): Likewise.
7715 (thunderx2t99_prefetch_tune): Likewise.
7716 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7718 (aarch64_override_options_internal): Update to set
7719 PARAM_PREFETCH_DYNAMIC_STRIDES.
7720 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7721 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7722 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7723 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7724 prefetch-dynamic-strides setting.
7726 2018-05-07 Luis Machado <luis.machado@linaro.org>
7728 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7729 <minimum_stride>: New const int field.
7730 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7731 minimum_stride field.
7732 (exynosm1_prefetch_tune): Likewise.
7733 (thunderxt88_prefetch_tune): Likewise.
7734 (thunderx_prefetch_tune): Likewise.
7735 (thunderx2t99_prefetch_tune): Likewise.
7736 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7737 (aarch64_override_options_internal): Update to set
7738 PARAM_PREFETCH_MINIMUM_STRIDE.
7739 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7740 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7741 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7742 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7743 stride is constant and is below the minimum stride threshold.
7745 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7748 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7749 the type is addressable. Don't force op into register if it has
7752 2018-05-05 Roland McGrath <mcgrathr@google.com>
7755 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7756 any section for which we don't know a specific type it should have,
7757 regardless of name. Previously this was done only for the exact
7758 names ".init_array", ".fini_array", and ".preinit_array".
7759 (default_elf_asm_named_section): Add comment about
7760 relationship with default_section_type_flags and SECTION_NOTYPE.
7761 (get_section): Don't consider it a type conflict if one side has
7762 SECTION_NOTYPE and the other doesn't, as long as neither has the
7763 SECTION_BSS et al used in the default_section_type_flags logic.
7765 2018-05-05 Tom de Vries <tom@codesourcery.com>
7768 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7769 (workaround_barsyncs): New function.
7770 (nvptx_reorg): Use workaround_barsyncs.
7771 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7772 (define_expand "nvptx_membar_cta"): New define_expand.
7773 (define_insn "*nvptx_membar_cta"): New insn.
7775 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7777 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7778 To improve optimization opportunities.
7779 * builtin-types.def: The new needed builtin types for the above.
7781 2018-05-04 Richard Biener <rguenther@suse.de>
7783 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7784 * gimple-ssa-store-merging.c
7785 (imm_store_chain_info::output_merged_store): Remove redundant create,
7786 release split_store vector contents on failure.
7787 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7788 scalar stmt vector on cache hit.
7790 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7792 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7794 * config.gcc (powerpc-xilinx-eabi*): Remove.
7795 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7797 (fusion_addis_mem_combo_load): Ditto.
7798 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7800 (rs6000_cpu_cpp_builtins): Ditto.
7801 * config/rs6000/rs6000-linux.c
7802 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7803 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7804 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7806 (rs6000_setup_reg_addr_masks): Ditto.
7807 (rs6000_init_hard_regno_mode_ok): Ditto.
7808 (rs6000_option_override_internal): Ditto.
7809 (legitimate_lo_sum_address_p): Ditto.
7810 (rs6000_legitimize_address): Ditto.
7811 (rs6000_legitimize_reload_address): Ditto.
7812 (rs6000_legitimate_address_p): Ditto.
7813 (abi_v4_pass_in_fpr): Ditto.
7814 (setup_incoming_varargs): Ditto.
7815 (rs6000_gimplify_va_arg): Ditto.
7816 (rs6000_split_multireg_move): Ditto.
7817 (rs6000_savres_strategy): Ditto.
7818 (rs6000_emit_prologue_components): Ditto.
7819 (rs6000_emit_epilogue_components): Ditto.
7820 (rs6000_emit_prologue): Ditto.
7821 (rs6000_emit_epilogue): Ditto.
7822 (rs6000_elf_file_end): Ditto.
7823 (rs6000_function_value): Ditto.
7824 (rs6000_libcall_value): Ditto.
7825 * config/rs6000/rs6000.h: Ditto.
7826 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7827 (TARGET_MINMAX): ... this. New.
7828 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7829 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7830 (*movsi_internal1_single): Delete.
7831 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7832 mfpu=, mxilinx-fpu): Delete.
7833 * config/rs6000/singlefp.h: Delete.
7834 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7835 * config/rs6000/t-rs6000: Ditto.
7836 * config/rs6000/t-xilinx: Delete.
7837 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7838 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7839 (VStype_simple): Delete.
7840 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7841 * config/rs6000/xfpu.h: Delete.
7842 * config/rs6000/xfpu.md: Delete.
7843 * config/rs6000/xilinx.h: Delete.
7844 * config/rs6000/xilinx.opt: Delete.
7845 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7846 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7848 2018-05-04 Tom de Vries <tom@codesourcery.com>
7851 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7854 2018-05-04 Richard Biener <rguenther@suse.de>
7857 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7858 (expand_complex_div_wide): Likewise.
7859 (expand_complex_operations_1): Likewise.
7860 (expand_complex_libcall): Preserve EH info of the original stmt.
7861 (tree_lower_complex): Handle removed blocks.
7862 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7863 on complex multiplication and division libcall builtins.
7865 2018-05-04 Richard Biener <rguenther@suse.de>
7868 * fold-const.c (negate_expr_p): Restrict negation of operand
7869 zero of a division to when we know that can happen without
7871 (fold_negate_expr_1): Likewise.
7873 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7876 * real.h (real_nextafter): Declare.
7877 * real.c (real_nextafter): New function.
7878 * fold-const-call.c (fold_const_nextafter): New function.
7879 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7880 CASE_CFN_NEXTTOWARD.
7881 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7882 even when arg1_mode is different from arg0_mode.
7884 2018-05-03 Nathan Sidwell <nathan@acm.org>
7886 * doc/extend.texi (Deprecated Features): Remove
7888 (Backwards Compatibility): Likewise.
7889 * doc/invoke.texi (C++ Language Options): Likewise.
7890 (C++ Dialect Options): Likewise.
7892 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7895 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7896 _mm512_mask_mullox_epi64): New intrinsics.
7898 2018-05-03 Tom de Vries <tom@codesourcery.com>
7901 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7902 dump files): Add offload-tree.
7904 2018-05-03 Richard Biener <rguenther@suse.de>
7906 PR tree-optimization/85615
7907 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7908 to loops not nested in BBs loop father to avoid creating multi-entry
7911 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7913 PR tree-optimization/70291
7914 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7915 arguments. Change return type to tree. Emit libcall as a new
7916 statement rather than replacing existing one when inplace_p is true.
7917 (expand_complex_multiplication_components): New function.
7918 (expand_complex_multiplication): Expand floating-point complex
7919 multiplication using the above.
7920 (expand_complex_division): Rename inner_type parameter to type.
7921 Update expand_complex_libcall call-site.
7922 (expand_complex_operations_1): Update expand_complex_multiplication
7923 and expand_complex_division call-sites.
7925 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7928 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7929 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7930 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7931 the highest significant bit of the shift count mask is clear. In
7932 check whether and[sq]i3 is needed verify that all significant bits
7933 of the shift count other than the highest are set.
7935 2018-05-02 Tom de Vries <tom@codesourcery.com>
7938 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7939 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7940 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7941 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7942 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7943 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7944 __builtin_goacc_parlevel_size.
7946 2018-05-02 Richard Biener <rguenther@suse.de>
7948 PR tree-optimization/85597
7949 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7950 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7953 2018-05-02 Tom de Vries <tom@codesourcery.com>
7956 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7957 dump files): Add ltrans-tree.
7959 2018-05-02 Tom de Vries <tom@codesourcery.com>
7962 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7963 dump files): Add wpa-ipa.
7965 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7967 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7968 powerpc*-*-linux*paired* target.
7969 * config/rs6000/750cl.h: Delete.
7970 * config/rs6000/paired.h: Delete.
7971 * config/rs6000/paired.md: Delete.
7972 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7974 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7975 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7976 comment. Remove paired float support.
7977 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7978 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7980 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7981 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7983 * config/rs6000/rs6000.c: Remove paired float support.
7984 (paired_expand_vector_init, paired_expand_vector_move,
7985 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7986 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7987 paired_expand_builtin, paired_expand_predicate_builtin,
7988 paired_init_builtins): Delete.
7989 * config/rs6000/rs6000.h: Remove paired float support.
7990 * config/rs6000/rs6000.md: Remove paired float support.
7991 (move_from_CR_ov_bit): Delete.
7992 * config/rs6000/rs6000.opt (mpaired): Delete.
7993 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7994 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7996 2018-05-02 Richard Biener <rguenther@suse.de>
7999 * gimplify.c (gimplify_save_expr): When in SSA form allow
8000 SAVE_EXPRs to compute to SSA vars.
8002 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8005 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8006 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8007 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8008 clobber operands[2], instead use a new pseudo. Formatting fixes.
8010 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8012 PR tree-optimization/85586
8013 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8014 exit early for statements in the same group if the accesses are
8017 2018-05-02 Tom de Vries <tom@codesourcery.com>
8020 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8023 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8025 PR tree-optimization/85143
8026 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8028 2018-05-01 Tom de Vries <tom@codesourcery.com>
8031 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8032 not found" error message.
8034 2018-05-01 Tom de Vries <tom@codesourcery.com>
8037 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8038 * vec.c (test_ordered_remove_if): New function.
8039 (vec_c_tests): Call test_ordered_remove_if.
8040 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8041 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8042 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8043 VEC_ORDERED_REMOVE_IF.
8045 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8047 PR tree-optimization/82665
8048 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8049 pointer subtraction where arguments come from a memchr call.
8051 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8053 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8054 --push-state --as-needed and --pop-state instead of --as-needed and
8055 --no-as-needed if ld supports it.
8056 * configure: Regenerated.
8059 * doc/install.texi2html: Replace _002d with - and _002a with * in
8060 generated html files using sed.
8062 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8065 * gcc-rich-location.c (blank_line_before_p): New function.
8066 (use_new_line): New function.
8067 (gcc_rich_location::add_fixit_insert_formatted): New function.
8068 * gcc-rich-location.h
8069 (gcc_rich_location::add_fixit_insert_formatted): New function.
8071 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8073 * selftest.c (assert_streq): Rename "expected" and "actual" to
8074 "val1" and "val2". Extend NULL-handling to cover both inputs
8075 symmetrically, while still requiring both to be non-NULL for a pass.
8076 * selftest.h (assert_streq): Rename "expected" and "actual" to
8078 (ASSERT_EQ): Likewise.
8079 (ASSERT_EQ_AT): Likewise.
8080 (ASSERT_KNOWN_EQ): Likewise.
8081 (ASSERT_KNOWN_EQ_AT): Likewise.
8082 (ASSERT_NE): Likewise.
8083 (ASSERT_MAYBE_NE): Likewise.
8084 (ASSERT_MAYBE_NE_AT): Likewise.
8085 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8086 the assertion to pass.
8087 (ASSERT_STREQ_AT): Likewise.
8089 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8091 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8092 interaction with -pie.
8094 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8096 * selftest.h: Fix alphabetization of per-source-file selftest
8099 2018-04-30 Jason Merrill <jason@redhat.com>
8101 PR c++/61982 - dead stores to destroyed objects.
8102 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8105 2018-04-30 Jason Merrill <jason@redhat.com>
8107 * tree.c (build_clobber): New.
8108 * tree.h: Declare it.
8109 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8111 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8113 * diagnostic-show-locus.c (layout::layout): Update for
8114 location_get_source_line returning a char_span.
8115 (struct char_span): Move to input.h.
8116 (struct correction): Update for fields in char_span becoming
8118 (struct source_line): Update for location_get_source_line
8119 returning a char_span.
8120 (layout::print_line): Likewise.
8121 * edit-context.c (edited_file::print_content): Likewise.
8122 (edited_file::print_diff_hunk): Likewise.
8123 (edited_file::print_run_of_changed_lines): Likewise.
8124 (edited_file::get_num_lines): Likewise.
8125 (edited_line::edited_line): Likewise.
8126 * final.c (asm_show_source): Likewise.
8127 * input.c (location_get_source_line): Convert return type
8128 from const char * to char_span, losing the final "line_len"
8130 (dump_location_info): Update for the above.
8131 (get_substring_ranges_for_loc): Likewise. Use a char_span
8132 when handling the literal within the line.
8133 (test_reading_source_line): Update for location_get_source_line
8134 returning a char_span.
8135 * input.h (class char_span): Move here from
8136 diagnostic-show-locus.c, converting from a struct to a class.
8137 Make data members private.
8138 (char_span::operator bool): New.
8139 (char_span::length): New.
8140 (char_span::get_buffer): New.
8141 (char_span::operator[]): New.
8142 (char_span::subspan): Make const.
8143 (char_span::xstrdup): New.
8144 (location_get_source_line): Convert return type from const char *
8145 to char_span, losing the final "line_size" param.
8147 2018-04-30 Jan Hubicka <jh@suse.cz>
8149 * lto-wrapper.c (ltrans_priorities): New static var.
8150 (cmp_priority): New.
8151 (run_gcc): Read priorities and if doing parallel build order
8152 the Makefile by them.
8154 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8156 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8158 2018-04-30 Richard Biener <rguenther@suse.de>
8160 * tree-cfg.c (verify_address): Remove base argument, add
8161 flag whether to check TREE_ADDRESSABLE and do that.
8162 (verify_expr): Remove.
8163 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8164 (verify_gimple_assign_single): Likewise.
8165 (verify_gimple_switch): Likewise.
8166 (verify_expr_location_1): Dereference tp once. Add (disabled)
8167 piece from verify_expr.
8168 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8170 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8172 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8174 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8176 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8177 (small_data_pattern): Likewise.
8178 (arc_rewrite_small_data): Likewise.
8179 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8180 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8181 (get_symbol_alignment): New function.
8182 (legitimate_small_data_address_p): Likewise.
8183 (legitimate_scaled_address): Update, call
8184 legitimate_small_data_address_p.
8185 (output_sdata): New static variable.
8186 (arc_print_operand): Update how we handle small data operands.
8187 (arc_print_operand_address): Likewise.
8188 (arc_legitimate_address_p): Update, use
8189 legitimate_small_data_address_p.
8190 (arc_rewrite_small_data_p): Remove.
8191 (arc_rewrite_small_data_1): Likewise.
8192 (arc_rewrite_small_data): Likewise.
8193 (small_data_pattern): Likewise.
8194 (compact_sda_memory_operand): Update to use
8195 legitimate_small_data_address_p and get_symbol_alignment.
8196 (prepare_move_operands): Don't rewite sdata pattern.
8197 (prepare_extend_operands): Remove.
8198 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8200 (zero_extendqisi2): Likewise.
8201 (zero_extendhisi2): Likewise.
8202 (extendqihi2): Likewise.
8203 (extendqisi2): Likewise.
8204 (extendhisi2): Likewise.
8208 * config/arc/constraints.md (Usd): Change it to memory constraint.
8210 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8212 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8213 as source of std instructions.
8214 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8215 allow 6-bit constants as source for store instructions.
8216 (movdi_insn): Update instruction pattern to allow 6-bit constants
8217 as source for store instructions.
8219 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8221 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8223 2018-04-30 Nathan Sidwell <nathan@acm.org>
8224 Sandra Loosemore <sandra@codesourcery.com>
8226 * dumpfile.c (dump_open): Allow '-' for stdout.
8227 * doc/invoke.texi (Developer Options): Document dump filename
8228 determination early. Document stdin/stdout selection.
8230 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8232 Microblaze Target: PIC data text relative
8234 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8235 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8237 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8238 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8239 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8240 New addressing mode for data-text relative position indepenedent code.
8241 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8242 'ADDRESS_SYMBOLIC_TXT_REL'.
8243 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8244 (microblaze_legitimate_pic_operand): Exclude function calls from
8245 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8246 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8248 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8249 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8250 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8251 for 'address + offset'.
8252 (microblaze_expand_prologue): Add new function prologue call for
8254 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8255 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8256 table in case of TARGET_PIC_DATA_TEXT_REL.
8257 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8258 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8259 Add new macros 'UNSPEC_TEXT',
8260 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8261 + exclude function calls from 'UNSPEC_PLT' in case of data text
8263 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8264 new target hook for generating address diff vector tables in case of
8266 * doc/tm.texi : Regenerate.
8267 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8268 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8269 of addr diff vector generation.
8270 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8271 target hook definition.
8272 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8273 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8274 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8275 option for data text relative.
8277 2018-04-30 Richard Biener <rguenther@suse.de>
8279 * tree-chrec.h (evolution_function_is_constant_p): Remove
8281 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8283 2018-04-30 Richard Biener <rguenther@suse.de>
8286 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8288 2018-04-30 Richard Biener <rguenther@suse.de>
8290 PR tree-optimization/28364
8291 PR tree-optimization/85275
8292 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8293 copying first exit test.
8295 2018-04-28 Mark Wielaard <mark@klomp.org>
8297 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8299 (dwarf_AT): Handle DW_AT_addr_base.
8300 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8302 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8305 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8306 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8307 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8308 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8310 2018-04-28 Richard Biener <rguenther@suse.de>
8312 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8313 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8314 to reflect use. Only add interesting stmts.
8316 2018-04-27 Martin Jambor <mjambor@suse.cz>
8319 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8320 the jump function allows for passing through aggregate values.
8322 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8324 * input.h (in_system_header_at): Convert from macro to inline
8326 (from_macro_expansion_at): Likewise.
8327 (from_macro_definition_at): Likewise.
8329 2018-04-27 Jeff Law <law@redhat.com>
8331 * config.gcc: Mark tile* targets as deprecated/obsolete.
8333 2018-04-27 Richard Biener <rguenther@suse.de>
8335 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8338 2018-04-27 Richard Biener <rguenther@suse.de>
8340 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8342 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8344 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8345 with Yd constraint. Set "preferred_for_speed" attribute from
8346 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8348 (*movdi_internal): Ditto.
8349 (movti_interunit splitters): Remove
8350 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8351 (movdi_interunit splitters): Ditto.
8352 * config/i386/constraints.md (Ye): Remove.
8353 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8355 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8358 * config/aarch64/constraints.md (Usg): Limit to 31.
8361 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8363 PR tree-optimization/85529
8364 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8365 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8366 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8367 zero extension or masking of the MSB bit.
8368 (optimize_range_tests): Add FIRST_BB argument, pass it through
8369 to optimize_range_tests_var_bound.
8370 (maybe_optimize_range_tests, reassociate_bb): Adjust
8371 optimize_range_tests callers.
8373 2018-04-26 Richard Biener <rguenther@suse.de>
8374 Jakub Jelinek <jakub@redhat.com>
8376 * cgraph.h (symbol_table): Just declare debug method here.
8377 * symtab.c (symbol_table::debug): Define.
8379 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8381 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8383 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8385 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8386 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8387 (*movdi_internal): Substitute Yi and Yj constraint with x
8388 and Ym and Yn constraint with y constraint. Update "isa"
8389 attribute and set "preferred_for_speed" attribute from
8390 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8391 (*movsi_internal): Ditto.
8392 (*movdf_internal): Ditto.
8393 (*movsf_internal): Ditto.
8394 (*zero_extendsidi2): Ditto.
8395 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8396 (sse2_loadld): Ditto.
8397 (*vec_extract<ssevecmodelower>_0): Ditto.
8398 (*vec_extractv4si_0_zext_sse4): Ditto.
8399 (vec_concatv2di): Ditto.
8400 (*vec_dup<mode>): Ditto.
8401 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8402 * config/i386/constraints.md (Yi): Remove.
8407 2018-04-26 Nathan Sidwell <nathan@acm.org>
8409 * dumpfile.c (dump_open): New.
8410 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8411 (dump_finish): Detect stdio/stderr by value not name.
8413 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8415 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8417 2018-04-26 Tom de Vries <tom@codesourcery.com>
8420 * config/nvptx/nvptx.c (verify_neutering_jumps)
8421 (verify_neutering_labels): New function
8422 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8424 2018-04-26 Tom de Vries <tom@codesourcery.com>
8427 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8428 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8431 2018-04-26 Richard Biener <rguenther@suse.de>
8432 Tom de Vries <tom@codesourcery.com>
8435 * lto-streamer-out.c (output_function): Fixup loops if required to match
8436 discovery done in the reader.
8438 2018-04-26 Richard Biener <rguenther@suse.de>
8440 PR tree-optimization/85116
8441 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8442 have a loop exit from the single latch predecessor. Remove
8443 case of header with just condition.
8444 (ch_base::copy_headers): Exclude infinite loops from any
8446 (pass_ch::execute): Record exits.
8448 2018-04-26 Richard Biener <rguenther@suse.de>
8450 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8451 prologue cost vector and pass it to vect_get_load_cost.
8452 (vect_get_peeling_costs_all_drs): Likewise.
8453 (vect_peeling_hash_get_lowest_cost): Likewise.
8454 (vect_enhance_data_refs_alignment): Likewise.
8456 2018-04-26 Richard Biener <rguenther@suse.de>
8459 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8460 checking of integer<->pointer conversions.
8461 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8462 sign-/zero-extending pointer types.
8463 (expand_omp_for_static_chunk): Likewise.
8465 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8466 Jean Lee <xiaoyur347@gmail.com>
8468 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8469 (TARGET_ASAN_SHADOW_OFFSET): Define.
8470 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8471 true for -fsanitize=address.
8473 2018-04-25 Mark Wielaard <mark@klomp.org>
8475 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8478 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8480 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8481 than "alu", remove explicit "memory" and "imm_disp" attributes.
8482 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8485 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8486 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8487 gen_lowpart_no_emit.
8489 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8492 * config/i386/i386.c (ix86_expand_builtin): Change memory
8493 operand to XI, extend p0 to Pmode.
8494 * config/i386/i386.md: Change unspec volatile and operand
8495 1 mode to XI, change operand 0 mode to P.
8497 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8499 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8500 GET_MODE_MASK before any checking.
8501 (nds32_can_use_bset_p): Likewise.
8502 (nds32_can_use_btgl_p): Likewise.
8504 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8506 * config/nds32/nds32-doubleword.md: New define_split pattern for
8507 illegal register number.
8509 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8511 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8513 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8515 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8517 2018-04-25 Richard Biener <rguenther@suse.de>
8519 * lto-streamer.h (LTO_major_version): Bump to 8.
8521 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8523 * BASE-VER: Set to 9.0.0.
8525 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8527 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8528 in __abskf2 and __powikf2.
8530 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8533 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8534 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8535 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8536 Use the above on operand 2. Reindent.
8537 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8539 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8542 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8544 * config/i386/i386.opt (mcet): Removed.
8545 * doc/install.texi: Remove -mcet documentation.
8546 * doc/invoke.texi: Likewise.
8548 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8551 * doc/install.texi: Remove -mcet from bootstrap-cet.
8553 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8556 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8557 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8561 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8562 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8563 containing a CONST_VECTOR.
8565 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8567 * doc/install.texi: Update newlib dependency for nvptx.
8569 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8572 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8573 instead of INTVAL when shifting x left.
8575 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8577 PR tree-optimization/85478
8578 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8579 vect_grouped_store_supported for single element vectors.
8581 2018-04-24 Richard Biener <rguenther@suse.de>
8584 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8585 load cost increase to the case of non-constant step.
8587 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8590 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8591 destination if any_malformed_asm.
8593 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8596 * expr.c (store_field): In the bitfield case, if the value comes from
8597 a function call and is returned in registers by means of a PARALLEL,
8598 do not change the mode of the temporary unless BLKmode and VOIDmode.
8600 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8602 PR rtl-optimization/85423
8603 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8604 dependencies to debug insns when the previous insn is non-debug.
8606 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8608 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8609 enums into a single definition.
8610 (fls): Fix predicates and printing.
8613 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8615 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8616 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8617 and short u6 immediate.
8618 (check_if_valid_sleep_operand): Remove.
8619 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8621 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8623 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8624 flag_always_save_lp condition.
8625 * config/nds32/nds32.opt (malways-save-lp): New option.
8627 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8629 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8630 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8631 * config/nds32/nds32.h
8632 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8633 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8635 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8637 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8638 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8640 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8641 Chung-Ju Wu <jasonwucj@gmail.com>
8643 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8645 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8646 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8648 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8650 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8652 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8654 * config/nds32/nds32-protos.h (nds32_data_alignment,
8655 nds32_local_alignment): Declare.
8656 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8657 nds32_local_alignment): New functions.
8658 (TARGET_CONSTANT_ALIGNMENT): Define.
8659 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8661 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8663 * config/nds32/nds32.c
8664 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8665 (TARGET_MODES_TIEABLE_P): Likewise.
8667 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8669 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8672 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8673 Chung-Ju Wu <jasonwucj@gmail.com>
8675 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8676 for unaligned access.
8677 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8679 * config/nds32/nds32-intrinsic.md: Likewise.
8680 * config/nds32/nds32_intrinsic.h: Likewise.
8681 * config/nds32/nds32.h (nds32_builtins): Likewise.
8682 * config/nds32/nds32.opt (munaligned-access): New option.
8683 * config/nds32/nds32.c (nds32_asm_file_start): Display
8684 flag_unaligned_access status.
8686 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8688 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8689 -mno-relax is present.
8690 * config/riscv/linux.h (LINK_SPEC): Ditto.
8692 2018-04-20 Martin Sebor <msebor@redhat.com>
8695 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8697 (gimple_fold_builtin_stxcpy_chk): Same.
8698 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8700 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8703 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8704 __powikf2 when long double is IEEE 128-bit.
8706 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8708 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8709 step to make sure stack always aligned.
8711 2018-04-20 Carl Love <cel@us.ibm.com>
8714 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8715 size check for arg0.
8717 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8718 Tom de Vries <tom@codesourcery.com>
8721 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8722 Emit insns for calls too.
8723 (nvptx_find_par): Always look for worker-level predecessor insn.
8724 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8726 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8727 (nvptx_process_pars): Propagate frames for calls.
8729 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8732 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8734 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8735 (ix86_handle_option): Don't handle OPT_mibt.
8736 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8738 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8740 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8741 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8742 (ix86_target_macros): Define __CET__ with flag_cf_protection
8743 for -fcf-protection.
8744 * config/i386/i386.c (isa2_opts): Remove -mibt.
8745 * config/i386/i386.h (TARGET_IBT): Removed.
8746 (TARGET_IBT_P): Likewise.
8747 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8748 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8749 * config/i386/i386.opt (mcet): Update help message.
8752 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8753 -mcet as an alias for -mshstk.
8755 2018-04-20 Richard Biener <rguenther@suse.de>
8758 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8759 complexity by forcing a single use of the multiply operand.
8761 2018-04-20 Martin Jambor <mjambor@suse.cz>
8764 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8765 recursion dependency to only apply to non-clones.
8767 2018-04-20 Martin Jambor <mjambor@suse.cz>
8770 * ipa-cp.c (create_specialized_node): Check that clones of
8771 self-recursive edges exist during IPA-CP.
8773 2018-04-19 Toon Moene <toon@moene.org>
8775 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8778 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8780 PR tree-optimization/85467
8781 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8782 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8783 VECTOR_CST element to type.
8785 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8788 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8789 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8790 (builtin_longjmp): Likewise.
8791 (save_stack_nonlocal): New pattern.
8792 (restore_stack_nonlocal): Likewise.
8794 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8797 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8798 Replace ASM_OUTPUT_LABEL with fprintf.
8800 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8803 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8804 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8805 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8806 define __IBT__ and __SHSTK__ for -fcf-protection.
8807 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8809 (ix86_trampoline_init): Likewise.
8810 (x86_output_mi_thunk): Likewise.
8811 (ix86_notrack_prefixed_insn_p): Likewise.
8812 (ix86_option_override_internal): Don't disallow -fcf-protection.
8813 * config/i386/i386.md (rdssp<mode>): Also enable for
8815 (incssp<mode>): Likewise.
8816 (nop_endbr): Likewise.
8817 * config/i386/i386.opt (mcet): Change help message to built-in
8821 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8822 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8823 enable CET built-in functions.
8825 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8827 * common/config/i386/i386-common.c
8828 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8829 OPTION_MASK_ISA_MOVDIRI_UNSET,
8830 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8831 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8832 * config.gcc (movdirintrin.h): New header.
8833 * config/i386/cpuid.h (bit_MOVDIRI,
8834 bit_MOVDIR64B): New bits.
8835 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8837 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8838 (VOID, PVOID, PCVOID)): New function types.
8839 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8840 __builtin_ia32_directstoreu_u64,
8841 __builtin_ia32_movdir64b): New builtins.
8842 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8843 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8845 (ix86_valid_target_attribute_inner_p): Ditto.
8846 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8847 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8848 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8849 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8850 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8851 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8852 (movdiri<mode>, movdir64b_<mode>): New.
8853 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8854 * config/i386/immintrin.h: Include movdirintrin.h.
8855 * config/i386/movdirintrin.h: New file.
8856 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8858 2018-04-19 Richard Biener <rguenther@suse.de>
8861 * cfg.c (clear_bb_flags): When loop state says we have
8862 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8864 2018-04-19 Richard Biener <rguenther@suse.de>
8866 PR tree-optimization/84737
8867 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8868 copying restrict info.
8869 (vect_setup_realignment): Use it.
8870 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8871 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8872 the first DR to all generated stores.
8873 (vectorizable_load): Likewise for loads.
8875 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8877 PR tree-optimization/85446
8878 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8879 the integral and pointer types to have the same precision.
8881 * doc/install.texi: Document --disable-cet being the default and
8884 2018-04-18 Martin Liska <mliska@suse.cz>
8886 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8889 2018-04-18 Martin Liska <mliska@suse.cz>
8892 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8895 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8896 arguments if they are comparable.
8898 2018-04-18 Martin Liska <mliska@suse.cz>
8901 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8904 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8907 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8910 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8911 ENDBR after calling __morestack.
8913 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8916 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8917 by using gcc_base_ver to generate a gcc_driver_version, and use
8918 it when generating GCC_DRIVER_NAME.
8919 * configure: Regenerate.
8921 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8924 * config.gcc: Obsolete powerpc*-*-*spe*.
8926 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8929 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8930 (stabstr_D): Change type of unum from unsigned int to
8931 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8934 2018-04-17 Jim Wilson <jimw@sifive.com>
8937 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8938 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8939 Set arg_pointer_offset after using pretend_args_size.
8941 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8943 PR rtl-optimization/85431
8944 * dse.c (record_store): Ignore zero width stores.
8947 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8948 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8949 __builtin_stack_restore rather than after it.
8950 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8951 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8952 argument instead of virtual_dynamic_stack_rtx.
8954 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8956 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8958 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8959 Add note to error message to explain internal mapping of overloaded
8960 built-in function name to non-overloaded built-in function name.
8961 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8964 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8967 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8968 where the inputs overlap with the output.
8970 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8972 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8973 (=v, v) alternative and explicit "memory" attribute.
8974 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8975 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8977 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8978 "sselog1" type instead of "sselog".
8979 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8980 "sselog". Remove explicit "memory" attribute.
8981 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8982 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8984 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8985 "sselog1" type instead of "sselog". Remove explicit "memory"
8987 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8988 use "sselog1" type instead of "sselog". Remove explicit "memory"
8990 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8991 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8993 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8994 "sselog1" type instead of "sselog". Remove explicit "memory"
8996 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8997 use "sselog1" type instead of "sselog". Remove explicit "memory"
9001 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9004 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9007 2018-04-17 Martin Jambor <mjambor@suse.cz>
9010 * ipa-cp.c (create_specialized_node): Call
9011 expand_all_artificial_thunks if necessary.
9013 2018-04-17 Martin Liska <mliska@suse.cz>
9016 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9017 in message, remote space in between '_G' and '('.
9019 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9022 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9023 avx512f_vmcmp<mode>3<round_saeonly_name>,
9024 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9025 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9026 avx512f_rndscale<mode><round_saeonly_name>,
9027 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9028 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9029 Use %<iptr>2 instead of %2 for -masm=intel.
9030 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9031 avx512f_vcvttss2usi<round_saeonly_name>,
9032 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9034 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9035 avx512f_vcvttsd2usi<round_saeonly_name>,
9036 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9037 Use %q1 instead of %1 for -masm=intel.
9038 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9039 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9040 of %3 for -masm=intel.
9041 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9043 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9045 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9047 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9048 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9050 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9052 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9053 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9054 %g1 and one with %0 and %1.
9055 (avx512er_vmrcp28<mode><round_saeonly_name>,
9056 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9058 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9059 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9060 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9061 of %0 and %{%4%} for -masm=intel.
9062 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9063 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9064 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9065 order of %0 and %{%5%}%{z%} for -masm=intel.
9067 2018-04-17 Jan Hubicka <jh@suse.cz>
9070 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9072 2018-04-17 Martin Liska <mliska@suse.cz>
9075 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9076 for target_clone error message. Make default implementation
9077 clone to be a local declaration.
9078 (separate_attrs): Add new argument and check for an empty
9080 (expand_target_clones): Handle it.
9081 (ipa_target_clone): Make redirection just for target_clones
9084 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9085 Tom de Vries <tom@codesourcery.com>
9088 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9089 tiled basic blocks without omp continue statements.
9091 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9094 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9095 vec_extract expression as having side effects to make sure it gets
9098 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9101 * config/i386/i386.c (get_builtin_code_for_version): Check
9104 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9107 * gcc/config.gcc: Support "skylake".
9108 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9110 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9111 (processor_target_table): Add "skylake".
9112 (ix86_option_override_internal): Add "skylake".
9113 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9114 PROCESSOR_CANNONLAKE.
9115 (get_builtin_code_for_version): Fix priority for
9116 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9117 PROCESSOR_SKYLAKE-AVX512.
9118 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9119 (processor_type): Add PROCESSOR_SKYLAKE.
9121 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9122 Jason Merrill <jason@redhat.com>
9125 * convert.c (convert_to_integer_1): Use direct recursion for
9126 enumeral types and types with a precision less than the number
9127 of bits in their mode.
9129 2018-04-16 Julia Koval <julia.koval@intel.com>
9132 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9133 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9135 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9138 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9139 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9140 and -mno-direct-move.
9142 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9145 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9146 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9147 Ensure negative shifts result in {0}.
9149 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9151 PR rtl-optimization/79916
9152 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9153 regs (if any) to define how to gnerate SD moves when LRA is in
9156 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9158 PR rtl-optimization/85393
9159 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9160 * except.c (expand_dw2_landing_pad_for_region): Make static.
9161 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9162 a label and unconditional jump to old_bb, rather than
9163 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9166 PR rtl-optimization/85376
9167 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9168 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9169 instead of a specific value.
9171 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9172 Bin Cheng <bin.cheng@arm.com>
9174 PR tree-optimization/82965
9175 PR tree-optimization/83991
9176 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9177 by_profile_only parameter.
9178 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9179 information if the loop was predicted to iterate too many times.
9180 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9182 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9185 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9188 2018-04-13 Martin Liska <mliska@suse.cz>
9189 Jakub Jelinek <jakub@redhat.com>
9192 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9193 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9194 * builtins.c (expand_builtin_memory_copy_args): Use
9195 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9196 handle dest_addr == pc_rtx.
9198 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9201 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9202 asked to not generate direct moves.
9203 (fix_trunc<mode>si2_stfiwx): Similar.
9204 (fix_trunc<mode>si2_internal): Similar.
9206 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9209 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9210 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9211 lookup if dest in some wider mode is known to be const0_rtx and
9212 if so, record permanent equivalence for it to be ZERO_EXTEND of
9213 the narrower mode destination.
9215 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9217 * lto-streamer-out.c (output_function): Revert 259346.
9218 * omp-expand.c (expand_oacc_for): Likewise.
9220 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9222 PR rtl-optimization/85354
9223 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9224 * sel-sched.c (sel_global_init): ... here.
9226 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9229 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9230 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9231 mode for PE-COFF targets.
9232 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9233 (i386_pe_asm_lto_end): Likewise.
9234 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9235 (TARGET_ASM_LTO_END): Likewise.
9236 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9237 (i386_pe_asm_lto_start): New function.
9238 (i386_pe_asm_lto_end): Likewise.
9240 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9241 Richard Biener <rguenther@suse.de>
9244 * lto-streamer-out.c (output_function): Fix CFG loop state before
9246 * omp-expand.c (expand_oacc_for): Handle calls to internal
9247 functions like regular functions.
9249 2018-04-12 Richard Biener <rguenther@suse.de>
9252 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9253 for the early LTO debug to properly generate references to it
9254 during DIE emission. Do not re-use that for the skeleton for
9256 (dwarf2out_early_finish): Likewise.
9258 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9261 * config/i386/sse.md
9262 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9263 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9264 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9265 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9266 and output is a reg, avoid creating invalid lowpart subreg, but
9267 instead split into a 512-bit move. Don't split if not AVX512VL,
9268 input is xmm16+ reg and output is a mem.
9269 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9270 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9271 xmm16+ reg and output is a mem.
9273 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9275 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9276 also for flag_dwarf2_cfi_asm.
9278 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9280 PR rtl-optimization/85342
9281 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9282 a bool scalar var inside of the loop instead. Don't try to update
9283 recog_data.operand after failed apply_change_group.
9285 2018-04-12 Tom de Vries <tom@codesourcery.com>
9288 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9289 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9290 array with flexible array member as array without given dimension.
9291 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9292 argument for undefined param to true.
9294 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9297 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9298 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9299 from PowerPC section.
9300 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9301 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9302 help text that is too long.
9303 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9304 help text that is too long.
9305 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9306 help text that is too long.
9308 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9310 * config/alpha/alpha.md (stack_probe_internal): Rename
9311 from "probe_stack". Update all callers.
9313 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9315 PR rtl-optimization/84566
9316 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9317 sched_macro_fuse_insns.
9319 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9322 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9323 (compute_block_dependences): ... from here.
9325 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9327 PR tree-optimization/85331
9328 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9329 from int to HOST_WIDE_INT.
9331 2018-04-11 Martin Jambor <mjambor@suse.cz>
9334 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9335 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9336 not the same as the source val.
9337 (cgraph_edge_brings_value_p): New parameter.
9338 (gather_edges_for_value): Pass destination value to
9339 cgraph_edge_brings_value_p.
9340 (perhaps_add_new_callers): Likewise.
9341 (get_info_about_necessary_edges): Likewise and exclude values brought
9342 only by self-recursive edges.
9343 (create_specialized_node): Redirect only clones of self-calling edges.
9344 (+self_recursive_pass_through_p): New function.
9345 (find_more_scalar_values_for_callers_subset): Use it.
9346 (find_aggregate_values_for_callers_subset): Likewise.
9347 (known_aggs_to_agg_replacement_list): Removed.
9348 (decide_whether_version_node): Re-calculate known constants for all
9349 remaining context clones.
9351 2018-04-11 Richard Biener <rguenther@suse.de>
9354 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9355 from early DWARF output.
9356 (dwarf2out_early_finish): Output line info unconditionally into
9357 early DWARF and add reference to it.
9359 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9362 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9363 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9364 other than V2DFmode using iptr mode attribute.
9365 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9367 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9369 PR rtl-optimization/84659
9370 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9372 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9375 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9377 (output_loc_list): Pass address of a dummy size variable even in the
9378 locview handling loop.
9379 (index_location_lists): Add comment on why skip_loc_list_entry can't
9382 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9385 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9388 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9391 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9392 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9393 and -mstring-compare-inline-limit.
9395 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9398 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9399 for stack clash protection in a register whenever we need it to be in
9402 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9404 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9405 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9407 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9410 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9412 (mlong-double-): Ditto.
9413 * config/rs6000/sysv4.opt (msdata=): Ditto.
9414 (mtls-size=): Ditto.
9416 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9418 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9419 erroneous entries for
9420 "vector int vec_ldl (int, long int *)", and
9421 "vector unsigned int vec_ldl (int, unsigned long int *)".
9422 Add comments and entries for
9423 "vector bool char vec_ldl (int, bool char *)",
9424 "vector bool short vec_ldl (int, bool short *)",
9425 "vector bool int vec_ldl (int, bool int *)",
9426 "vector bool long long vec_ldl (int, bool long long *)",
9427 "vector pixel vec_ldl (int, pixel *)",
9428 "vector long long vec_ldl (int, long long *)",
9429 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9430 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9431 type tree bool_long_long_type_node and correct definition of
9432 bool_V2DI_type_node to make reference to this new type tree.
9433 (rs6000_mangle_type): Replace erroneous reference to
9434 bool_long_type_node with bool_long_long_type_node.
9435 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9436 comments to emphasize sign distinctions for char and int types and
9437 replace RS6000_BTI_bool_long constant with
9438 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9439 use of RS6000_BTI_pixel.
9440 (bool_long_type_node): Remove this macro definition.
9441 (bool_long_long_type_node): New macro definition
9443 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9445 PR rtl-optimization/85300
9446 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9447 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9448 simplify_unary_operation fails.
9450 2018-04-10 Martin Liska <mliska@suse.cz>
9452 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9453 cgraph_edge and ipa_ref.
9455 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9459 * config/i386/sse.md
9460 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9461 computation of the VEC_MERGE selector from mask.
9462 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9463 Fix decoding of the VEC_MERGE selector into mask.
9465 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9467 PR tree-optimization/85286
9468 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9470 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9472 * final.c (final_1): Set insn_last_address as well as
9473 insn_current_address.
9475 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9478 * explow.c (emit_stack_probe): Call validize_mem on memory location
9479 before passing it to gen_probe_stack. Create address operand and
9480 legitimize it for the probe_stack_address case.
9482 2018-04-09 Jan Hubicka <jh@suse.cz>
9485 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9486 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9487 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9489 (free_lang_data): Rebuild type inheritance graph.
9491 2018-04-09 Martin Sebor <msebor@redhat.com>
9493 * invoke.texi (-finline-small-functions): Mention other optimization
9495 (-findirect-inlining, -fpartial-inlining): Same.
9496 (-finline-functions-called-once): Same.
9497 (-freorder-blocks-and-partition): Same.
9499 2018-04-09 Jan Hubicka <jh@suse.cz>
9502 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9503 jumps; choose last target that matches the criteria (i.e.
9504 no partition changes for non-crossing jumps).
9505 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9506 support for redirecting crossing jumps to non-crossing.
9508 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9510 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9511 also for naked functions.
9513 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9515 * config/arc/arc.md (add_shift): New pattern.
9516 (add_shift2): Likewise.
9517 (sub_shift): Likewise.
9518 (sub_shift_cmp0_noout): Likewise.
9519 (compare_si_ashiftsi): Likewise.
9520 (xbfu_cmp0_noout): New combine pattern.
9521 (xbfu_cmp0"): Likewise.
9522 (movsi_set_cc_insn): Place the predicable variant first.
9523 (commutative_binary_cmp0_noout): Remove clobber.
9524 (commutative_binary_cmp0): New pattern.
9525 (noncommutative_binary_cmp0): Likewise.
9526 (noncommutative_binary_cmp0_noout): Likewise.
9527 (noncommutative_binary_comparison_result_used): Removed.
9528 (rsub_cmp0): New pattern.
9529 (rsub_cmp0_noout): Likewise.
9530 (extzvsi): Changed, keep only meaningful variants.
9531 (SQH, SEZ): New iterators.
9532 (SQH_postfix): New mode attribute.
9533 (SEZ_prefix): New code attribute.
9534 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9535 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9536 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9538 (noncommutative_operator): Check the availability of barrel
9541 2018-04-09 Richard Biener <rguenther@suse.de>
9543 PR tree-optimization/85284
9544 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9545 Only use the niter constraining form of simple_iv when the exit
9548 2018-04-09 Tom de Vries <tom@codesourcery.com>
9551 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9552 (define_expand "*memory_barrier"): New define_expand.
9553 (define_insn "memory_barrier"): New insn.
9555 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9557 PR rtl-optimization/80463
9558 PR rtl-optimization/83972
9559 PR rtl-optimization/83480
9561 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9562 correct producer for the insn.
9563 (tidy_control_flow): Fixup seqnos in case of debug insns.
9565 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9567 PR rtl-optimization/83913
9569 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9570 different sched-times when merging exprs.
9572 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9574 PR rtl-optimization/83962
9576 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9577 tidy_fallthru_edge and tidy_control_flow.
9579 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9581 PR rtl-optimization/83530
9583 * sel-sched.c (force_next_insn): New global variable.
9584 (remove_insn_for_debug): When force_next_insn is true, also leave only
9585 next insn in the ready list.
9586 (sel_sched_region): When the region wasn't scheduled, make another pass
9587 over it with force_next_insn set to 1.
9589 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9591 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9593 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9594 for interrupt control.
9595 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9596 functions for interrupt control.
9597 * config/nds32/nds32-intrinsic.md: Likewise.
9598 * config/nds32/nds32_intrinsic.h: Likewise.
9599 * config/nds32/nds32.h (nds32_builtins): Likewise.
9601 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9603 * config/nds32/nds32.c (nds32_init_machine_status,
9604 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9605 strict_aligned_p field.
9606 (nds32_expand_to_rtl_hook): New function.
9607 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9608 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9610 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9611 Chung-Ju Wu <jasonwucj@gmail.com>
9613 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9614 * config/nds32/nds32-n7.md: New file.
9615 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9616 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9618 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9619 * config/nds32/nds32.md (pipeline_model): Add n7.
9620 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9621 * config/nds32/pipelines.md: Include n7 settings.
9623 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9624 Chung-Ju Wu <jasonwucj@gmail.com>
9626 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9627 * config/nds32/nds32-e8.md: New file.
9628 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9629 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9631 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9632 * config/nds32/nds32.md (pipeline_model): Add e8.
9633 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9634 * config/nds32/pipelines.md: Include e8 settings.
9636 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9637 Chung-Ju Wu <jasonwucj@gmail.com>
9639 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9640 * config/nds32/nds32-n8.md: New file.
9641 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9642 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9644 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9645 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9646 * config/nds32/nds32.md (pipeline_model): Add n8.
9647 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9648 * config/nds32/pipelines.md: Include n8 settings.
9650 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9651 Chung-Ju Wu <jasonwucj@gmail.com>
9653 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9654 * config/nds32/nds32-n9-2r1w.md: New file.
9655 * config/nds32/nds32-n9-3r2w.md: New file.
9656 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9657 nds32_register_ports): New or modify for cpu n9.
9658 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9660 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9661 * config/nds32/nds32-utils.c: New file.
9662 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9663 TARGET_MUL_SLOW): Define.
9664 * config/nds32/nds32.md (pipeline_model): New attribute.
9665 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9666 New options that support cpu n9.
9667 * config/nds32/pipelines.md: Include n9 settings.
9668 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9670 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9672 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9673 information if necessary.
9674 (output_cond_branch_compare_zero): Likewise.
9675 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9676 (nds32_target_alignment): Refine for alignment.
9677 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9678 (FUNCTION_BOUNDARY): Modify.
9679 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9681 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9683 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9685 * config/nds32/constants.md (unspec_volatile_element): Add values for
9686 TLB operation and data prefetch.
9687 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9688 functions for TLB operation and data prefetch.
9689 * config/nds32/nds32-intrinsic.md: Likewise.
9690 * config/nds32/nds32_intrinsic.h: Likewise.
9691 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9692 (nds32_print_operand): Likewise.
9693 * config/nds32/nds32.h (nds32_builtins): Likewise.
9695 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9696 Andrew Pinski <pinsika@gcc.gnu.org>
9699 * match.pd: Use constant_boolean_node of correct type instead of
9700 boolean_true_node or boolean_false_node for simplifying
9701 pointer comparisons to zero.
9703 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9705 PR tree-optimization/80021
9706 * tree.c (verify_type_variant): Make error call in verify_variant_match
9707 translatable and remove final full stop.
9709 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9711 * config/nds32/constants.md (unspec_volatile_element): Add
9712 UNSPEC_VOLATILE_EH_RETURN.
9713 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9714 nds32_output_stack_pop): Support dwarf exception handling process.
9715 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9716 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9717 exception handling process.
9718 (nds32_compute_stack_frame): Likewise.
9719 (nds32_return_addr_rtx): Likewise.
9720 (nds32_initial_elimination_offset): Likewise.
9721 (nds32_expand_prologue): Likewise.
9722 (nds32_expand_epilogue): Likewise.
9723 (nds32_dynamic_chain_address): New function.
9724 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9726 (DYNAMIC_CHAIN_ADDRESS): Define.
9727 (EH_RETURN_DATA_REGNO): Define.
9728 (EH_RETURN_STACKADJ_RTX): Define.
9729 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9730 patterns for dwarf exception handling.
9732 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9734 * config/nds32/nds32.h: Clean up obsolete macros.
9736 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9738 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9739 Add enum values for particular instructions.
9740 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9741 particular intrinsic functions.
9742 * config/nds32/nds32-intrinsic.md: Likewise.
9743 * config/nds32/nds32_intrinsic.h: Likewise.
9744 * config/nds32/nds32.h (nds32_builtins): Likewise.
9745 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9746 (btst, ave): New patterns for particular instructions.
9748 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9750 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9751 Add enum values for atomic load/store and memory sync.
9752 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9754 * config/nds32/nds32-intrinsic.md: Likewise.
9755 * config/nds32/nds32_intrinsic.h: Likewise.
9756 * config/nds32/nds32.h (nds32_builtins): Likewise.
9758 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9760 PR tree-optimization/85257
9761 * fold-const.c (native_encode_vector): If not all elts could fit
9762 and off is -1, return 0 rather than offset.
9763 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9764 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9765 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9766 adjust buffer in native_interpret_expr call.
9768 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9770 * config/nds32/constants.md (unspec_volatile_element): Add cache
9771 control enum values.
9772 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9773 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9774 * config/nds32/nds32.c (nds32_cctl_names): New.
9775 (nds32_print_operand): Handle cache control register names.
9776 * config/nds32/nds32.h (nds32_builtins): New enum values.
9777 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9779 * config/nds32/nds32.md (type): Add mmu.
9780 * config/nds32/pipelines.md (simple_insn): Add mmu.
9782 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9784 * config/nds32/nds32.md (type): Remove call.
9785 * config/nds32/pipelines.md (simple_insn): Likewise.
9787 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9789 * config/nds32/constants.md (unspec_volatile_element): Add
9790 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9791 UNSPEC_VOLATILE_FMFCFG.
9792 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9793 description for fmfcfg and fmfcsr.
9794 (bdesc_1arg): Add fmtcsr.
9795 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9796 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9797 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9798 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9799 unspec_fmfcfg): New patterns.
9800 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9801 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9802 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9803 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9804 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9805 __nds32__fmfcfg): Define.
9807 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9809 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9810 intrinsic register names.
9811 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9812 intrinsic register enum values and macros.
9814 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9816 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9817 for load/store addressing form.
9818 (nds32_print_operand_address): Likewise.
9820 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9823 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9824 based on LABEL_REF. Remove useless assertion.
9825 (pic_address_needs_scratch): Fix formatting.
9826 (sparc_legitimize_pic_address): Minor tweaks.
9827 (sparc_delegitimize_address): Adjust assertion accordingly.
9828 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9829 into symbolic_operand.
9830 (movsi_high_pic_label_ref): Likewise.
9831 (movsi_lo_sum_pic_label_ref): Likewise.
9832 (movdi_pic_label_ref): Likewise.
9833 (movdi_high_pic_label_ref): Likewise.
9834 (movdi_lo_sum_pic_label_ref): Likewise.
9836 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9838 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9839 custom LIB_SPEC setup.
9841 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9842 Kito Cheng <kito.cheng@gmail.com>
9844 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9845 * config/riscv/freebsd.h: New.
9847 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9849 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9850 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9853 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9854 Kito Cheng <kito.cheng@gmail.com>
9856 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9857 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9858 * config/nds32/nds32-protos.h (nds32_output_call,
9859 nds32_output_return): Declare.
9860 * config/nds32/nds32.md: Refine all the call and return patterns.
9862 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9865 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9866 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9868 PR rtl-optimization/84872
9869 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9870 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9873 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9875 * expr.c (copy_blkmode_to_reg): Revert 254862.
9876 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9878 2018-04-06 Richard Biener <rguenther@suse.de>
9881 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9882 after seeing a component reference with an adjacent field. Treat
9883 refs to arrays at struct end of external decls similar to
9884 refs to unconstrained commons.
9886 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9889 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9890 look through SAVE_EXPRs with non-side-effects argument. Adjust
9892 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9895 2018-04-06 Richard Biener <rguenther@suse.de>
9898 * alias.c (find_base_term): New wrapper around find_base_term
9899 unwinding CSELIB_VAL_PTR changes.
9900 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9903 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9905 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9907 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9908 constant definitions.
9909 ("nop"): lr 0,0 -> nopr r0
9910 ("nop_lr0", "nop_lr1"): New insn definitions.
9912 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9914 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9915 NDS32_V3PUSH_AVAILABLE_P macro.
9917 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9918 Chung-Ju Wu <jasonwucj@gmail.com>
9920 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9921 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9922 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9923 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9924 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9925 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9926 * config/nds32/constraints.md: New constraints and checking for hard
9927 float configuration.
9928 * config/nds32/iterators.md: New mode iterator and attribute for hard
9929 float configuration.
9930 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9932 * config/nds32/nds32-fpu.md: New file.
9933 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9934 deal with hard float code generation.
9935 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9937 (abi_type, float_reg_number): New enum type.
9938 * config/nds32/nds32-predicates.c: New predicates for hard float.
9939 * config/nds32/nds32-protos.h: Declare functions for hard float.
9940 * config/nds32/nds32.c: Implementation for hard float configuration.
9941 * config/nds32/nds32.h: Definitions for hard float configuration.
9942 * config/nds32/nds32.md: Include hard float machine description and
9943 modify patterns for hard float configuration.
9944 * config/nds32/nds32.opt: New options for hard float configuration.
9945 * config/nds32/predicates.md: New predicates for hard float
9948 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9950 * common/config/nds32/nds32-common.c
9951 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9953 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9956 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9957 CONSTRUCTOR_ELT (ctor, ...)->value.
9959 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9962 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9964 2018-04-05 Tom de Vries <tom@codesourcery.com>
9967 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9970 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9971 Kito Cheng <kito.cheng@gmail.com>
9973 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9974 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9975 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9976 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9978 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9979 Kito Cheng <kito.cheng@gmail.com>
9981 * config/nds32/constraints.md (Ufe): New memory constraint.
9982 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9983 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9984 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9986 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9987 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9989 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9991 * config/nds32/nds32.md: Use optimize_size in the condition for
9992 alu-shift instructions.
9994 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9996 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9998 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10000 * config/nds32/nds32.md (negsi2): Refine pattern.
10002 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10003 Chung-Ju Wu <jasonwucj@gmail.com>
10005 * config/nds32/iterators.md (shift_rotate): New code iterator.
10006 (shift): New code attribute.
10007 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10008 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10009 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10010 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10011 bit-wise operations.
10012 (andsi3, *andsi3): Ditto.
10013 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10014 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10015 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10016 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10017 nds32_ior_operand, nds32_xor_operand): New predicates.
10019 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10021 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10022 (addsi3, subsi3): ... this.
10024 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10026 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10028 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10030 * config/nds32/nds32.md: Adjust indention.
10032 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10034 * config/nds32/nds32.md (feature): New attribute.
10036 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10038 * config/nds32/nds32.md (subtype): New attribute.
10040 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10043 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10044 expansion to perform a bitwise AND of the argument followed by a
10045 boolean negation of the result.
10047 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10049 PR rtl-optimization/84878
10050 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10051 the basic block. Assert the use reference is not artificial and that
10052 it has an associated insn.
10054 2018-04-04 Michael Matz <matz@suse.de>
10056 * builtins.c (compute_objsize): Pass correct operand
10057 to array_at_struct_end_p.
10059 2018-04-04 Richard Biener <rguenther@suse.de>
10062 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10063 from contexts for DINFO_LEVEL_TERSE and below.
10065 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10067 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10068 resiter_operand condition.
10069 * config/nds32/nds32.md (*move<mode>): Ditto.
10071 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10072 Monk Chiang <sh.chiang04@gmail.com>
10074 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10076 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10078 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10080 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10081 Kito Cheng <kito.cheng@gmail.com>
10083 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10084 nds32_cond_code_str, output_cond_branch,
10085 output_cond_branch_compare_zero, nds32_expand_cbranch,
10086 nds32_expand_cstore, nds32_expand_movcc,
10087 nds32_output_cbranchsi4_equality_zero,
10088 nds32_output_cbranchsi4_equality_reg,
10089 nds32_output_cbranchsi4_equality_reg_or_const_int,
10090 nds32_output_cbranchsi4_greater_less_zero: New functions.
10091 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10092 nds32_expand_cstore, nds32_expand_movcc,
10093 nds32_output_cbranchsi4_equality_zero,
10094 nds32_output_cbranchsi4_equality_reg,
10095 nds32_output_cbranchsi4_equality_reg_or_const_int,
10096 nds32_output_cbranchsi4_greater_less_zero): Declare.
10097 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10098 nds32_rimm11s_operand): New predicates.
10099 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10100 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10103 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10105 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10106 * config/nds32/nds32.md: Ditto.
10107 * config/nds32/pipelines.md: Ditto.
10109 2018-04-04 Richard Biener <rguenther@suse.de>
10111 PR tree-optimization/85168
10112 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10113 propagating abnormals.
10115 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10117 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10119 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10120 Kito Cheng <kito.cheng@gmail.com>
10122 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10123 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10124 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10125 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10126 * config/nds32/nds32.md (sibcall_internal): New.
10127 (sibcall_register): Remove.
10128 (sibcall_immediate): Remove.
10129 (sibcall_value_internal): New.
10130 (sibcall_value_register): Remove.
10131 (sibcall_value_immediate): Remove.
10132 * config/nds32/predicates.md (nds32_general_register_operand): New.
10133 (nds32_call_address_operand): New.
10135 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10137 PR rtl-optimization/85167
10138 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10139 bb_defs if *split_p, instead preinitialize it to NULL.
10141 PR tree-optimization/85156
10142 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10143 evaluating the argument multiple times.
10145 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10147 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10149 (_mm_cvtpd_ps): Likewise.
10150 (_mm_cvttpd_epi32): Likewise.
10151 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10152 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10153 vector, pixel, and bool following altivec.h include.
10155 2018-04-03 Martin Sebor <msebor@redhat.com>
10157 * doc/extend.texi (Common Function Attributes): Clarify.
10158 (const attribute): Likewise.
10159 (pure attribute): Likewise.
10161 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10164 * config/i386/i386.c (ix86_expand_vector_set): Use
10165 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10167 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10169 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10170 instructions when changing rounding bits to preserve precision bits
10171 in the x87 control word.
10173 2018-04-03 Martin Liska <mliska@suse.cz>
10175 PR tree-optimization/82491
10176 * rtl.h (strip_offset_and_add): Replace += suboffset with
10177 poly_uint64 () + suboffset.
10179 2018-03-29 Martin Liska <mliska@suse.cz>
10180 Martin Jambor <mjambor@suse.cz>
10183 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10184 param_type is not an integral or pointer type.
10186 2018-04-03 Richard Biener <rguenther@suse.de>
10188 * sese.h (recompute_all_dominators): Remove.
10190 2018-04-02 Martin Sebor <msebor@redhat.com>
10192 * doc/invoke.texi (-Wrestrict): Fix typos.
10194 2018-04-02 Jim Wilson <jimw@sifive.com>
10196 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10197 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10198 (<optab>di3, <optab>si3_extend): Likewise.
10199 (<optab>si3_mask, <optab>si3_mask_1): New.
10200 (<optab>di3_mask, <optab>di3_mask_1): New.
10201 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10202 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10203 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10205 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10207 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10210 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10212 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10213 (nds32_canonicalize_comparison): New function.
10215 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10216 Kito Cheng <kito.cheng@gmail.com>
10217 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10219 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10220 * config/nds32/constants.md (unspec_volatile_element): Add
10221 UNSPEC_VOLATILE_RELAX_GROUP.
10222 * config/nds32/nds32-relax-opt.c: New file.
10223 * config/nds32/nds32-predicates.c
10224 (nds32_symbol_load_store_p): New function.
10225 * config/nds32/nds32-protos.h
10226 (nds32_symbol_load_store_p): Declare function.
10227 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10228 * config/nds32/nds32.c
10229 (nds32_register_pass): New function to register pass.
10230 (nds32_register_passes): New function to register passes.
10231 * config/nds32/nds32.md (relax_group): New pattern.
10232 * config/nds32/nds32.opt (mrelax-hint): New option.
10233 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10235 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10237 * config/nds32/t-nds32: Modify files dependency.
10239 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10241 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10242 (PROFILE_HOOK): Define its implementation.
10244 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10246 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10247 type and 32-bit size.
10249 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10251 PR middle-end/85090
10252 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10253 (V_128_256): New mode iterator.
10254 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10255 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10256 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10258 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10259 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10261 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10264 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10265 NaN inputs correctly.
10267 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10270 * config/rs6000/vsx.md (??r): New mode attribute.
10271 (*vsx_mov<mode>_64bit): Use it.
10272 (*vsx_mov<mode>_32bit): Likewise.
10274 2018-03-30 Martin Sebor <msebor@redhat.com>
10276 PR tree-optimization/84818
10277 * builtins.c (check_access): Use warning_n.
10279 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10282 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10284 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10287 2018-03-30 Julia Koval <julia.koval@intel.com>
10290 * x86-tune.def (movx, partial_reg_dependency): Enable for
10293 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10295 PR inline-asm/84985
10296 * lra-constraints.c (process_alt_operands): Move setting
10297 this_alternative_matches below.
10299 2018-03-29 Martin Liska <mliska@suse.cz>
10302 * doc/invoke.texi: Document how LTO works with debug info.
10303 Describe auto-load support of binutils. Mention 'x86-64'
10304 as valid option value of -march option.
10306 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10308 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10311 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10312 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10313 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10316 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10319 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10320 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10321 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10322 for RS6000_BTM_POWERPC64.
10323 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10324 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10325 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10328 (DIVDEU): Likewise.
10330 2018-03-28 Carl Love <cel@us.ibm.com>
10333 2017-09-27 Carl Love <cel@us.ibm.com>
10335 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10336 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10337 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10340 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10342 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10343 instead of __vector bool.
10344 (_mm_max_pu8): Likewise.
10345 (_mm_min_pi16): Likewise.
10347 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10350 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10351 (DIVWEUO): Likewise.
10352 (DIVDEO): Likewise.
10353 (DIVDEUO): Likewise.
10354 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10355 DIVWEUO and DIVDEUO.
10356 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10357 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10358 (div_extend): Likewise.
10359 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10361 (__builtin_divweuo): Likewise.
10362 (__builtin_divdeo): Likewise.
10363 (__builtin_divdeuo): Likewise.
10365 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10368 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10369 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10371 PR tree-optimization/82004
10372 * gimple-match-head.c (optimize_pow_to_exp): New function.
10373 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10374 Don't fold to exp if optimize_pow_to_exp is false.
10376 2018-03-28 Martin Liska <mliska@suse.cz>
10379 * calls.c (initialize_argument_information): Fix trailing space.
10380 * common.opt: Fix typo and provide better explanation for
10381 -fsanitize-coverage option.
10382 * config/i386/i386.opt: Fix typo.
10384 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10385 Martin Liska <mliska@suse.cz>
10388 * gimplify.c (asan_poison_variable): Don't do the check for
10389 gimplify_omp_ctxp here.
10390 (gimplify_decl_expr): Do it here.
10391 (gimplify_target_expr): Likewise.
10393 2018-03-28 Martin Liska <mliska@suse.cz>
10396 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10397 chkp_type_bounds_count if MPX is not enabled.
10399 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10401 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10403 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10406 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10407 function to create the function decl for complex long double
10408 multiply and divide for -mabi=ieeelongdouble.
10409 (init_float128_ieee): Call it.
10411 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10414 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10415 -fcf-protection=branch -mibt.
10416 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10418 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10421 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10422 arm_word_relocations.
10424 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10427 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10428 extern array declarations.
10430 2018-03-27 Richard Biener <rguenther@suse.de>
10432 PR middle-end/84067
10433 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10434 explicit single_use checks.
10436 2018-03-27 Richard Biener <rguenther@suse.de>
10438 PR tree-optimization/85082
10439 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10442 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10444 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10445 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10446 Turn on fasynchronous-unwind-tables and funwind-tables.
10448 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10451 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10452 (*bmi_blsr_<mode>_ccz): Ditto.
10454 2018-03-26 Tom de Vries <tom@codesourcery.com>
10456 PR tree-optimization/85063
10457 * omp-general.c (offloading_function_p): New function. Factor out
10459 * omp-offload.c (pass_omp_target_link::gate): ... here.
10460 * omp-general.h (offloading_function_p): Declare.
10461 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10462 with attribute omp declare target for offloading functions.
10464 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10466 PR tree-optimization/84005
10467 * tree-data-ref.h (get_base_for_alignment): Declare.
10468 * tree-data-ref.c (get_base_for_alignment_1): New function.
10469 (get_base_for_alignment): Likewise.
10470 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10471 get_base_for_alignment to find a suitable base object, instead
10472 of always using drb->base_address.
10474 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10476 PR inline-asm/85022
10477 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10478 known size by default.
10480 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10482 PR inline-asm/85030
10483 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10484 and non BLKmode operands.
10486 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10489 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10490 Clean up attributes.
10492 2018-03-23 Richard Biener <rguenther@suse.de>
10495 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10496 we are going to emit early debug for LTO.
10498 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10500 PR inline-asm/85034
10501 * function.c (match_asm_constraints_1): Don't optimize if input
10502 doesn't satisfy general_operand predicate for output's mode.
10504 PR inline-asm/85022
10505 * alias.c (write_dependence_p): Don't require for x_canonicalized
10506 non-VOIDmode if x has VOIDmode.
10509 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10510 just don't try to optimize it rather than assert it never happens.
10512 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10514 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10515 macro expansions for definition of ST_INTERNAL_<mode> and
10516 LD_INTERNAL_<mode> builtins.
10517 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10519 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10521 (altivec_expand_st_builtin): Likewise.
10522 (altivec_expand_builtin): Remove calls to deleted functions.
10523 (rs6000_address_for_altivec): Delete this function.
10524 * config/rs6000/vector.md: Remove expands for
10525 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10527 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10530 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10531 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10532 re-computing once computed.
10533 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10534 (arm_init_machine_status): Initialize
10535 machine->static_chain_stack_bytes.
10537 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10540 * doc/extend.texi: Add four new prototypes for vec_ld.
10541 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10542 definitions for more logical presentation.
10543 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10544 entries for V1TI variants of __builtin_altivec_ld builtin.
10545 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10546 handling of V1TI variant of LVX icode pattern.
10547 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10548 (rs6000_gimple_fold_builtin): Likewise.
10549 (altivec_init_builtins): Add code to define
10550 __builtin_altivec_lvx_v1ti function.
10552 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10554 PR inline-asm/84941
10555 * function.c (match_asm_constraints_1): Don't do the optimization
10556 if input isn't a REG, SUBREG, MEM or constant.
10558 2018-03-22 Tom de Vries <tom@codesourcery.com>
10560 PR tree-optimization/84956
10561 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10562 bb_has_abnormal_pred.
10564 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10567 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10568 DECL_INITIAL (decl) to decl at the end.
10569 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10570 adjust the comment.
10572 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10574 * doc/extend.texi (__builtin_tgmath): Document when complex
10575 integer types are treated as _Complex _Float64.
10577 2018-03-21 Tom de Vries <tom@codesourcery.com>
10579 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10581 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10583 PR tree-optimization/84960
10584 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10585 if it is ENTRY block, move them into single succ of ENTRY in that case.
10587 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10589 PR tree-optimization/84811
10590 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10591 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10592 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10593 (known_subrange_p): Update accordingly. Cast each value involved
10594 in the size comparison, rather than casting the result of the
10597 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10599 PR tree-optimization/84982
10600 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10601 by flipping the least significant bit rather than all bits from
10602 bitpos to bitpos + bitsize - 1.
10604 2018-03-21 Nathan Sidwell <nathan@acm.org>
10606 * doc/extend.texi (Deprecated Features): Remove mention of
10607 long-deleted deprecations.
10609 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10612 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10613 * configure: Regenerate.
10615 2018-03-21 Tom de Vries <tom@codesourcery.com>
10617 PR tree-optimization/83126
10618 * tree-parloops.c (num_phis): New function.
10619 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10621 2018-03-21 Nathan Sidwell <nathan@acm.org>
10623 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10624 mention anon-struct/union members and trailing attributes.
10626 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10628 PR tree-optimization/84969
10629 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10630 builtin memset partitions if they set different rhs values.
10632 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10634 PR rtl-optimization/84989
10635 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10636 VEC_DUPLICATE with scalar result mode.
10638 2018-03-21 Martin Liska <mliska@suse.cz>
10641 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10642 not intended return statement.
10644 2018-03-21 Martin Liska <mliska@suse.cz>
10647 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10648 (chkp_find_bound_slots_1): Limit number of iterations.
10650 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10653 * Minor grammar fixes for x86 options.
10655 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10658 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10659 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10661 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10664 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10665 (altivec_lvx_<mode>_1op): Likewise.
10666 (altivec_stvx_<mode>_2op): Likewise.
10667 (altivec_stvx_<mode>_1op): Likewise.
10668 (altivec_lvx_<VM2:mode>): New define_expand.
10669 (altivec_stvx_<VM2:mode>): Likewise.
10670 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10671 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10672 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10673 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10674 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10675 (rs6000_gen_lvx): Likewise.
10676 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10677 (altivec_expand_stv_builtin): Likewise.
10678 (altivec_expand_builtin): Likewise.
10679 * config/rs6000/vector.md: Likewise.
10681 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10684 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10687 2018-03-20 Richard Biener <rguenther@suse.de>
10690 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10691 sign-conversions as zero, fall back to standard scalar_stmt
10694 2018-03-20 Martin Liska <mliska@suse.cz>
10697 * predict.c (rebuild_frequencies): Handle case when we have
10698 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10700 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10703 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10704 flag_section_anchors.
10705 * varasm.c (use_blocks_for_decl_p): Remove hack for
10706 dw2_force_const_mem.
10709 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10711 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10712 be created, use lowpart_subreg of operands[0] rather than operands[0]
10714 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10715 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10716 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10717 and n constraint instead of aarch64_shift_imm_di and Usd.
10718 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10719 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10721 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10724 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10725 to favor GPR over NEON registers.
10726 (<shift>di3_neon): Likewise.
10728 2018-03-20 Tom de Vries <tom@codesourcery.com>
10731 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10732 (nvptx_process_pars): Emit bar.sync asap and alap.
10734 2018-03-20 Tom de Vries <tom@codesourcery.com>
10737 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10738 seen_label if seen_label is already set.
10740 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10743 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10744 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10745 Use 1U instead of 1. Formatting fixes.
10748 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10749 instead of TREE_TYPE (s1) for the return value.
10751 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10753 PR tree-optimization/84946
10754 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10755 bitsize + bitsize in poly_uint64 rather than poly_int64.
10758 * dwarf2asm.c: Include fold-const.c.
10759 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10760 of decl rather than decl itself.
10762 PR rtl-optimization/84643
10763 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10765 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10768 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10769 calling assemble_variable.
10771 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10774 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10775 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10777 2018-03-19 Jim Wilson <jimw@sifive.com>
10780 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10781 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10782 (riscv_first_stack_step): Likewise.
10783 (riscv_option_override): Use STACK_BOUNDARY instead of
10784 MIN_STACK_BOUNDARY.
10785 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10786 MIN_STACK_BOUNDARY.
10787 (BIGGEST_ALIGNMENT): Set to 128.
10788 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10789 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10792 2018-03-19 Richard Biener <rguenther@suse.de>
10794 PR tree-optimization/84933
10795 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10796 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10798 2018-03-19 Richard Biener <rguenther@suse.de>
10800 PR tree-optimization/84859
10801 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10802 (cond_if_else_store_replacement): Perform sinking operation on
10803 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10804 Generalize what a BB with a single eligible store is.
10806 2018-03-19 Richard Biener <rguenther@suse.de>
10808 PR tree-optimization/84929
10809 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10810 chrec_is_positive against non-chrec arg.
10812 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10815 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10817 2018-03-18 Martin Liska <mliska@suse.cz>
10819 PR rtl-optimization/84635
10820 * regrename.c (build_def_use): Use matches_mode only when
10823 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10825 PR tree-optimization/84913
10826 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10827 vectorize chains of COND_EXPRs.
10829 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10831 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10833 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10835 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10837 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10839 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10841 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10842 Kito Cheng <kito.cheng@gmail.com>
10844 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10845 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10846 (nds32_adjust_reg_alloc_order): New function.
10847 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10849 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10851 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10852 nds32_print_operand, nds32_print_operand_address): Use
10853 HOST_WIDE_INT_PRINT_DEC instead.
10855 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10857 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10859 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10862 * config/i386/i386.c (initial_ix86_tune_features,
10863 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10864 unsigned long long.
10865 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10866 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10867 rather than 1u << ix86_tune. Formatting fix.
10868 (ix86_option_override_internal): Change ix86_arch_mask from
10869 unsigned int to unsigned HOST_WIDE_INT, initialize to
10870 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10871 (ix86_function_specific_restore): Likewise.
10873 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10876 * postreload.c (reload_combine_recognize_pattern): Perform
10877 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10878 truncate_int_for_mode the result for the destination's mode.
10881 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10882 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10884 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10887 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10890 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10891 regno_allocno_class_array and sorted_pseudos.
10892 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10893 insns where regno is used.
10895 2018-03-16 Martin Liska <mliska@suse.cz>
10898 * multiple_target.c (create_dispatcher_calls): Redirect
10899 reference in the symbol table.
10901 2018-03-16 Martin Liska <mliska@suse.cz>
10904 * multiple_target.c (create_dispatcher_calls): Redirect also
10907 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10911 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10912 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10913 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10915 2018-03-16 Julia Koval <julia.koval@intel.com>
10917 * doc/invoke.texi (Skylake Server): Add CLWB.
10918 Cannonlake): Remove CLWB.
10920 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10922 PR tree-optimization/84841
10923 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10925 (FLOAT_ONE_CONST_TYPE): Define.
10926 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10927 (sort_by_operand_rank): Put entries with higher constant_type last
10928 rather than first to match comments.
10930 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10932 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10935 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10938 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10939 check and use address of target always.
10941 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10944 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10945 (indirect_thunk_bnd_needed): Likewise.
10946 (indirect_thunks_used): Likewise.
10947 (indirect_thunks_bnd_used): Likewise.
10948 (indirect_return_needed): New.
10949 (indirect_return_bnd_needed): Likewise.
10950 (output_indirect_thunk_function): Add a bool argument for
10952 (output_indirect_thunk_function): Don't generate alias for
10953 function return thunk.
10954 (ix86_code_end): Call output_indirect_thunk_function to generate
10955 function return thunks.
10956 (ix86_output_function_return): Set indirect_return_bnd_needed
10957 and indirect_return_needed instead of indirect_thunk_bnd_needed
10958 and indirect_thunk_needed.
10960 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10962 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10963 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10964 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10966 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10967 Paul Hua <paul.hua.gm@gmail.com>
10970 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10972 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10974 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10975 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10976 resp. SFmode cases.
10978 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10981 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10982 instead of GET_MODE_SIZE when comparing Units.
10984 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10987 * varasm.c (hash_section): Return an unchangeble hash value
10988 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10989 Return !aarch64_can_use_per_function_literal_pools_p ().
10991 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10994 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10995 rather than address of cmode as last argument to prepare_cmp_insn.
10997 2018-03-15 Julia Koval <julia.koval@intel.com>
10999 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11000 F_AVX512VNNI, F_AVX512BITALG): New.
11002 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11005 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11006 insn for floating-point loads and stores.
11008 2018-03-14 Carl Love <cel@us.ibm.com>
11010 * config/rs6000/rs6000-c.c: Add macro definitions for
11011 ALTIVEC_BUILTIN_VEC_PERMXOR.
11012 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11013 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11014 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11015 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11017 * config/doc/extend.texi: Add prototypes for vec_permxor.
11019 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11022 * diagnostic-show-locus.c (class layout_point): Convert m_line
11023 from int to linenum_type.
11024 (line_span::comparator): Use linenum "compare" function when
11025 comparing line numbers.
11026 (test_line_span): New function.
11027 (layout_range::contains_point): Convert param "row" from int to
11029 (layout_range::intersects_line_p): Likewise.
11030 (layout::will_show_line_p): Likewise.
11031 (layout::print_source_line): Likewise.
11032 (layout::should_print_annotation_line_p): Likewise.
11033 (layout::print_annotation_line): Likewise.
11034 (layout::print_leading_fixits): Likewise.
11035 (layout::annotation_line_showed_range_p): Likewise.
11036 (struct line_corrections): Likewise for field m_row.
11037 (line_corrections::line_corrections): Likewise for param "row".
11038 (layout::print_trailing_fixits): Likewise.
11039 (layout::get_state_at_point): Likewise.
11040 (layout::get_x_bound_for_row): Likewise.
11041 (layout::print_line): Likewise.
11042 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11043 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11044 * input.c (selftest::test_linenum_comparisons): New function.
11045 (selftest::input_c_tests): Call it.
11046 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11047 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11048 * selftest.h (ASSERT_GT): New macro.
11049 (ASSERT_GT_AT): New macro.
11050 (ASSERT_LT): New macro.
11051 (ASSERT_LT_AT): New macro.
11053 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11055 PR rtl-optimization/84780
11056 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11058 2018-03-14 Martin Liska <mliska@suse.cz>
11060 * tree.c (record_node_allocation_statistics): Use
11061 get_stats_node_kind.
11062 (get_stats_node_kind): New function extracted from
11063 record_node_allocation_statistics.
11064 (free_node): Use get_stats_node_kind.
11066 2018-03-14 Richard Biener <rguenther@suse.de>
11068 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11069 that the value-set of ANTIC_IN doesn't grow.
11072 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11074 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11075 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11077 2018-03-14 Julia Koval <julia.koval@intel.com>
11079 * config.gcc (icelake-client, icelake-server): New.
11081 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11082 (initial_ix86_arch_features): Ditto.
11083 (PTA_SKYLAKE): Add SGX.
11084 (PTA_ICELAKE): Remove.
11085 (PTA_ICELAKE_CLIENT): New.
11086 (PTA_ICELAKE_SERVER): New.
11087 (ix86_option_override_internal): Split up icelake on icelake client and
11089 (get_builtin_code_for_version): Ditto.
11090 (fold_builtin_cpu): Ditto.
11091 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11092 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11093 * config/i386/i386.h (processor_type): Ditto.
11094 * doc/invoke.texi: Ditto.
11096 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11099 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11100 INTEGER_CST offset, add it together with bitpos / 8 and
11101 sign extend based on POINTER_SIZE.
11105 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11108 * config/i386/constraints.md (Yc): New register constraint.
11109 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11110 Use Yc constraint for alternative 2 of operand 0. Remove
11111 preferred_for_speed attribute.
11113 2018-03-14 Richard Biener <rguenther@suse.de>
11115 PR tree-optimization/84830
11116 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11117 with the old one to avoid oscillations.
11119 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11122 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11124 (assign_by_spills): Return a flag of reload assignment failure.
11125 Do not process the reload assignment failures. Do not spill other
11126 reload pseudos if they has the same reg class. Update n if
11128 (lra_assign): Add a return arg. Set up from the result of
11129 assign_by_spills call.
11130 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11131 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11132 usage_insns if it is not NULL.
11133 (spill_hard_reg_in_range): New function.
11134 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11135 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11136 function prototypes.
11137 (lra_assign): Change prototype.
11138 * lra.c (lra): Add code to deal with fails by splitting hard reg
11141 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11143 * config/riscv/riscv.opt (mrelax): New option.
11144 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11145 "norelax" when riscv_mrelax is disabled.
11146 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11148 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11151 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11152 reassociation for int modes.
11154 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11156 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11157 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11159 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11160 * config/aarch64/aarch64-sve.md
11161 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11162 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11163 (*extend<mode><Vwide>2): Rename to...
11164 (aarch64_sve_extend<mode><Vwide>2): ...this.
11165 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11166 renaming the old pattern to...
11167 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11169 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11170 define_expand, renaming the old pattern to...
11171 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11172 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11173 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11174 account when deciding which SVE instruction the optab should use.
11175 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11177 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11179 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11180 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11181 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11182 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11183 (tlsdesc_small_<mode>): Turn a define_expand and use
11184 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11185 (tlsdesc_small_advsimd_<mode>): ...this.
11186 (tlsdesc_small_sve_<mode>): New pattern.
11188 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11190 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11191 (UNSPEC_UMUL_HIGHPART): New constants.
11192 (MUL_HIGHPART): New int iteraor.
11193 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11194 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11196 (*<su>mul<mode>3_highpart): New define_insn.
11198 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11201 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11204 2018-03-13 Martin Liska <mliska@suse.cz>
11207 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11209 (sem_item_optimizer::~sem_item_optimizer): Release it.
11210 (sem_item_optimizer::merge_classes): Register variable aliases.
11211 (sem_item_optimizer::fixup_pt_set): New function.
11212 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11213 * ipa-icf.h: Declare new variables and functions.
11215 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11217 PR middle-end/84834
11218 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11219 integer_pow2p@2 and test integer_pow2p in condition.
11220 (A < 0 ? C : 0): Similarly for @1.
11222 PR middle-end/84831
11223 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11224 characters starting at p contain '\0' character, don't look beyond
11228 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11229 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11232 * reg-stack.c (change_stack): Change update_end var from int to
11233 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11234 also call set_block_for_insn on the newly added insns and rescan.
11237 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11238 on the last operand.
11241 * tree.c (stabilize_reference_1): Return save_expr (e) for
11242 STATEMENT_LIST even if it doesn't have side-effects.
11244 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11246 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11248 2018-03-12 Renlin Li <renlin.li@arm.com>
11250 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11251 aarch64_output_scalar_simd_mov_immediate.
11253 2018-03-12 Martin Sebor <msebor@redhat.com>
11255 PR tree-optimization/83456
11256 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11257 for perfectly overlapping calls to memcpy.
11258 (gimple_fold_builtin_memory_chk): Same.
11259 (gimple_fold_builtin_strcpy): Handle no-warning.
11260 (gimple_fold_builtin_stxcpy_chk): Same.
11261 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11263 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11265 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11266 parameter. Use it for SFmode.
11267 (rs6000_function_arg_advance_1): Adjust.
11268 (rs6000_function_arg): Adjust.
11269 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11271 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11273 PR rtl-optimization/84169
11274 PR rtl-optimization/84780
11275 * combine.c (can_combine_p): Check for a 2-insn combination whether
11276 the destination register is used between the two insns, too.
11278 2018-03-12 Richard Biener <rguenther@suse.de>
11280 PR tree-optimization/84803
11281 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11282 for refs DR analysis didn't process.
11284 2018-03-12 Richard Biener <rguenther@suse.de>
11286 PR tree-optimization/84777
11287 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11288 force-vectorize loops ignore whether we are optimizing for size.
11290 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11292 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11293 (TARGET_MD_ASM_ADJUST): Define.
11295 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11296 Kito Cheng <kito.cheng@gmail.com>
11297 Chung-Ju Wu <jasonwucj@gmail.com>
11299 * config/nds32/nds32.c (nds32_compute_stack_frame,
11300 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11301 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11302 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11303 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11304 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11305 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11306 * config/nds32/nds32.md (prologue, epilogue): Use macro
11307 NDS32_V3PUSH_AVAILABLE_P to do checking.
11309 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11312 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11313 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11314 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11315 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11316 addition of most attributes on !orig_type_die or the attribute not
11317 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11319 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11320 Chung-Ju Wu <jasonwucj@gmail.com>
11322 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11323 __NDS32_VH__ macro.
11324 * config/nds32/nds32.opt (mvh): New option.
11326 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11327 Chung-Ju Wu <jasonwucj@gmail.com>
11329 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11331 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11332 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11335 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11336 Chung-Ju Wu <jasonwucj@gmail.com>
11338 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11340 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11341 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11343 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11344 Kito Cheng <kito.cheng@gmail.com>
11345 Chung-Ju Wu <jasonwucj@gmail.com>
11347 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11348 UNSPEC_FFMISM and UNSPEC_FLMISM.
11349 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11350 for ffb, ffmism and flmism.
11351 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11352 (unspec_ffmism): Ditto.
11353 (unspec_flmism): Ditto.
11354 (nds32_expand_builtin_impl): Check if string extension is available.
11355 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11356 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11358 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11361 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11364 * lra-assigns.c (assign_by_spills): Return a flag of reload
11365 assignment failure. Do not process the reload assignment
11366 failures. Do not spill other reload pseudos if they has the same
11368 (lra_assign): Add a return arg. Set up from the result of
11369 assign_by_spills call.
11370 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11371 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11372 usage_insns if it is not NULL.
11373 (spill_hard_reg_in_range): New function.
11374 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11375 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11376 function prototypes.
11377 (lra_assign): Change prototype.
11378 * lra.c (lra): Add code to deal with fails by splitting hard reg
11381 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11384 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11386 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11389 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11390 (dw_val_node): Add val_symbolic_view.
11391 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11392 (symview_upper_bound): New.
11393 (new_line_info_table): Initialize symviews_since_reset.
11394 (dwarf2out_source_line): Count symviews_since_reset and set
11395 symview_upper_bound.
11396 (dw_val_equal_p): Handle symview.
11397 (add_AT_symview): New.
11398 (print_dw_val): Handle symview.
11399 (attr_checksum, attr_checksum_ordered): Likewise.
11400 (same_dw_val_p, size_of_die): Likewise.
11401 (value_format, output_die): Likewise.
11402 (add_high_low_attributes): Use add_AT_symview for entry_view.
11403 (dwarf2out_finish): Reset symview_upper_bound, clear
11406 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11409 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11410 Add strict argument and use it.
11411 (rs6000_split_multireg_move): Update for new strict argument.
11412 (mem_operand_gpr): Disallow all non-offsettable addresses.
11413 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11415 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11418 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11419 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11420 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11423 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11424 decl, use remap_type if we want to use the type.
11426 2018-03-09 Martin Sebor <msebor@redhat.com>
11428 PR tree-optimization/84526
11429 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11431 (builtin_access::generic_overlap): Be prepared to handle non-array
11434 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11436 PR rtl-optimization/84682
11437 * lra-constraints.c (process_address_1): Check is_address flag
11438 for address constraints.
11439 (process_alt_operands): Likewise.
11440 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11441 preprocess_constraints.
11442 * recog.h (preprocess_constraints): Add oploc parameter.
11444 * recog.c (preprocess_constraints): Test address_operand for
11445 CT_ADDRESS constraints.
11447 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11450 * lra-assigns.c (assign_by_spills): Return a flag of reload
11451 assignment failure. Do not process the reload assignment
11452 failures. Do not spill other reload pseudos if they has the same
11454 (lra_assign): Add a return arg. Set up from the result of
11455 assign_by_spills call.
11456 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11457 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11458 usage_insns if it is not NULL.
11459 (spill_hard_reg_in_range): New function.
11460 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11461 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11462 function prototypes.
11463 (lra_assign): Change prototype.
11464 * lra.c (lra): Add code to deal with fails by splitting hard reg
11467 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11470 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11471 Accept complain bool parameter. Only emit errors if it is true.
11472 (arm_parse_cpu_option_name): Likewise.
11473 (arm_target_thumb_only): Adjust callers of the above.
11474 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11475 prototype to take a default true bool parameter.
11476 (arm_parse_arch_option_name): Likewise.
11478 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11479 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11483 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11484 * configure: Regenerate.
11485 * configure.ac ("linker --version-script option"): New.
11486 ("linker soname option"): New.
11488 2018-03-09 Richard Biener <rguenther@suse.de>
11490 PR tree-optimization/84775
11491 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11492 immediate uses of predicate stmts and mark them modified.
11495 PR tree-optimization/84178
11496 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11498 (version_loop_for_if_conversion): Delay update_ssa call.
11499 (tree_if_conversion): Delay update_ssa until after predicate
11502 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11505 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11506 when the function accesses prior frames.
11508 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11511 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11512 gen_llsym, otherwise call maybe_gen_llsym.
11514 PR inline-asm/84742
11515 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11516 has ',' character inside of it.
11518 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11521 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11522 as clobbering CC_REGNUM.
11524 2018-03-08 Richard Biener <rguenther@suse.de>
11526 PR middle-end/84552
11527 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11528 (follow_copies_to_constant): Do not follow SSA names registered
11531 2018-03-08 Richard Biener <rguenther@suse.de>
11533 PR tree-optimization/84178
11534 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11536 (version_loop_for_if_conversion): Delay update_ssa call.
11537 (tree_if_conversion): Delay update_ssa until after predicate
11540 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11542 PR tree-optimization/84178
11543 * tree-if-conv.c (release_bb_predicate): Remove the
11544 the assertion that the stmts have NULL use_ops.
11545 Discard the statements, asserting that they haven't
11546 yet been added to a BB.
11548 2018-03-08 Richard Biener <rguenther@suse.de>
11550 PR tree-optimization/84746
11551 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11552 (phi_translate): Pass in destination ANTIC_OUT set.
11553 (phi_translate_1): Likewise. For a simplified result lookup
11554 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11555 (phi_translate_set): Adjust.
11556 (do_pre_regular_insertion): Likewise.
11557 (do_pre_partial_partial_insertion): Likewise.
11559 2018-03-08 Martin Liska <mliska@suse.cz>
11561 PR gcov-profile/84735
11562 * doc/gcov.texi: Document usage of profile files.
11563 * gcov-io.h: Document changes in the format.
11565 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11569 * dwarf2out.c (struct dw_line_info_table): Update comments for
11571 (FORCE_RESET_NEXT_VIEW): New.
11572 (FORCE_RESETTING_VIEW_P): New.
11573 (RESETTING_VIEW_P): Check for -1 too.
11574 (ZERO_VIEW_P): Likewise.
11575 (new_line_info_table): Force-reset next view.
11576 (dwarf2out_begin_function): Likewise.
11577 (dwarf2out_source_line): Simplify zero_view_p initialization.
11578 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11579 view directly. Omit view when omitting .loc at line 0.
11581 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11583 PR tree-optimization/84740
11584 * tree-switch-conversion.c (process_switch): Call build_constructors
11585 only if info.phi_count is non-zero.
11587 PR tree-optimization/84739
11588 * tree-tailcall.c (find_tail_calls): Check call arguments against
11589 DECL_ARGUMENTS (current_function_decl) rather than
11590 DECL_ARGUMENTS (func) when checking for tail recursion.
11592 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11594 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11595 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11596 Volker Reichelt's entry and add entries for people that perform
11597 GCC fuzzy testing and report numerous bugs.
11599 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11602 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11603 readonly data in sdata, if that is disabled.
11604 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11605 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11606 -mreadonly-in-sdata option.
11608 2018-03-07 Martin Sebor <msebor@redhat.com>
11610 PR tree-optimization/84468
11611 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11612 basic block when looking for nul assignment.
11614 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11617 * except.h (output_function_exception_table): Adjust prototype.
11618 * except.c (output_function_exception_table): Remove FNNAME parameter
11619 and add SECTION parameter. Ouput one part of the table at a time.
11620 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11621 the first part of the exception table and emit unwind directives.
11622 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11623 (i386_pe_seh_cold_init): Likewise.
11624 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11625 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11626 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11627 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11628 * config/i386/winnt.c: Include except.h.
11629 (struct seh_frame_state): Add reg_offset, after_prologue and
11630 in_cold_section fields.
11631 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11632 (i386_pe_seh_cold_init): New function.
11633 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11634 to seh->in_cold_section.
11635 (seh_emit_push): Record the offset of the push.
11636 (seh_emit_save): Record the offet of the save.
11637 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11638 Test seh->after_prologue to disregard the epilogue.
11639 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11640 (i386_pe_end_cold_function): New function.
11642 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11645 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11646 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11649 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11651 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11652 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11654 PR middle-end/84723
11655 * multiple_target.c: Include tree-inline.h and intl.h.
11656 (expand_target_clones): Diagnose and fail if node->definition and
11657 !tree_versionable_function_p (node->decl).
11659 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11661 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11663 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11664 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11665 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11667 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11670 * combine.c (try_combine): Use reg_or_subregno instead of handling
11671 just paradoxical SUBREGs and REGs.
11673 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11675 * config/arc/arc.c (arc_finalize_pic): Remove function.
11676 (arc_must_save_register): We use single base PIC register, remove
11677 checks to save/restore the PIC register.
11678 (arc_expand_prologue): Likewise.
11679 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11681 (arc_verify_short): Likewise.
11682 (arc_attr_type): Likewise.
11683 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11684 (walk_stores): Likewise.
11685 (arc_address_cost): Make it static.
11686 (arc_verify_short): Likewise.
11687 (branch_dest): Likewise.
11688 (arc_attr_type): Likewise.
11689 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11690 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11691 (arc_final_prescan_insn): Remove inserting the nops due to
11692 hardware hazards. It is done in reorg step.
11693 (insn_length_variant_t): Remove.
11694 (insn_length_parameters_t): Likewise.
11695 (arc_insn_length_parameters): Likewise.
11696 (arc_get_insn_variants): Likewise.
11697 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11699 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11701 PR inline-asm/84683
11702 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11705 PR tree-optimization/84687
11706 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11708 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11710 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11712 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11713 Rename to ppc_speculation_barrier.
11714 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11715 __builtin_ppc_speculation_barrier.
11717 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11720 * combine.c (combine_simplify_rtx): Don't try to simplify if
11721 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11724 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11726 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11727 to 32 bytes when compiling for POWER9.
11729 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11732 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11733 regparm >= 3 with no arg reg available also for calls with
11734 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11737 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11739 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11741 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11744 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11746 2018-03-05 Richard Biener <rguenther@suse.de>
11748 PR tree-optimization/84486
11749 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11750 When inserting a __builtin_assume_aligned call set the LHS
11751 SSA name alignment info accordingly.
11753 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11755 PR tree-optimization/84114
11756 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11757 Avoid reassociation of FLOAT_MODE addition.
11759 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11761 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11762 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11763 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11764 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11765 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11766 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11767 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11769 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11770 __builtin_ia32_wbinvd): New builtins.
11771 (SPECIAL_ARGS2): New.
11772 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11773 (SPECIAL_ARGS2): New.
11774 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11775 (ix86_valid_target_attribute_inner_p): Ditto.
11776 (ix86_init_mmx_sse_builtins): Add special_args2.
11777 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11778 TARGET_WBNOINVD_P): New.
11779 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11780 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11781 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11782 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11783 * config/i386/pconfigintrin.h: New file.
11784 * config/i386/wbnoinvdintrin.h: Ditto.
11785 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11787 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11789 2018-03-05 Richard Biener <rguenther@suse.de>
11791 PR tree-optimization/84670
11792 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11794 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11795 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11796 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11797 doesn't grow if all successors (recursively) were visited at least
11800 2018-03-05 Richard Biener <rguenther@suse.de>
11802 PR tree-optimization/84650
11803 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11804 if executed in the loop pipeline.
11806 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11808 * doc/configfiles.texi (Configuration Files): Move info about
11809 conditionalizing $target-protos.h to...
11810 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11811 differs from $target-protos.h.
11813 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11814 Chung-Ju Wu <jasonwucj@gmail.com>
11816 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11817 * config/nds32/nds32-multiple.md (setmemsi): Define.
11818 * config/nds32/nds32-memory-manipulation.c
11819 (nds32_gen_dup_4_byte_to_word_value): New.
11820 (emit_setmem_word_loop): New.
11821 (emit_setmem_byte_loop): New.
11822 (nds32_expand_setmem_loop): New.
11823 (nds32_expand_setmem_loop_v3m): New.
11824 (nds32_expand_setmem_unroll): New.
11825 (nds32_expand_setmem): New.
11827 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11828 Chung-Ju Wu <jasonwucj@gmail.com>
11830 * config/nds32/nds32-memory-manipulation.c
11831 (nds32_emit_load_store): New.
11832 (nds32_emit_post_inc_load_store): New.
11833 (nds32_emit_mem_move): New.
11834 (nds32_emit_mem_move_block): New.
11835 (nds32_expand_movmemsi_loop_unknown_size): New.
11836 (nds32_expand_movmemsi_loop_known_size): New.
11837 (nds32_expand_movmemsi_loop): New.
11838 (nds32_expand_movmemsi_unroll): New.
11839 (nds32_expand_movmemqi): Rename ...
11840 (nds32_expand_movmemsi): ... to this.
11841 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11842 (movmemsi): ... to this.
11843 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11844 (nds32_expand_movmemsi): ... to this.
11846 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11847 Monk Chiang <sh.chiang04@gmail.com>
11848 Chung-Ju Wu <jasonwucj@gmail.com>
11850 * config/nds32/nds32-protos.h
11851 (nds32_expand_load_multiple): New arguments.
11852 (nds32_expand_store_multiple): Ditto.
11853 (nds32_valid_multiple_load_store): Rename ...
11854 (nds32_valid_multiple_load_store_p): ... to this.
11855 * config/nds32/nds32-memory-manipulation.c
11856 (nds32_expand_load_multiple): Refine implementation.
11857 (nds32_expand_store_multiple): Ditto.
11858 * config/nds32/nds32-multiple.md
11859 (load_multiple): Update nds32_expand_load_multiple interface.
11860 (store_multiple): Update nds32_expand_store_multiple interface.
11861 * config/nds32/nds32-predicates.c
11862 (nds32_valid_multiple_load_store): Rename ...
11863 (nds32_valid_multiple_load_store_p): ... to this and refine
11865 * config/nds32/predicates.md
11866 (nds32_load_multiple_and_update_address_operation): New predicate.
11867 (nds32_store_multiple_and_update_address_operation): New predicate.
11869 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11870 Chung-Ju Wu <jasonwucj@gmail.com>
11872 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11873 (combo): New attribute.
11874 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11876 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11878 * config/nds32/nds32.opt: Change -mcmodel= default value.
11880 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11881 Monk Chiang <sh.chiang04@gmail.com>
11882 Chung-Ju Wu <jasonwucj@gmail.com>
11884 * config/nds32/constants.md (unspec_element): New enum.
11885 * config/nds32/constraints.md (Umw): New constraint.
11886 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11887 * config/nds32/nds32-intrinsic.md: Likewise.
11888 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11889 (nds32_valid_smw_lwm_base_p): New.
11890 (nds32_output_smw_single_word): New.
11891 (nds32_output_lmw_single_word): New.
11892 (nds32_expand_unaligned_load): New.
11893 (nds32_expand_unaligned_store): New.
11894 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11895 (nds32_output_smw_single_word): Declare.
11896 (nds32_output_lmw_single_word): Declare.
11897 (nds32_expand_unaligned_load): Declare.
11898 (nds32_expand_unaligned_store): Declare.
11899 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11900 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11901 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11902 NDS32_BUILTIN_UASTORE_DW.
11903 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11906 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11907 Kito Cheng <kito.cheng@gmail.com>
11908 Chung-Ju Wu <jasonwucj@gmail.com>
11910 * config/nds32/nds32-intrinsic.c
11911 (nds32_expand_builtin_null_ftype_reg): Delete.
11912 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11913 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11914 (nds32_read_argument): New.
11915 (nds32_legitimize_target): Ditto.
11916 (nds32_legitimize_argument): Ditto.
11917 (nds32_check_constant_argument): Ditto.
11918 (nds32_expand_unop_builtin): Ditto.
11919 (nds32_expand_unopimm_builtin): Ditto.
11920 (nds32_expand_binop_builtin): Ditto.
11921 (nds32_builtin_decl_impl): Ditto.
11922 (builtin_description): Ditto.
11923 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11924 (nds32_init_builtins_impl): Ditto.
11925 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11926 (nds32_builtin_decl): New.
11927 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11928 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11930 2018-03-02 Jeff Law <law@redhat.com>
11932 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11933 (redundant_insn, fill_simple_delay_slots): Likewise.
11934 (fill_slots_from_thread): Likewise.
11935 * resource.c (mark_referenced_resources): Likewise.
11936 (mark_set_resources, find_dead_or_set_registers): Likewise.
11938 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11940 * substring-locations.h (format_warning_va): Formatting fix for
11941 ATTRIBUTE_GCC_DIAG.
11942 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11944 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11945 * substring-locations.c: Include intl.h.
11946 (format_warning_va): Turned into small wrapper around
11947 format_warning_n_va, renamed to ...
11948 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11949 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11951 (format_warning_at_substring_n): New function.
11952 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11953 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11954 format_warning_at_substring with just a shorter name instead of
11955 const function pointer.
11956 (fmtwarn_n): New function.
11957 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11958 appropriate, get rid of all the fmtstr temporaries, move conditionals
11959 with G_() wrapped string literals directly into fmtwarn arguments,
11960 cast dir.len to (int), formatting fixes.
11962 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11964 * doc/invoke.texi: Remove "Cilk Plus" references.
11966 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11967 Richard Biener <rguenther@suse.de>
11970 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11971 for error or warning attributes if CALL_FROM_THUNK_P is set.
11974 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11977 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11978 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11981 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11982 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11984 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11985 instead of -1U in last predictors element's probability member.
11987 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11990 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11991 arguments if they are comparable.
11993 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11995 PR tree-optimization/84634
11996 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11997 masks and masked_loop_p with a single loop_masks, making sure it's
11998 null for bb vectorization.
12000 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12002 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12003 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12004 loop->force_vectorize to check whether there is no alias.
12006 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12009 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12011 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12013 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12014 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12015 instead of a loop around prev_real_insn.
12016 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12019 PR inline-asm/84625
12020 * config/i386/i386.c (ix86_print_operand): Use conditional
12021 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12024 2018-03-02 Richard Biener <rguenther@suse.de>
12026 PR tree-optimization/84427
12027 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12028 (bitmap_set_subtract_values): Rewrite to handle multiple
12031 (prune_clobbered_mems): Likewise.
12032 (phi_translate): Take edge instead of pred/phiblock.
12033 (phi_translate_1): Likewise.
12034 (phi_translate_set): Likewise. Insert all translated
12035 exprs for a value into the set, keeping possibly multiple
12036 expressions per value.
12037 (compute_antic_aux): Adjust for phi_translate changes.
12038 When intersecting union the expressions and prune those
12039 not in the final value set, keeping possibly multiple
12040 expressions per value. Do not use value-insertion
12041 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12042 all expressions. Add verification that the value-sets
12043 only shrink during iteration.
12044 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12045 (do_pre_regular_insertion): Likewise.
12046 (do_pre_partial_partial_insertion): Likewise.
12048 2018-03-02 Richard Biener <rguenther@suse.de>
12051 * config/darwin.c (saved_debug_info_level): New static global.
12052 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12053 (darwin_asm_lto_end): Restore debug info generation settings.
12055 2018-03-01 Martin Liska <mliska@suse.cz>
12058 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12059 volatile arguments.
12061 2018-03-01 Richard Biener <rguenther@suse.de>
12064 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12065 type annotation with LTO when debug was disabled at compile-time.
12067 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12069 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12071 (mips_final_postscan_insn): Likewise.
12073 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12075 PR rtl-optimization/84528
12076 * alias.c (init_alias_target): Add commentary.
12077 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12078 a unique base value if the frame pointer is not eliminated
12079 to the stack pointer.
12081 2018-03-01 Tom de Vries <tom@codesourcery.com>
12083 PR rtl-optimization/83327
12084 * lra-int.h (hard_regs_spilled_into): Declare.
12085 * lra.c (hard_regs_spilled_into): Define.
12086 (init_reg_info): Init hard_regs_spilled_into.
12087 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12088 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12089 (process_bb_lives): Handle hard_regs_spilled_into.
12090 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12091 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12093 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12095 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12096 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12097 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12098 * config/rs6000/aix72.h: New file.
12100 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12102 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12103 instead of warning_at with conditional singular and plural messages
12107 * stor-layout.c (update_alignment_for_field): For
12108 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12109 && !DECL_PACKED (field), do the alignment update, just use
12110 only desired_align instead of MAX (type_align, desired_align)
12112 (place_field): Don't do known_align < desired_align handling
12113 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12114 is non-NULL, instead do it after rli->prev_field handling and
12115 only if not within a bitfield word. For DECL_PACKED (field)
12116 use type_align of BITS_PER_UNIT.
12118 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12120 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12121 superfluous parentheses and trailing spaces.
12123 2018-02-28 Richard Biener <rguenther@suse.de>
12125 PR tree-optimization/84584
12126 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12127 SCoPs with fake exit edge.
12129 2018-02-28 Martin Liska <mliska@suse.cz>
12132 * timevar.c (timer::print): Fix format to properly print 100%
12135 2018-02-28 Richard Biener <rguenther@suse.de>
12137 PR middle-end/84607
12138 * genmatch.c (capture_info::walk_match): Do not mark
12139 captured expressions without operands as expr_p given
12140 they act more like predicates and should be subject to
12141 "lost tail" side-effect preserving.
12143 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12145 PR rtl-optimization/81611
12146 * auto-inc-dec.c (attempt_change): Move dead note from
12147 mem_insn if it's the next use of regno
12148 (find_address): Take address use of reg holding
12149 non-incremented value. Add parm to limit search to the named
12151 (merge_in_block): Attempt to use a mem insn that is the next
12152 use of the original regno.
12154 2018-02-27 Martin Sebor <msebor@redhat.com>
12157 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12158 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12160 2018-02-27 Martin Sebor <msebor@redhat.com>
12162 PR translation/84207
12163 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12164 n argument to unsigned HOST_WIDE_INT.
12165 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12166 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12167 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12170 2018-02-27 Richard Biener <rguenther@suse.de>
12172 PR tree-optimization/84512
12173 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12174 Do not use the estimate returned from record_stmt_cost for
12175 the scalar iteration cost but sum properly using add_stmt_cost.
12177 2018-02-27 Richard Biener <rguenther@suse.de>
12179 PR tree-optimization/84466
12180 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12181 Adjust last change to less strictly validate use operands.
12183 2018-02-27 Martin Liska <mliska@suse.cz>
12185 PR gcov-profile/84548
12186 * gcov.c (process_file): Allow partial overlap and consider it
12187 also as group functions.
12188 (output_lines): Properly calculate range of lines for a group.
12190 2018-02-27 Martin Liska <mliska@suse.cz>
12192 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12193 'ggc' suffixes. Change first column width.
12194 (timer::print): Fix formatting of the column.
12196 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12198 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12199 preserve inline entry blocks for the sake of debug inline
12200 entry point markers alone.
12201 (remove_unused_locals): Suggest in comments a better place to
12202 force the preservation of inline entry blocks that are
12203 otherwise unused, but do not preserve them.
12205 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12207 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12209 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12212 * config/i386/constraints.md (Bs): Replace
12213 ix86_indirect_branch_register with
12214 TARGET_INDIRECT_BRANCH_REGISTER.
12216 * config/i386/i386.md (indirect_jump): Likewise.
12217 (tablejump): Likewise.
12218 (*sibcall_memory): Likewise.
12219 (*sibcall_value_memory): Likewise.
12220 Peepholes of indirect call and jump via memory: Likewise.
12221 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12222 (*sibcall_value_GOT_32): Likewise.
12223 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12224 (GOT_memory_operand): Likewise.
12225 (call_insn_operand): Likewise.
12226 (sibcall_insn_operand): Likewise.
12227 (GOT32_symbol_operand): Likewise.
12228 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12230 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12232 PR rtl-optimization/83496
12233 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12234 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12235 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12236 redundant insn, if any.
12237 (relax_delay_slots): Likewise.
12238 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12240 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12242 PR tree-optimization/83965
12243 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12244 that grouped statements are part of a reduction chain. Return
12245 true if the statement is not marked as a reduction itself but
12246 is part of a group.
12247 (vect_recog_dot_prod_pattern): Don't check whether the statement
12248 is part of a group here.
12249 (vect_recog_sad_pattern): Likewise.
12250 (vect_recog_widen_sum_pattern): Likewise.
12252 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12255 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12257 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12260 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12262 (ix86_output_indirect_function_return): New prototype.
12263 (ix86_split_simple_return_pop_internal): Likewise.
12264 * config/i386/i386.c (indirect_return_via_cx): New.
12265 (indirect_return_via_cx_bnd): Likewise.
12266 (indirect_thunk_name): Handle return va CX_REG.
12267 (output_indirect_thunk_function): Create alias for
12268 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12269 (ix86_output_indirect_jmp): Remove the bool argument.
12270 (ix86_output_indirect_function_return): New function.
12271 (ix86_split_simple_return_pop_internal): Likewise.
12272 * config/i386/i386.md (*indirect_jump): Don't pass false
12273 to ix86_output_indirect_jmp.
12274 (*tablejump_1): Likewise.
12275 (simple_return_pop_internal): Change it to define_insn_and_split.
12276 Call ix86_split_simple_return_pop_internal to split it for
12277 -mfunction-return=.
12278 (simple_return_indirect_internal): Call
12279 ix86_output_indirect_function_return instead of
12280 ix86_output_indirect_jmp.
12282 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12285 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12286 memset and value initialization afterwards.
12288 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12290 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12292 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12295 * common/config/aarch64/aarch64-common.c
12296 (aarch_option_optimization_table[]): Switch
12297 off fomit-frame-pointer
12299 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12300 Chung-Ju Wu <jasonwucj@gmail.com>
12302 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12304 (store_multiple): Ditto.
12306 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12308 * config.gcc: Add --with-cpu support for nds32 target.
12309 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12310 * config/nds32/nds32.opt: Add -mcpu= option.
12312 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12314 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12315 isel=yes): Warn for these deprecated options.
12317 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12319 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12320 ISA_2_5_MASKS_EMBEDDED.
12322 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12324 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12325 p->max as pointers rather than using iterative_hash_expr.
12327 2018-02-23 Carl Love <cel@us.ibm.com>
12329 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12330 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12332 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12333 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12334 P8V_BUILTIN_VEC_VUNSIGNED2.
12336 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12339 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12340 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12342 * lra-constraints.c (curr_insn_transform): Set up
12343 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12345 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12346 * lra-eliminations.c (spill_pseudos): Ditto.
12347 (process_insn_for_elimination): Ditto.
12348 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12349 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12350 LRA_NON_CLOBBERED_ALT.
12352 2018-02-22 Martin Sebor <msebor@redhat.com>
12354 PR tree-optimization/84480
12355 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12356 to maybe_diag_stxncpy_trunc. Call it.
12357 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12358 from gimple_fold_builtin_strcpy. Print inlining stack.
12359 (handle_builtin_stxncpy): Print inlining stack.
12360 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12362 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12365 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12366 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12367 and -fcheck-pointer-bounds are used together.
12368 (indirect_thunk_prefix): New enum.
12369 (indirect_thunk_need_prefix): New function.
12370 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12371 "_nt" instead of "_bnd" for NOTRACK prefix.
12372 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12373 (output_indirect_thunk_function): Likewise.
12375 (ix86_code_end): Update output_indirect_thunk_function calls.
12376 (ix86_output_indirect_branch_via_reg): Replace
12377 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12378 (ix86_output_indirect_branch_via_push): Likewise.
12379 (ix86_output_function_return): Likewise.
12380 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12381 incompatible with -fcf-protection=branch and
12382 -fcheck-pointer-bounds.
12384 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12387 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12388 Change gcc_assert call to output_operand_lossage.
12390 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12392 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12394 2018-02-22 DJ Delorie <dj@redhat.com>
12395 Sebastian Perta <sebastian.perta@renesas.com>
12396 Oleg Endo <olegendo@gcc.gnu.org>
12398 * config/rx/rx.c (rx_rtx_costs): New function.
12399 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12401 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12403 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12405 2018-02-22 Martin Liska <mliska@suse.cz>
12408 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12409 Add "native" as a possible value.
12411 2018-02-22 Martin Liska <mliska@suse.cz>
12414 * config/i386/i386.c (ix86_option_override_internal):
12415 Add "native" as a possible value for -march and -mtune.
12417 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12420 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12421 to all type variants.
12423 PR tree-optimization/84503
12424 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12425 width as info->bitpos + info->bitsize - start.
12426 (merged_store_group::merge_overlapping): Simplify width computation.
12427 (check_no_overlap): New function.
12428 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12429 start + width and last_order of the group, fail if check_no_overlap
12431 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12432 to group if check_no_overlap fails.
12434 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12436 * config/rs6000/altivec.md: Delete contraint arguments to
12437 define_expand, define_split, and define_peephole2, and in
12438 define_insn_and_split if always unused.
12439 * config/rs6000/darwin.md: Ditto.
12440 * config/rs6000/dfp.md: Ditto.
12441 * config/rs6000/rs6000.md: Ditto.
12442 * config/rs6000/sync.md: Ditto.
12443 * config/rs6000/vector.md: Ditto.
12444 * config/rs6000/vsx.md: Ditto.
12446 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12448 * config/rs6000/altivec.md: Write output control strings as braced
12449 blocks instead of double-quoted strings.
12450 * config/rs6000/darwin.md: Ditto.
12451 * config/rs6000/rs6000.md: Ditto.
12452 * config/rs6000/vector.md: Ditto.
12453 * config/rs6000/vsx.md: Ditto.
12455 2018-02-21 Jason Merrill <jason@redhat.com>
12457 PR c++/84314 - ICE with templates and fastcall attribute.
12458 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12460 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12462 * ipa-cp.c (determine_versionability): Fix comment typos.
12464 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12467 * ipa-cp.c (determine_versionability): Do not version functions caling
12470 2018-02-21 Martin Liska <mliska@suse.cz>
12473 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12474 Add "native" as a possible value.
12475 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12476 the macro when native cpu detection is available.
12478 2018-02-21 Martin Liska <mliska@suse.cz>
12481 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12482 Add "native" as a possible value.
12483 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12484 when native cpu detection is available.
12486 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12487 Martin Sebor <msebor@redhat.com>
12489 PR tree-optimization/84478
12490 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12492 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12493 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12494 support which is conservatively correct, for 2 only stay conservative
12495 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12496 argument to the 2 argument get_range_strlen, adjust 6 arg
12497 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12499 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12500 (gimple_fold_builtin_strlen): Pass true as last argument to
12503 2018-02-20 Martin Sebor <msebor@redhat.com>
12505 PR middle-end/84095
12506 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12507 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12508 (builtin_memref::builtin_memref): Factor out parts into
12509 set_base_and_offset and call it.
12511 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12513 PR middle-end/84406
12514 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12515 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12516 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12517 search at the associated MODE_INT.
12519 2018-02-20 Jeff Law <law@redhat.com>
12521 PR middle-end/82123
12522 PR tree-optimization/81592
12523 PR middle-end/79257
12524 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12525 for range data rather than using global data.
12526 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12527 range data rather than using global data.
12528 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12529 pass it to children as needed.
12530 (struct directive::fmtresult): Similarly.
12531 (struct directive::set_width): Similarly.
12532 (struct directive::set_precision): Similarly.
12533 (format_integer, format_directive, parse_directive): Similarly.
12534 (format_none): Accept unnamed vr_values parameter.
12535 (format_percent, format_floating, format_character): Similarly.
12536 (format_string, format_plain): Similarly.
12537 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12538 the EVRP range analyzer for range data rather than using global data.
12539 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12540 gimple-ssa-evrp-analyze.h
12541 (class sprintf_dom_walker): Add after_dom_children member function.
12542 Add evrp_range_analyzer member.
12543 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12544 range analyzer as needed.
12545 (sprintf_dom_walker::after_dom_children): New member function.
12546 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12548 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12549 (evrp_range_analyzer::pop_to_marker): Likewise.
12551 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12553 PR tree-optimization/84419
12554 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12555 with the required type if its current type is compatible but
12558 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12560 PR middle-end/82004
12561 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12562 after vectorization.
12564 2018-02-20 Martin Liska <mliska@suse.cz>
12567 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12568 possible values if we don't have a hint.
12570 2018-02-20 Martin Liska <mliska@suse.cz>
12574 * final.c (shorten_branches): Build align_tab array with one
12576 * opts.c (finish_options): Add alignment option limit check.
12577 (MAX_CODE_ALIGN): Likewise.
12578 (MAX_CODE_ALIGN_VALUE): Likewise.
12579 * doc/invoke.texi: Document maximum allowed option value for
12580 all -falign-* options.
12582 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12585 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12586 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12587 * var-tracking.c (emit_note_insn_var_location): Remove all references
12588 to NOTE_INSN_CALL_ARG_LOCATION.
12589 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12590 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12591 Use copy_rtx_if_shared.
12592 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12593 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12594 (dwarf2out_var_location): Remove handling of
12595 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12597 * final.c (final_scan_insn): Remove all references to
12598 NOTE_INSN_CALL_ARG_LOCATION.
12599 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12600 before dumping final insns.
12601 * except.c (emit_note_eh_region_end): Remove all references to
12602 NOTE_INSN_CALL_ARG_LOCATION.
12603 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12604 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12605 * config/arc/arc.c (hwloop_optimize): Likewise.
12606 * config/arm/arm.c (create_fix_barrier): Likewise.
12607 * config/s390/s390.c (s390_chunkify_start): Likewise.
12608 * config/sh/sh.c (find_barrier): Likewise.
12609 * config/i386/i386.c (rest_of_insert_endbranch,
12610 ix86_seh_fixup_eh_fallthru): Likewise.
12611 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12612 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12613 * config/frv/frv.c (frv_function_prologue): Likewise.
12614 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12616 (note_outside_basic_block_p): Remove all references to
12617 NOTE_INSN_CALL_ARG_LOCATION.
12618 * gengtype.c (adjust_field_rtx_def): Likewise.
12619 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12621 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12622 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12625 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12628 PR tree-optimization/84452
12629 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12630 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12633 2018-02-19 Martin Liska <mliska@suse.cz>
12636 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12638 2018-02-19 Martin Liska <mliska@suse.cz>
12639 Richard Sandiford <richard.sandiford@linaro.org>
12641 PR tree-optimization/82491
12642 * gimple-fold.c (get_base_constructor): Make earlier bail out
12645 2018-02-19 Carl Love <cel@us.ibm.com>
12647 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12648 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12650 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12651 P8V_BUILTIN_VEC_NEG.
12653 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12655 * config/rl78/rl78.md (movdf): New define expand.
12657 2018-02-19 Martin Liska <mliska@suse.cz>
12660 * doc/invoke.texi: Fix typo.
12661 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12663 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12665 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12666 handle rs6000_single_float and rs6000_double_float specially for
12669 2018-02-16 Jeff Law <law@redhat.com>
12671 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12672 (pop_regs): Use it.
12674 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12677 * ipa-inline.c (inline_small_functions): Fix a typo.
12679 2018-02-16 Nathan Sidwell <nathan@acm.org>
12681 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12683 2018-02-16 Carl Love <cel@us.ibm.com>
12685 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12686 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12687 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12688 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12689 expansion to P8V_BUILTIN_VEC_FLOAT2.
12691 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12693 PR rtl-optimization/70023
12694 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12695 src_regno into account.
12697 2018-02-16 Carl Love <cel@us.ibm.com>
12699 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12700 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12701 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12702 * config/rs6000/rs6000.c: Remove case statements for
12703 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12704 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12705 and P9V_BUILTIN_VEC_VINSERT4B.
12706 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12707 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12708 * config/rs6000/vsx.md:
12709 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12712 2018-02-16 Carl Love <cel@us.ibm.com>
12714 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12716 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12718 * config/rs6000/rs6000-c.c: Add the definitions for
12719 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12720 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12721 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12722 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12723 definition for insert4b and define insn *insert3b_internal.
12724 * doc/extend.texi: Add documentation for vec_extract4b.
12726 2018-02-16 Nathan Sidwell <nathan@acm.org>
12728 * doc/extend.texi (Backwards Compatibility): Mention friend
12729 injection. Note for-scope is deprecated.
12730 * doc/invoke.texi (-ffriend-injection): Deprecate.
12732 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12734 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12735 that moved to I2, also allow destinations that are a paradoxical
12736 subreg (instead of a normal reg).
12738 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12741 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12744 2018-02-16 Richard Biener <rguenther@suse.de>
12746 PR tree-optimization/84037
12747 PR tree-optimization/84016
12749 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12750 Adjust vec_construct for the fact we need additional higher latency
12751 128bit inserts for AVX256 and AVX512 vector builds.
12752 (ix86_add_stmt_cost): Scale vector construction cost for
12755 2018-02-16 Richard Biener <rguenther@suse.de>
12757 PR tree-optimization/84417
12758 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12759 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12760 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12762 2018-02-16 Martin Liska <mliska@suse.cz>
12765 * internal-fn.def (ASAN_CHECK): Set proper flags.
12766 (ASAN_MARK): Likewise.
12768 2018-02-16 Julia Koval <julia.koval@intel.com>
12770 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12771 from PTA_CANNONLAKE.
12773 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12776 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12777 Use ++iter rather than iter++ for std::list iterators.
12778 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12779 defer deleting them until all nodes in the forest are processed. Do
12780 free even leaf nodes. Change to_process into auto_vec.
12783 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12784 * vec.h (vec_default_construct): Use memset instead of placement new
12785 if BROKEN_VALUE_INITIALIZATION is defined.
12786 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12787 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12790 PR rtl-optimization/83723
12791 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12792 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12793 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12795 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12797 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12799 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12801 PR rtl-optimization/81443
12802 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12803 from inner REGs to paradoxical SUBREGs.
12805 2018-02-16 Richard Biener <rguenther@suse.de>
12807 PR tree-optimization/84399
12808 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12809 For operands we can analyze at their definition make sure we can
12810 analyze them at each use as well.
12812 2018-02-16 Richard Biener <rguenther@suse.de>
12814 PR tree-optimization/84190
12815 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12816 volatile accesses if the decl isn't volatile.
12818 2018-02-15 Jason Merrill <jason@redhat.com>
12820 PR c++/84314 - ICE with templates and fastcall attribute.
12821 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12822 TYPE_CANONICAL on an existing type.
12824 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12826 PR tree-optimization/84383
12827 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12828 dstoff nor call operand_equal_p if dstbase is NULL.
12830 PR tree-optimization/84334
12831 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12832 also a CONSTANT_CLASS_P, punt.
12834 2018-02-14 Jim Wilson <jimw@sifive.com>
12836 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12837 first SMALL_OPERAND check. New local min_second_step. Move assert
12838 to where locals are set. Add TARGET_RVC support.
12839 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12841 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12843 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12845 2018-02-14 Martin Sebor <msebor@redhat.com>
12847 PR tree-optimization/83698
12848 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12849 arrays constrain the offset range to their bounds.
12850 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12851 (builtin_access::overlap): Avoid setting the size of overlap if it's
12853 (maybe_diag_overlap): Also consider arrays when deciding what values
12854 of offsets to include in diagnostics.
12856 2018-02-14 Martin Sebor <msebor@redhat.com>
12859 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12860 that correspond to the kind of a declaration.
12862 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12865 * config/pa/pa.md: Load address of PIC label using the linkage table
12866 if the label is nonlocal.
12868 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12870 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12871 warning message if user requests -maltivec=be.
12872 * doc/invoke.texi: Document deprecation of -maltivec=be.
12874 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12877 * config/rs6000/rs6000-c.c: Update definitions for
12878 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12879 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12881 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12884 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12885 add _get_ssp intrinsics. Remove argument from
12886 __builtin_ia32_rdssp[d|q].
12887 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12888 * config/i386/i386-builtin.def: Remove argument from
12889 __builtin_ia32_rdssp[d|q].
12890 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12891 ix86_expand_special_args_builtin for _rdssp[d|q].
12892 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12893 Clear register before usage.
12894 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12895 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12897 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12899 PR tree-optimization/84357
12900 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12901 operand 1 of an ARRAY_REF too.
12903 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12906 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12907 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12909 (set_of_reg): New struct.
12910 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12911 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12912 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12914 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12915 Split into bitclr, bitset, bitinvert patterns if appropriate.
12916 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12917 use rx_fuse_in_memory_bitop.
12918 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12919 to named insn, correct maximum insn length.
12921 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12924 * machmode.def: Define a complex mode for PARTIAL_INT.
12925 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12927 * doc/rtl.texi: Document CSPImode.
12928 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12930 (msp430_hard_regno_nregs_with_padding): Likewise.
12932 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12935 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12937 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12939 PR rtl-optimization/84169
12940 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12941 we generated a parallel as new i3 and we split that to new i2 and i3
12942 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12943 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12944 those to i2, not i1. Partially rewrite this scan code.
12946 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12949 * stor-layout.c (place_field): For variable length fields, adjust
12950 offset_align afterwards not just based on the field's alignment,
12951 but also on the size.
12953 PR middle-end/84309
12954 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12955 of exps and logs in the use_exp2 case.
12957 2018-02-13 Jeff Law <law@redhat.com>
12959 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12960 entry for "vector".
12962 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12965 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12969 * common.opt (gas-loc-support, gas-locview-support): New.
12970 (ginline-points, ginternal-reset-location-views): New.
12971 * doc/invoke.texi: Document them. Use @itemx where intended.
12972 (gvariable-location-views): Adjust.
12973 * target.def (reset_location_view): New.
12974 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12975 (TARGET_RESET_LOCATION_VIEW): New.
12976 * doc/tm.texi: Rebuilt.
12977 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12978 (dwarf2out_default_as_locview_support): New.
12979 (output_asm_line_debug_info): Use option variables.
12980 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12981 (output_loc_list): Likewise.
12982 (add_high_low_attributes): Check option variables.
12983 Don't output entry view attribute in strict mode.
12984 (gen_inlined_subroutine_die): Check option variables.
12985 (dwarf2out_inline_entry): Likewise.
12986 (init_sections_and_labels): Likewise.
12987 (dwarf2out_early_finish): Likewise.
12988 (maybe_reset_location_view): New, from...
12989 (dwarf2out_var_location): ... here. Call it.
12990 * debug.h (dwarf2out_default_as_loc_support): Declare.
12991 (dwarf2out_default_as_locview_support): Declare.
12992 * hooks.c (hook_int_rtx_insn_0): New.
12993 * hooks.h (hook_int_rtx_insn_0): Declare.
12994 * toplev.c (process_options): Take -gas-loc-support and
12995 -gas-locview-support from dwarf2out. Enable
12996 -gvariable-location-views by default only with locview
12997 assembler support. Enable -ginternal-reset-location-views by
12998 default only if the target defines the corresponding hook.
12999 Enable -ginline-points by default if location views are
13000 enabled; force it disabled if statement frontiers are
13002 * tree-inline.c (expand_call_inline): Check option variables.
13003 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13005 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13007 PR tree-optimization/84321
13008 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13009 handling. Also check whether the anti-range contains any values
13010 that satisfy the mask; switch to a VR_RANGE if not.
13012 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13015 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13017 2018-02-13 Martin Jambor <mjambor@suse.cz>
13020 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13021 of call statements, also set location of a load to a temporary.
13023 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13025 * config/rl78/rl78.c (add_vector_labels): New function.
13026 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13027 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13028 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13029 which checks that no arguments are passed.
13030 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13031 * doc/extend.texi: Documentation for the new attribute.
13033 2018-02-13 Andreas Schwab <schwab@suse.de>
13035 * config/riscv/linux.h (CPP_SPEC): Define.
13037 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13040 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13041 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13042 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13043 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13044 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13045 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13046 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13047 temporarily for AES and PCLMUL builtins.
13049 PR tree-optimization/84339
13050 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13051 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13054 PR middle-end/84309
13055 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13056 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13057 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13059 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13061 * omp-simd-clone.h: New file.
13062 * omp-simd-clone.c: Include omp-simd-clone.h.
13063 (expand_simd_clones): No longer static.
13064 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13065 cgraph.h and omp-simd-clone.h.
13066 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13067 (vect_recog_widen_shift_pattern): Formatting fix.
13068 (vect_pattern_recog_1): Don't check optab for calls.
13071 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13072 operands[2] into a REG before using gen_lowpart on it.
13074 2018-02-12 Jeff Law <law@redhat.com>
13077 * config/sh/sh.c (find_barrier): Consider a sibling call
13080 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13081 successfully back substituting a reg.
13083 2018-02-12 Richard Biener <rguenther@suse.de>
13085 PR tree-optimization/84037
13086 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13087 parameter, move visited init to caller.
13088 (vect_slp_analyze_operations): Separate cost from validity
13089 check, initialize visited once for all instances.
13090 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13092 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13094 (vect_model_promotion_demotion_cost): Likewise.
13095 (vectorizable_bswap): Guard cost modeling with !slp_node
13096 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13098 (vectorizable_call): Likewise.
13099 (vectorizable_conversion): Likewise.
13100 (vectorizable_assignment): Likewise.
13101 (vectorizable_shift): Likewise.
13102 (vectorizable_operation): Likewise.
13103 (vectorizable_store): Likewise.
13104 (vectorizable_load): Likewise.
13105 (vectorizable_condition): Likewise.
13106 (vectorizable_comparison): Likewise.
13108 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13111 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13112 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13113 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13115 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13117 PR middle-end/83665
13118 * params.def (inline-min-speedup): Increase from 8 to 15.
13119 (max-inline-insns-auto): Decrease from 40 to 30.
13120 * ipa-split.c (consider_split): Add some buffer for function to
13121 be considered inlining candidate.
13122 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13125 2018-02-12 Richard Biener <rguenther@suse.de>
13127 PR tree-optimization/84037
13128 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13129 matched stmts if we cannot swap the non-matched ones.
13131 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13133 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13134 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13135 _mm_maskz_scalef_round_ss): New intrinsics.
13136 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13137 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13138 __builtin_ia32_scalefss_round): Remove.
13139 (__builtin_ia32_scalefsd_mask_round,
13140 __builtin_ia32_scalefss_mask_round): New intrinsics.
13141 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13142 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13143 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13144 "<round_constraint>")): Changed to ...
13145 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13146 "<round_scalar_constraint>")): ... this.
13147 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13148 %0, %1, %2<round_op3>}"): Changed to ...
13149 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13150 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13151 %2<round_scalar_mask_op3>}"): ... this.
13152 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13154 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13156 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13157 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13158 (_mm_maskz_sqrt_round_ss): New intrinsics.
13159 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13160 (__builtin_ia32_sqrtsd_mask_round)
13161 (__builtin_ia32_sqrtss_mask_round): New builtins.
13162 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13163 (__builtin_ia32_sqrtss_round): Remove.
13164 (__builtin_ia32_sqrtsd_mask_round)
13165 (__builtin_ia32_sqrtss_mask_round): New builtins.
13166 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13167 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13168 ((match_operand:VF_128 1 "vector_operand"
13169 "xBm,<round_constraint>")): Changed to ...
13170 ((match_operand:VF_128 1 "vector_operand"
13171 "xBm,<round_scalar_constraint>")): ... this.
13172 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13173 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13174 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13175 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13176 %<iptr>1<round_scalar_mask_op3>}): ... this.
13177 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13178 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13180 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13183 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13184 Cast vec_cmpeq result to correct type.
13185 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13186 Cast vec_cmpgt result to correct type.
13188 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13190 * final.c (final_scan_insn_1): Renamed from...
13191 (final_scan_insn): ... this. New wrapper, to recover
13192 seen from the outermost call in recursive ones.
13193 * config/sparc/sparc.c (output_return): Drop seen from call.
13194 (output_sibcall): Likewise.
13195 * config/visium/visium.c (output_branch): Likewise.
13197 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13199 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13202 2018-02-10 Alan Modra <amodra@gmail.com>
13205 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13206 Specify LR as an input.
13208 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13211 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13212 remove_member_access_dummy_vars): New functions.
13213 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13214 lower_omp_1, execute_lower_omp): Use them.
13216 PR rtl-optimization/84308
13217 * shrink-wrap.c (spread_components): Release todo vector.
13219 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13221 PR rtl-optimization/57193
13222 * ira-color.c (struct allocno_color_data): Add member
13223 conflict_allocno_hard_prefs.
13224 (update_conflict_allocno_hard_prefs): New.
13225 (bucket_allocno_compare_func): Add a preference based on
13226 conflict_allocno_hard_prefs.
13227 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13228 (color_allocnos): Remove a dead code. Initiate
13229 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13231 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13234 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13235 constraint from =wa to wa. Avoid a subreg on the output operand,
13236 instead use a pseudo and subreg it in a move.
13237 (p9_xxbrd_<mode>): Changed to ...
13238 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13239 (p9_xxbrd_v2df): New expander.
13240 (p9_xxbrw_<mode>): Changed to ...
13241 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13242 (p9_xxbrw_v4sf): New expander.
13244 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13246 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13248 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13251 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13252 multiply in 32-bit mode.
13253 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13254 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13257 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13259 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13260 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13261 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13262 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13264 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13267 * dwarf2out.c (is_trivial_indirect_ref): New function.
13268 (dwarf2out_late_global_decl): Do not generate a location
13269 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13270 and that are not defined in the current unit.
13272 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13274 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13275 instead of a libcall for UNORDERED.
13277 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13280 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13281 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13283 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13286 * config/s390/s390.c (s390_set_current_function): Invoke
13287 s390_indirect_branch_settings also if fndecl didn't change.
13289 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13291 * config/rs6000/rs6000.md (blockage): Set length to zero.
13293 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13295 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13297 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13300 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13301 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13305 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13306 PARALLEL incoming that failed vt_get_decl_and_offset check.
13308 PR middle-end/84237
13309 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13310 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13312 (get_variable_section): For decls in named .bss* sections pass true as
13313 second argument to bss_initializer_p.
13315 2018-02-09 Marek Polacek <polacek@redhat.com>
13316 Jakub Jelinek <jakub@redhat.com>
13319 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13320 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13321 Sync some changes from cxx_fold_indirect_ref.
13323 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13325 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13327 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13328 (BLOCK_INLINE_ENTRY_LABEL): New.
13329 (dwarf2out_var_location): Disregard inline entry markers.
13330 (inline_entry_data): New struct.
13331 (inline_entry_data_hasher): New hashtable type.
13332 (inline_entry_data_hasher::hash): New.
13333 (inline_entry_data_hasher::equal): New.
13334 (inline_entry_data_table): New variable.
13335 (add_high_low_attributes): Add DW_AT_entry_pc and
13336 DW_AT_GNU_entry_view attributes if a pending entry is found
13337 in inline_entry_data_table. Add old entry_pc attribute only
13338 if debug nonbinding markers are disabled.
13339 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13340 markers are enabled.
13341 (block_within_block_p, dwarf2out_inline_entry): New.
13342 (dwarf2out_finish): Check that no entries remained in
13343 inline_entry_data_table.
13344 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13345 (final_scan_insn, notice_source_line): Likewise.
13346 (rest_of_clean_state): Skip inline entry markers.
13347 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13349 * gimple.c (gimple_build_debug_inline_entry): New.
13350 * gimple.h (enum gimple_debug_subcode): Add
13351 GIMPLE_DEBUG_INLINE_ENTRY.
13352 (gimple_build_debug_inline_entry): Declare.
13353 (gimple_debug_inline_entry_p): New.
13354 (gimple_debug_nonbind_marker_p): Adjust.
13355 * insn-notes.def (INLINE_ENTRY): New.
13356 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13357 inline entry marker notes.
13358 (print_insn): Likewise.
13359 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13360 (INSN_DEBUG_MARKER_KIND): Likewise.
13361 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13362 * tree-inline.c (expand_call_inline): Build and insert
13363 debug_inline_entry stmt.
13364 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13365 inline entry blocks early, if nonbind markers are enabled.
13366 (dump_scope_block): Dump fragment info.
13367 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13368 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13369 (gimple_build_debug_inline_entry): New.
13370 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13371 Enable/disable inline entry points too.
13372 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13373 (DEBUG_INSN): Describe inline entry markers.
13375 * common.opt (gvariable-location-views): New.
13376 (gvariable-location-views=incompat5): New.
13377 * config.in: Rebuilt.
13378 * configure: Rebuilt.
13379 * configure.ac: Test assembler for view support.
13380 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13381 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13382 * dwarf2out.c (var_loc_view): New typedef.
13383 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13384 (dwarf2out_locviews_in_attribute): New.
13385 (dwarf2out_locviews_in_loclist): New.
13386 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13387 (enum dw_line_info_opcode): Add LI_adv_address.
13388 (struct dw_line_info_table): Add view.
13389 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13390 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13391 (zero_view_p): New variable.
13392 (ZERO_VIEW_P): New macro.
13393 (output_asm_line_debug_info): New.
13394 (struct var_loc_node): Add view.
13395 (add_AT_view_list, AT_loc_list): New.
13396 (add_var_loc_to_decl): Add view param. Test it against last.
13397 (new_loc_list): Add view params. Record them.
13398 (AT_loc_list_ptr): Handle loc and view lists.
13399 (view_list_to_loc_list_val_node): New.
13400 (print_dw_val): Handle dw_val_class_view_list.
13401 (size_of_die): Likewise.
13402 (value_format): Likewise.
13403 (loc_list_has_views): New.
13404 (gen_llsym): Set vl_symbol too.
13405 (maybe_gen_llsym, skip_loc_list_entry): New.
13406 (dwarf2out_maybe_output_loclist_view_pair): New.
13407 (output_loc_list): Output view list or entries too.
13408 (output_view_list_offset): New.
13409 (output_die): Handle dw_val_class_view_list.
13410 (output_dwarf_version): New.
13411 (output_compilation_unit_header): Use it.
13412 (output_skeleton_debug_sections): Likewise.
13413 (output_rnglists, output_line_info): Likewise.
13414 (output_pubnames, output_aranges): Update version comments.
13415 (output_one_line_info_table): Output view numbers in asm comments.
13416 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13417 Call maybe_gen_llsym.
13418 (loc_list_from_tree_1): Adjust.
13419 (add_AT_location_description): Create view list attribute if
13420 needed, check it's absent otherwise.
13421 (convert_cfa_to_fb_loc_list): Adjust.
13422 (maybe_emit_file): Call output_asm_line_debug_info for test.
13423 (dwarf2out_var_location): Reset views as needed. Precompute
13424 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13425 attribute. Set view.
13426 (new_line_info_table): Reset next view.
13427 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13428 (dwarf2out_source_line): Likewise. Output view resets and labels to
13429 the assembler, or select appropriate line info opcodes.
13430 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13431 (optimize_string_length): Catch it. Adjust.
13432 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13433 dw_val_class_view_list, and remove it if no longer needed.
13434 (hash_loc_list): Hash view numbers.
13435 (loc_list_hasher::equal): Compare them.
13436 (optimize_location_lists): Check whether a view list symbol is
13437 needed, and whether the locview attribute is present, and
13438 whether they match. Remove the locview attribute if no longer
13440 (index_location_lists): Call skip_loc_list_entry for test.
13441 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13442 Use output_dwarf_version.
13443 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13444 (struct dw_val_node): Add val_view_list.
13445 * final.c (SEEN_NEXT_VIEW): New.
13446 (set_next_view_needed): New.
13447 (clear_next_view_needed): New.
13448 (maybe_output_next_view): New.
13449 (final_start_function): Rename to...
13450 (final_start_function_1): ... this. Take pointer to FIRST,
13451 add SEEN parameter. Emit param bindings in the initial view.
13452 (final_start_function): Reintroduce SEEN-less interface.
13453 (final): Rename to...
13454 (final_1): ... this. Take SEEN parameter. Output final pending
13455 next view at the end.
13456 (final): Reintroduce seen-less interface.
13457 (final_scan_insn): Output pending next view before switching
13458 sections or ending a block. Mark the next view as needed when
13459 outputting variable locations. Notify debug backend of section
13460 changes, and of location view changes.
13461 (rest_of_handle_final): Adjust.
13462 * toplev.c (process_options): Autodetect value for debug variable
13463 location views option. Warn on incompat5 without -gdwarf-5.
13464 * doc/invoke.texi (gvariable-location-views): New.
13465 (gvariable-location-views=incompat5): New.
13466 (gno-variable-location-views): New.
13468 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13470 PR tree-optimization/84136
13471 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13472 that the result of find_edge is non-NULL.
13474 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13477 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13478 storing integer register in SImode. Fix cost of 256 and 512
13479 byte aligned SSE register store.
13481 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13483 * config/i386/i386.c (ix86_multiplication_cost): Fix
13484 multiplication cost for TARGET_AVX512DQ.
13486 2018-02-08 Marek Polacek <polacek@redhat.com>
13488 PR tree-optimization/84238
13489 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13492 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13494 PR tree-optimization/84265
13495 * tree-vect-stmts.c (vectorizable_store): Don't treat
13496 VMAT_CONTIGUOUS accesses as grouped.
13497 (vectorizable_load): Likewise.
13499 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13501 PR tree-optimization/81635
13502 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13503 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13504 (test_round_for_mask): New functions.
13505 (wide_int_cc_tests): Call test_round_for_mask.
13506 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13507 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13508 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13509 range returned by get_range_info.
13511 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13514 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13515 * symtab.c: Include builtins.h
13516 (symtab_node::output_to_lto_symbol_table_p): Move here
13517 from lto-streamer-out.c:output_symbol_p.
13518 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13519 (output_symbol_p): Move all logic to symtab.c
13520 (produce_symtab): Update.
13522 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13524 * config/s390/s390-opts.h (enum indirect_branch): Define.
13525 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13526 (s390_indirect_branch_via_thunk)
13527 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13528 (enum s390_indirect_branch_type): Define.
13529 * config/s390/s390.c (struct s390_frame_layout, struct
13530 machine_function): Remove.
13531 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13532 (indirect_branch_table_label_no, indirect_branch_table_name):
13534 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13535 (enum s390_indirect_branch_option): Define.
13536 (s390_return_addr_from_memory): New function.
13537 (s390_handle_string_attribute): New function.
13538 (s390_attribute_table): Add new attribute handler.
13539 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13540 (s390_indirect_branch_via_thunk): New function.
13541 (s390_indirect_branch_via_inline_thunk): New function.
13542 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13543 sibling call optimization for non z10 compiles.
13544 (s390_emit_call): Force indirect branch target to be a single
13545 register. Add r1 clobber for non-z10 compiles.
13546 (s390_emit_epilogue): Emit return jump via return_use expander.
13547 (s390_reorg): Handle JUMP_INSNs as execute targets.
13548 (s390_option_override_internal): Perform validity checks for the
13549 new command line options.
13550 (s390_indirect_branch_attrvalue): New function.
13551 (s390_indirect_branch_settings): New function.
13552 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13553 (s390_output_indirect_thunk_function): New function.
13554 (s390_code_end): Implement target hook.
13555 (s390_case_values_threshold): Implement target hook.
13556 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13558 * config/s390/s390.h (struct s390_frame_layout)
13559 (struct machine_function): Move here from s390.c.
13560 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13561 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13562 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13563 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13564 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13565 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13566 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13567 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13568 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13569 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13570 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13571 (mnemonic attribute): Add values which aren't recognized
13573 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13574 pattern for branch conversion. Fix mnemonic attribute.
13575 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13576 indirect branch via thunk if requested.
13577 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13578 ("*indirect_jump"): Disable for branch conversion using out of
13580 ("indirect_jump_via_thunk<mode>_z10")
13581 ("indirect_jump_via_thunk<mode>")
13582 ("indirect_jump_via_inlinethunk<mode>_z10")
13583 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13584 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13585 ("casesi_jump_via_inlinethunk<mode>_z10")
13586 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13587 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13588 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13589 ("*indirect2_jump"): Disable for branch conversion.
13590 ("casesi_jump"): Turn into expander and expand patterns for branch
13592 ("return_use"): New expander.
13593 ("*return"): Emit return via thunk and rename it to ...
13594 ("*return<mode>"): ... this one.
13595 * config/s390/s390.opt: Add new options and and enum for the
13598 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13600 * lra-constraints.c (match_reload): Unconditionally use
13601 gen_lowpart_SUBREG, rather than selecting between that
13602 and equivalent gen_rtx_SUBREG code.
13604 2018-02-08 Richard Biener <rguenther@suse.de>
13606 PR tree-optimization/84233
13607 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13608 changed flag instead of boguously re-using phi_inserted.
13610 2018-02-08 Martin Jambor <mjambor@suse.cz>
13612 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13613 static local variables.
13615 2018-02-08 Richard Biener <rguenther@suse.de>
13617 PR tree-optimization/84278
13618 * tree-vect-stmts.c (vectorizable_store): When looking for
13619 smaller vector types to perform grouped strided loads/stores
13620 make sure the mode is supported by the target.
13621 (vectorizable_load): Likewise.
13623 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13625 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13626 Increase LDP/STP opportunities by adding adjacent callee-saves.
13628 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13630 PR rtl-optimization/84068
13631 PR rtl-optimization/83459
13632 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13634 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13636 PR tree-optimization/84224
13637 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13638 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13639 non-zero arguments.
13641 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13644 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13645 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13646 count, remove one USE.
13648 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13650 * doc/install.texi (Configuration): Document the
13651 --with-long-double-format={ibm,ieee} PowerPC configuration
13655 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13656 Convert from define_expand to be define_insn_and_split. Rework
13657 float/double/_Float128 conversions to QI/HI/SImode to work with
13658 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13659 conversions to QI/HImode types did a store and then a load to
13660 truncate the value. For conversions to VSX registers, don't split
13661 the insn, instead emit the code directly. Use the code iterator
13662 any_fix to combine signed and unsigned conversions.
13663 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13664 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13665 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13666 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13667 (fix_<mode>di2_hw): Likewise.
13668 (fixuns_<mode>di2_hw): Likewise.
13669 (fix_<mode>si2_hw): Likewise.
13670 (fixuns_<mode>si2_hw): Likewise.
13671 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13672 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13673 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13674 fix<uns>_trunc<SFDF:mode>si2_p8.
13675 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13676 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13677 (fix<uns>_<mode>_mem): Likewise.
13678 (fctiw<u>z_<mode>_mem): Likewise.
13679 (fix<uns>_<mode>_mem): Likewise.
13680 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13681 the register allocator from doing a direct move to the GPRs to do
13682 a store, and instead use the ISA 3.0 store byte/half-word from
13683 vector register instruction. For IEEE 128-bit floating point,
13684 also optimize stores of 32-bit ints.
13685 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13687 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13689 * genextract.c (push_pathstr_operand): New function to support
13691 (walk_rtx): Call push_pathstr_operand.
13692 (print_path): Support [a-zA-Z].
13694 2018-02-07 Richard Biener <rguenther@suse.de>
13696 PR tree-optimization/84037
13697 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13698 (cse_and_gimplify_to_preheader): Declare.
13699 (vect_get_place_in_interleaving_chain): Likewise.
13700 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13702 (_loop_vec_info::~_loop_vec_info): Delete it.
13703 (cse_and_gimplify_to_preheader): New function.
13704 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13705 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13706 (vectorizable_load): Likewise. For grouped stores always base
13707 the IV on the first element.
13708 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13709 condition before gimplifying.
13711 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13713 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13714 *DIV_EXPR and *MOD_EXPR.
13716 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13719 * config/i386/i386.c (ix86_option_override_internal): Mask out
13720 the CF_SET bit when checking -fcf-protection.
13722 2018-02-07 Tom de Vries <tom@codesourcery.com>
13725 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13728 2018-02-07 Richard Biener <rguenther@suse.de>
13730 PR tree-optimization/84204
13731 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13734 PR tree-optimization/84205
13735 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13736 special-case isl_ast_op_zdiv_r.
13738 PR tree-optimization/84223
13739 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13740 Only add conditions from within the region.
13741 (gather_bbs::after_dom_children): Adjust.
13743 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13746 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13747 * config/avr/avr.md: Only post-reload split REG-REG moves if
13748 either register is GENERAL_REG_P.
13750 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13752 PR tree-optimization/84235
13753 * tree-ssa-scopedtables.c
13754 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13755 if the subtraction is performed in floating point type where NaNs are
13756 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13757 build 1. Formatting fix.
13759 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13762 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13763 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13764 and skip it regardless of bb boundaries. Use CALL_P macro,
13765 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13766 unnecessarily, formatting fix.
13768 2018-02-06 Michael Collison <michael.collison@arm.com>
13770 * config/arm/thumb2.md:
13771 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13772 (*thumb_mov_notscc): Ditto.
13774 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13777 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13778 unsigned_fix, not "s".
13780 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13782 * configure.ac (gcc_fn_eh_frame_ro): New function.
13783 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13784 correct .eh_frame permissions.
13785 * configure: Regenerate.
13787 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13789 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13790 irrelevant options.
13792 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13794 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13795 Display warning message for -mno-speculate-indirect-jumps.
13797 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13799 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13801 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13803 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13805 PR tree-optimization/84225
13806 * tree-eh.c (find_trapping_overflow): Only call
13807 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13809 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13812 * config/i386/i386.c: Reimplement the check of possible options
13813 -mibt/-mshstk conbination. Change error messages.
13814 * doc/invoke.texi: Fix a typo: remove extra '='.
13816 2018-02-06 Marek Polacek <polacek@redhat.com>
13818 PR tree-optimization/84228
13819 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13821 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13824 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13825 emitted arch directives.
13826 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13827 __ARM_FEATURE_COPROC before changing architectures.
13829 2018-02-06 Richard Biener <rguenther@suse.de>
13831 * config/i386/i386.c (print_reg): Fix typo.
13832 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13834 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13836 * configure: Regenerate.
13838 2018-02-05 Martin Sebor <msebor@redhat.com>
13840 PR tree-optimization/83369
13841 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13844 2018-02-05 Martin Liska <mliska@suse.cz>
13846 * doc/invoke.texi: Cherry-pick upstream r323995.
13848 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13850 * ira.c (ira_init_register_move_cost): Adjust comment.
13852 2018-02-05 Martin Liska <mliska@suse.cz>
13854 PR gcov-profile/84137
13855 * doc/gcov.texi: Fix typo in documentation.
13857 2018-02-05 Martin Liska <mliska@suse.cz>
13859 PR gcov-profile/83879
13860 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13861 using dlopen functionality.
13863 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13865 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13866 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13867 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13868 _mm_maskz_range_round_ss): New intrinsics.
13869 (__builtin_ia32_rangesd128_round)
13870 (__builtin_ia32_rangess128_round): Remove.
13871 (__builtin_ia32_rangesd128_mask_round,
13872 __builtin_ia32_rangess128_mask_round): New builtins.
13873 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13874 __builtin_ia32_rangess128_round): Remove.
13875 (__builtin_ia32_rangesd128_mask_round,
13876 __builtin_ia32_rangess128_mask_round): New builtins.
13877 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13878 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13879 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13880 "<round_saeonly_constraint>")): Changed to ...
13881 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13882 "<round_saeonly_scalar_constraint>")): ... this.
13883 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13884 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13885 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13886 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13887 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13889 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13891 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13893 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13894 Remove all values except native, 8540 and 8548.
13896 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13898 * config/i386/i386.c (ix86_output_function_return): Pass
13899 INVALID_REGNUM, instead of -1, as invalid register number to
13900 indirect_thunk_name and output_indirect_thunk.
13902 2018-02-02 Julia Koval <julia.koval@intel.com>
13904 * config.gcc: Add -march=icelake.
13905 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13906 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13907 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13908 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13909 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13910 (processor_target_table): Add icelake.
13911 (ix86_option_override_internal): Handle new PTAs.
13912 (get_builtin_code_for_version): Handle icelake.
13913 (M_INTEL_COREI7_ICELAKE): New.
13914 (fold_builtin_cpu): Handle icelake.
13915 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13916 * doc/invoke.texi: Add -march=icelake.
13918 2018-02-02 Julia Koval <julia.koval@intel.com>
13920 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13921 to wide_int_bitmask.
13922 * wide-int-bitmask.h: New.
13924 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13927 * config/i386/i386.md: Replace Pmode with word_mode in
13928 builtin_setjmp_setup and builtin_longjmp to support x32.
13930 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13934 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13936 (rs6000_supported_cpu_names): New static variable.
13937 (linux_cpu_translation_table): Likewise.
13938 (elf_platform) <cpu>: Define new static variable and use it.
13939 Translate kernel AT_PLATFORM name to canonical name if needed.
13940 Error if platform name is unknown.
13942 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13945 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13947 2018-02-01 Jeff Law <law@redhat.com>
13950 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13951 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13952 the scratch if RELEASE_VIA_POP is false.
13953 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13954 If we have to save a temporary register, decrement SIZE appropriately.
13955 Pass new arguments to release_scratch_register_on_entry.
13956 (ix86_adjust_stack_and_probe): Likewise.
13957 (ix86_emit_probe_stack_range): Pass new arguments to
13958 release_scratch_register_on_entry.
13960 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13962 PR rtl-optimization/84157
13963 * combine.c (change_zero_ext): Use REG_P predicate in
13964 front of HARD_REGISTER_P predicate.
13966 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13968 * config/avr/avr.c (avr_option_override): Move disabling of
13969 -fdelete-null-pointer-checks to...
13970 * common/config/avr/avr-common.c (avr_option_optimization_table):
13973 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13975 PR tree-optimization/81635
13976 * tree-data-ref.c (split_constant_offset_1): For types that
13977 wrap on overflow, try to use range info to prove that wrapping
13980 2018-02-01 Renlin Li <renlin.li@arm.com>
13983 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13984 TAILCALL_ADDR_REGS.
13985 (aarch64_register_move_cost): Likewise.
13986 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13987 TAILCALL_ADDR_REGS.
13988 (REG_CLASS_NAMES): Likewise.
13989 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13990 TAILCALL_ADDR_REGS. Remove IP registers.
13991 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13993 2018-02-01 Richard Biener <rguenther@suse.de>
13995 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13996 for specifying RPO order and allow NULL for that.
13997 * domwalk.c (dom_walker::dom_walker): Likewise.
13998 (dom_walker::walk): Handle NULL RPO order.
13999 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14001 (rewrite_update_dom_walker): Likewise.
14002 (mark_def_dom_walker): Likewise.
14004 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14006 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14007 (aarch64_maybe_expand_sve_subreg_move): Declare.
14008 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14009 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14011 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14012 that are semantically a reverse operation.
14013 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14014 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14015 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14017 (aarch64_can_change_mode_class): For big-endian, forbid changes
14018 between two SVE modes if they have different element sizes.
14020 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14022 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14023 the TImode handling for big-endian targets.
14025 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14027 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14028 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14030 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14031 Remove BSWAP handing for big-endian targets and use the form of
14032 LD1RQ appropariate for the mode.
14034 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14036 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14037 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14038 duplicated element.
14040 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14043 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14044 check for operands that need to go through aarch64_sve_reload_be.
14046 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14048 PR tree-optimization/81661
14049 PR tree-optimization/84117
14050 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14051 * tree-eh.c: Include gimplify.h.
14052 (find_trapping_overflow, replace_trapping_overflow,
14053 rewrite_to_non_trapping_overflow): New functions.
14054 * tree-vect-loop.c: Include tree-eh.h.
14055 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14056 * tree-data-ref.c: Include tree-eh.h.
14057 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14059 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14061 PR rtl-optimization/84123
14062 * combine.c (change_zero_ext): Check if hard register satisfies
14063 can_change_dest_mode before calling gen_lowpart_SUBREG.
14065 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14068 * ira.c (ira_init_register_move_cost): Remove assert.
14070 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14072 PR rtl-optimization/84071
14073 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14074 * doc/tm.texi: Regenerate.
14076 2018-01-31 Richard Biener <rguenther@suse.de>
14078 PR tree-optimization/84132
14079 * tree-data-ref.c (analyze_miv_subscript): Properly
14080 check whether evolution_function_is_affine_multivariate_p
14081 before calling gcd_of_steps_may_divide_p.
14083 2018-01-31 Julia Koval <julia.koval@intel.com>
14086 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14087 * config/i386/i386.md (rdpid_rex64) New.
14088 (rdpid): Make 32bit only.
14090 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14093 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14094 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14096 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14099 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14101 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14103 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14105 PR rtl-optimization/84071
14106 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14107 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14109 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14111 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14112 (arc_attribute_table): Add 'aux' attribute.
14113 (arc_in_small_data_p): Consider aux like variables.
14114 (arc_is_aux_reg_p): New function.
14115 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14116 (arc_get_aux_arg): New function.
14117 (prepare_move_operands): Handle aux-register access.
14118 (arc_handle_aux_attribute): New function.
14119 * doc/extend.texi (ARC Variable attributes): Add subsection.
14121 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14123 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14124 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14125 (arc_attribute_table): Add 'uncached' attribute.
14126 (arc_print_operand): Print '.di' flag for uncached memory
14128 (arc_in_small_data_p): Do not consider for small data the uncached
14130 (arc_is_uncached_mem_p): New function.
14131 * config/arc/predicates.md (compact_store_memory_operand): Check
14132 for uncached memory accesses.
14133 (nonvol_nonimm_operand): Likewise.
14134 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14136 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14139 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14140 falign-loops=): Add Optimization flag.
14142 2018-01-30 Jeff Law <law@redhat.com>
14145 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14146 INT_REGISTERS_SAVED. Check it prior to calling
14147 get_scratch_register_on_entry.
14148 (ix86_adjust_stack_and_probe): Similarly.
14149 (ix86_emit_probe_stack_range): Similarly.
14150 (ix86_expand_prologue): Corresponding changes.
14152 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14155 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14156 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14158 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14161 * lra-constraints.c (curr_insn_transform): Process AND in the
14164 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14166 PR rtl-optimization/83986
14167 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14168 dependence against last_pending_memory_flush in addition to
14169 pending_jump_insns.
14171 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14173 PR tree-optimization/81611
14174 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14177 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14180 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14183 2018-01-30 Richard Biener <rguenther@suse.de>
14184 Jakub Jelinek <jakub@redhat.com>
14186 PR tree-optimization/84111
14187 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14188 inner loops added during recursion, as they don't have up-to-date
14191 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14194 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14195 (can_inline_edge_by_limits_p): ... here.
14196 (can_early_inline_edge_p, check_callers,
14197 update_caller_keys, update_callee_keys, recursive_inlining,
14198 add_new_edges_to_heap, speculation_useful_p,
14199 inline_small_functions,
14200 inline_small_functions, flatten_function,
14201 inline_to_all_callers_1): Update.
14203 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14205 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14208 2018-01-30 Richard Biener <rguenther@suse.de>
14210 PR tree-optimization/83008
14211 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14212 invariant and constant vector uses in stmts when they need
14213 more than one stmt.
14215 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14218 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14219 * configure: Regenerate.
14221 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14223 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14225 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14226 Use gen_rtx_REG rather than gen_lowpart.
14228 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14230 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14231 rather than 0 when creating partial subregs.
14233 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14235 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14238 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14241 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14242 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14243 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14244 flags. This restores the settings used before the 2017-07-24.
14245 Turning off pre increment/decrement/modify allows IVOPTS to
14246 optimize DF/SF loops where the index is an int.
14248 2018-01-29 Richard Biener <rguenther@suse.de>
14249 Kelvin Nilsen <kelvin@gcc.gnu.org>
14252 * tree-vect-stmts.c (vectorizable_call): Don't call
14253 targetm.vectorize_builtin_md_vectorized_function if callee is
14256 2018-01-22 Carl Love <cel@us.ibm.com>
14258 * doc/extend.tex: Fix typo in second arg in
14259 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14261 2018-01-29 Richard Biener <rguenther@suse.de>
14263 PR tree-optimization/84086
14264 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14265 (flush_ssaname_freelist): When SSA names were released reset
14266 the SCEV hash table.
14268 2018-01-29 Richard Biener <rguenther@suse.de>
14270 PR tree-optimization/84057
14271 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14272 removed paths when removing edges.
14274 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14276 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14277 -mfunction-return=@var{choice}.
14279 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14281 PR diagnostic/84034
14282 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14283 Handle CR like TAB.
14284 (layout::print_source_line): Likewise.
14285 (test_get_line_width_without_trailing_whitespace): Add test cases.
14287 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14289 PR middle-end/84040
14290 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14293 2018-01-26 Jim Wilson <jimw@sifive.com>
14295 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14297 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14300 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14302 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14303 and CMP + SUB-immediate -> SUBS.
14305 2018-01-26 Martin Sebor <msebor@redhat.com>
14307 PR tree-optimization/83896
14308 * tree-ssa-strlen.c (get_string_len): Rename...
14309 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14310 Avoid assuming length is constant.
14311 (handle_char_store): Use HOST_WIDE_INT for string length.
14313 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14316 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14317 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14319 2018-01-26 Richard Biener <rguenther@suse.de>
14321 PR rtl-optimization/84003
14322 * dse.c (record_store): Only record redundant stores when
14323 the earlier store aliases at least all accesses the later one does.
14325 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14327 PR rtl-optimization/83985
14328 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14329 REG_CFA_RESTORE insns.
14330 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14331 REG_CFA_RESTORE insns here.
14334 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14335 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14337 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14339 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14340 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14341 (arc_init): Likewise.
14342 (arc_override_options): Likewise.
14343 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14345 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14347 * config/arc/arc.h (TARGET_DBNZ): Define.
14348 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14349 properly set the tune attribute.
14350 (dbnz): Use TARGET_DBNZ guard.
14351 * config/arc/arc.opt (mtune): Add core3 option.
14353 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14355 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14356 recognize new pic like addresses.
14357 (arc_delegitimize_address): Clean up.
14359 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14361 * config/arc/arc-arches.def: Option mrf16 valid for all
14363 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14364 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14365 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14366 * config/arc/arc-tables.opt: Regenerate.
14367 * config/arc/arc.c (arc_conditional_register_usage): Handle
14368 reduced register file case.
14369 (arc_file_start): Set must have build attributes.
14370 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14371 mrf16 option value.
14372 * config/arc/arc.opt (mrf16): Add new option.
14373 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14374 * config/arc/genmultilib.awk: Handle new mrf16 option.
14375 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14376 * config/arc/t-multilib: Regenerate.
14377 * doc/invoke.texi (ARC Options): Document mrf16 option.
14379 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14381 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14382 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14383 (arc_attribute_table): Add 'secure_call' attribute.
14384 (arc_print_operand): Print secure call operand.
14385 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14387 (arc_is_secure_call_p): New function. * config/arc/arc.md
14388 (call_i): Add support for sjli instruction.
14389 (call_value_i): Likewise.
14390 * config/arc/constraints.md (Csc): New constraint.
14392 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14393 John Eric Martin <John.Martin@emmicro-us.com>
14395 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14396 * config/arc/arc.c (_arc_jli_section): New struct.
14397 (arc_jli_section): New type.
14398 (rc_jli_sections): New static variable.
14399 (arc_handle_jli_attribute): New function.
14400 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14401 (arc_file_end): New function.
14402 (TARGET_ASM_FILE_END): Define.
14403 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14404 (arc_add_jli_section): New function.
14405 (jli_call_scan): Likewise.
14406 (arc_reorg): Call jli_call_scan.
14407 (arc_output_addsi): Remove 'S' from printing asm operand.
14408 (arc_is_jli_call_p): New function.
14409 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14411 (movhi_insn): Likewise.
14412 (movsi_insn): Likewise.
14413 (movsi_set_cc_insn): Likewise.
14414 (loadqi_update): Likewise.
14415 (load_zeroextendqisi_update): Likewise.
14416 (load_signextendqisi_update): Likewise.
14417 (loadhi_update): Likewise.
14418 (load_zeroextendhisi_update): Likewise.
14419 (load_signextendhisi_update): Likewise.
14420 (loadsi_update): Likewise.
14421 (loadsf_update): Likewise.
14422 (movsicc_insn): Likewise.
14423 (bset_insn): Likewise.
14424 (bxor_insn): Likewise.
14425 (bclr_insn): Likewise.
14426 (bmsk_insn): Likewise.
14427 (bicsi3_insn): Likewise.
14428 (cmpsi_cc_c_insn): Likewise.
14429 (movsi_ne): Likewise.
14430 (movsi_cond_exec): Likewise.
14431 (clrsbsi2): Likewise.
14432 (norm_f): Likewise.
14442 (call_i): Remove 'S' asm letter, add jli instruction.
14443 (call_value_i): Likewise.
14444 * config/arc/arc.op (mjli-always): New option.
14445 * config/arc/constraints.md (Cji): New constraint.
14446 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14448 (subsf3_fpx): Likewise.
14449 (mulsf3_fpx): Likewise.
14450 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14452 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14453 function attrbutes.
14454 * doc/invoke.texi (ARC): Document mjli-always option.
14456 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14458 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14459 avoid addition with 0 and use incw and decw where possible.
14461 2018-01-26 Richard Biener <rguenther@suse.de>
14463 PR tree-optimization/81082
14464 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14465 association if it requires casting to unsigned.
14466 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14467 from fold_plusminus_mult_expr to catch important cases late when
14468 range info is available.
14470 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14472 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14473 * configure.ac (hidden_linkonce): New test.
14474 * configure: Regenerate.
14475 * config.in: Regenerate.
14477 2018-01-26 Julia Koval <julia.koval@intel.com>
14479 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14480 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14481 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14482 _mm_mask_bitshuffle_epi64_mask): Fix type.
14483 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14484 USI_FTYPE_V4DI_V4DI_USI): Remove.
14485 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14486 __builtin_ia32_vpshufbitqmb256_mask,
14487 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14488 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14489 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14491 2018-01-26 Alan Modra <amodra@gmail.com>
14494 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14495 UNSPEC_VBPERMQ. Sort other unspecs.
14497 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14499 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14501 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14503 PR middle-end/83055
14504 * predict.c (drop_profile): Do not push/pop cfun; update also
14506 (handle_missing_profiles): Fix logic looking for zero profiles.
14508 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14510 PR middle-end/83977
14511 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14512 on functions with #pragma omp declare simd or functions with simd
14514 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14515 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14516 Remove trailing \n from warning_at calls.
14518 2018-01-25 Tom de Vries <tom@codesourcery.com>
14521 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14522 for neutered workers.
14524 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14527 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14528 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14530 2018-01-24 Jeff Law <law@redhat.com>
14533 * i386.c (get_probe_interval): Move to earlier point.
14534 (ix86_compute_frame_layout): If -fstack-clash-protection and
14535 the frame is larger than the probe interval, then use pushes
14536 to save registers rather than reg->mem moves.
14537 (ix86_expand_prologue): Remove conditional for int_registers_saved
14540 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14543 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14544 min/max for never referenced object.
14546 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14548 PR middle-end/83977
14549 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14551 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14552 attributes from DECL_ATTRIBUTES (decl) without affecting
14553 DECL_ATTRIBUTES (current_function_decl).
14554 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14555 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14557 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14559 PR tree-optimization/83979
14560 * fold-const.c (fold_comparison): Use constant_boolean_node
14561 instead of boolean_{true,false}_node.
14563 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14565 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14568 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14570 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14571 Simplify the clause that sets the length attribute.
14572 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14573 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14574 clause that sets the length attribute.
14575 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14577 2018-01-24 Tom de Vries <tom@codesourcery.com>
14580 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14581 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14582 Add strict parameter.
14583 (prevent_branch_around_nothing): Insert dummy insn between branch to
14584 label and label with no ptx insn inbetween.
14585 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14587 2018-01-24 Tom de Vries <tom@codesourcery.com>
14590 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14591 for neutered threads in warp.
14592 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14594 2018-01-24 Richard Biener <rguenther@suse.de>
14596 PR tree-optimization/83176
14597 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14600 2018-01-24 Richard Biener <rguenther@suse.de>
14602 PR tree-optimization/82819
14603 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14604 code generating pluses that are no-ops in the target precision.
14606 2018-01-24 Richard Biener <rguenther@suse.de>
14608 PR middle-end/84000
14609 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14611 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14613 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14614 to merge probabilities.
14615 * predict.c (probably_never_executed): Also mark as cold functions
14616 with global 0 profile and guessed local profile.
14617 * profile-count.c (profile_probability::combine_with_count): New
14619 * profile-count.h (profile_probability::operator*,
14620 profile_probability::operator*=, profile_probability::operator/,
14621 profile_probability::operator/=): Reduce precision to adjusted
14622 and set value to guessed on contradictory divisions.
14623 (profile_probability::combine_with_freq): Remove.
14624 (profile_probability::combine_wiht_count): Declare.
14625 (profile_count::force_nonzero):: Set to adjusted.
14626 (profile_count::probability_in):: Set quality to adjusted.
14627 * tree-ssa-tail-merge.c (replace_block_by): Use
14628 combine_with_count.
14630 2018-01-23 Andrew Waterman <andrew@sifive.com>
14631 Jim Wilson <jimw@sifive.com>
14633 * config/riscv/riscv.c (riscv_stack_boundary): New.
14634 (riscv_option_override): Set riscv_stack_boundary. Handle
14635 riscv_preferred_stack_boundary_arg.
14636 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14637 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14638 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14639 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14640 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14641 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14643 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14646 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14647 of struct ix86_frame.
14648 (ix86_expand_epilogue): Likewise. Add a local variable for
14649 the reg_save_offset field in struct ix86_frame.
14651 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14653 PR tree-optimization/82604
14654 * tree-loop-distribution.c (enum partition_kind): New enum item
14655 PKIND_PARTIAL_MEMSET.
14656 (partition_builtin_p): Support above new enum item.
14657 (generate_code_for_partition): Ditto.
14658 (compute_access_range): Differentiate cases that equality can be
14659 proven at all loops, the innermost loops or no loops.
14660 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14661 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14662 (finalize_partitions, distribute_loop): Don't fuse partition of
14663 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14664 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14665 parloop is enabled.
14667 2018-01-23 Martin Liska <mliska@suse.cz>
14669 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14670 order to ignore the predictor.
14671 (PRED_POLYMORPHIC_CALL): Likewise.
14672 (PRED_RECURSIVE_CALL): Likewise.
14674 2018-01-23 Martin Liska <mliska@suse.cz>
14676 * tree-profile.c (tree_profiling): Print function header to
14677 aware reader which function we are working on.
14678 * value-prof.c (gimple_find_values_to_profile): Do not print
14679 not interesting value histograms.
14681 2018-01-23 Martin Liska <mliska@suse.cz>
14683 * profile-count.h (enum profile_quality): Add
14684 profile_uninitialized as the first value. Do not number values
14685 as they are zero based.
14686 (profile_count::verify): Update sanity check.
14687 (profile_probability::verify): Likewise.
14689 2018-01-23 Nathan Sidwell <nathan@acm.org>
14691 * doc/invoke.texi (ffor-scope): Deprecate.
14693 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14695 PR tree-optimization/83510
14696 * domwalk.c (set_all_edges_as_executable): New function.
14697 (dom_walker::dom_walker): Convert bool param
14698 "skip_unreachable_blocks" to enum reachability. Move setup of
14699 edge flags to set_all_edges_as_executable and only do it when
14700 reachability is REACHABLE_BLOCKS.
14701 * domwalk.h (enum dom_walker::reachability): New enum.
14702 (dom_walker::dom_walker): Convert bool param
14703 "skip_unreachable_blocks" to enum reachability.
14704 (set_all_edges_as_executable): New decl.
14705 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14706 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14708 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14709 but converting true to REACHABLE_BLOCKS.
14710 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14712 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14713 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14714 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14716 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14717 if check_all_array_refs will be called.
14719 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14721 * tree.c (selftest::test_location_wrappers): Add more test
14724 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14726 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14727 (selftest::test_bit_in_range): Likewise.
14729 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14732 * doc/sourcebuild.texi (vect_float): Say that the selector
14733 only describes the situation when -funsafe-math-optimizations is on.
14734 (vect_float_strict): Document.
14736 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14738 PR tree-optimization/83965
14739 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14740 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14741 instead of checking only for a reduction.
14742 (vect_recog_widen_sum_pattern): Likewise.
14744 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14746 * predict.c (probably_never_executed): Only use precise profile info.
14747 (compute_function_frequency): Skip after inlining hack since we now
14748 have quality checking.
14750 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14752 * profile-count.h (profile_probability::very_unlikely,
14753 profile_probability::unlikely, profile_probability::even): Set
14754 precision to guessed.
14756 2018-01-23 Richard Biener <rguenther@suse.de>
14758 PR tree-optimization/83963
14759 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14760 Properly terminate dominator walk when crossing the exit edge not
14761 when visiting its source block.
14763 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14766 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14767 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14769 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14771 PR tree-optimization/83957
14772 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14773 semicolon after for body surrounded by braces.
14775 PR tree-optimization/83081
14776 * profile-count.h (profile_probability::split): New method.
14777 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14778 Use profile_probability::split.
14779 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14780 when splitting a single conditional jump into 2.
14782 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14784 PR tree-optimization/69452
14785 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14788 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14790 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14791 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14792 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14794 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14796 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14798 * config/rl78/rl78.md (movdi): New define_expand.
14799 * config/rl78/rl78.c (rl78_split_movdi): New function.
14801 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14804 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14806 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14807 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14808 128-bit to produce an UNSPEC move to get the double word with the
14809 signbit and then a shift directly to do signbit.
14810 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14811 implementation with a new version that just does either a direct
14812 move or a regular move. Move memory interface to separate insns.
14813 Move insns so they are next to the expander.
14814 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14815 with signbit move. Split big and little endian case.
14816 (signbit<mode>2_dm_mem_le): Likewise.
14817 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14818 (signbit<mode>2_dm2): Likewise.
14820 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14822 * config/rl78/rl78.md (anddi3): New define_expand.
14824 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14826 * config/rl78/rl78.md (umindi3): New define_expand.
14828 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14830 * config/rl78/rl78.md (smindi3): New define_expand.
14832 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14834 * config/rl78/rl78.md (smaxdi3): New define_expand.
14836 2018-01-22 Carl Love <cel@us.ibm.com>
14838 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14839 LVX_V1TI): Add macro expansion.
14840 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14841 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14842 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14843 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14844 Change check to determine if the instruction is a byte reversing
14845 entry. Fix typo in comment.
14846 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14847 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14848 Add def_builtin calls for new builtins.
14849 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14850 Add define_insn expansion.
14852 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14854 * config/rl78/rl78.md (umaxdi3): New define_expand.
14856 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14858 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14859 for non-QImode registers.
14861 2018-01-22 Richard Biener <rguenther@suse.de>
14863 PR tree-optimization/83963
14864 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14865 including the loop exit block.
14866 (scop_detection::merge_sese): Likewise.
14867 (scop_detection::add_scop): Do it here instead.
14869 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14871 * doc/sourcebuild.texi (arm_softfloat): Document.
14873 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14876 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14877 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14878 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14880 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14881 David Edelsohn <dje.gcc@gmail.com>
14884 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14885 Change "crset eq" to "crset 2".
14886 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14887 (*call_indirect_aix<mode>_nospec): Likewise.
14888 (*call_value_indirect_aix<mode>_nospec): Likewise.
14889 (*call_indirect_elfv2<mode>_nospec): Likewise.
14890 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14891 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14892 change assembly output from . to $.
14893 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14894 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14895 (*tablejump<mode>_internal1_nospec): Likewise.
14897 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14900 * config/sh/sh_optimize_sett_clrt.cc:
14901 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14903 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14905 PR tree-optimization/83940
14906 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14907 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14908 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14909 use it to pass back the definition type.
14910 (vect_check_store_rhs): Likewise rhs_dt_out.
14911 (vect_build_gather_load_calls): Add a mask_dt argument and use
14912 it instead of a call to vect_is_simple_use.
14913 (vectorizable_store): Update calls to vect_check_load_store_mask
14914 and vect_check_store_rhs. Use the dt returned by the latter instead
14915 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14916 instead of calls to vect_is_simple_use. Pass the scalar rather
14917 than the vector operand to vect_is_simple_use when handling
14918 second and subsequent copies of an rhs value.
14919 (vectorizable_load): Update calls to vect_check_load_store_mask
14920 and vect_build_gather_load_calls. Use the cached mask_dt and
14921 gs_info.offset_dt instead of calls to vect_is_simple_use.
14923 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14925 PR middle-end/83945
14926 * tree-emutls.c: Include gimplify.h.
14927 (lower_emutls_2): New function.
14928 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14929 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14930 it before further processing.
14933 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14934 UINTVAL (trueop1) instead of INTVAL (op1).
14936 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14940 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14941 INCOMING_FRAME_SP_OFFSET if not defined.
14942 (scan_trace): Add ENTRY argument. If true and
14943 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14944 emit a note to adjust the CFA offset.
14945 (create_cfi_notes): Adjust scan_trace callers.
14946 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14947 INCOMING_FRAME_SP_OFFSET in the CIE.
14948 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14949 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14951 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14952 * doc/tm.texi: Regenerated.
14954 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14956 PR rtl-optimization/83147
14957 * lra-constraints.c (remove_inheritance_pseudos): Use
14958 lra_substitute_pseudo_within_insn.
14960 2018-01-19 Tom de Vries <tom@codesourcery.com>
14961 Cesar Philippidis <cesar@codesourcery.com>
14964 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14966 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14969 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14970 spaces for function labels.
14972 2018-01-19 Martin Liska <mliska@suse.cz>
14974 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14975 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14976 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14977 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14978 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14979 (PRED_CONST_RETURN): Change from 69 to 65.
14980 (PRED_NULL_RETURN): Change from 91 to 71.
14981 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14982 (PRED_LOOP_GUARD): Change from 66 to 73.
14984 2018-01-19 Martin Liska <mliska@suse.cz>
14986 * predict.c (predict_insn_def): Add new assert.
14987 (struct branch_predictor): Change type to signed integer.
14988 (test_prediction_value_range): Amend test to cover
14989 PROB_UNINITIALIZED.
14990 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14991 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14992 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14993 (PRED_LOOP_IV_COMPARE): Likewise.
14994 * predict.h (PROB_UNINITIALIZED): Define new constant.
14996 2018-01-19 Martin Liska <mliska@suse.cz>
14998 * predict.c (dump_prediction): Add new format for
14999 analyze_brprob.py script which is enabled with -details
15001 * profile-count.h (precise_p): New function.
15003 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15005 PR tree-optimization/83922
15006 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15007 there are no statements that need masking.
15008 (vect_active_double_reduction_p): New function.
15009 (vect_analyze_loop_operations): Use it when handling phis that
15010 are not in the loop header.
15012 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15014 PR tree-optimization/83914
15015 * tree-vect-loop.c (vectorizable_induction): Don't convert
15016 init_expr or apply the peeling adjustment for inductions
15017 that are nested within the vectorized loop.
15019 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15021 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15024 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15028 * function.h (gimplify_parameters): Add gimple_seq * argument.
15029 * function.c: Include gimple.h and options.h.
15030 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15031 for the added local temporaries if needed.
15032 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15033 if there are any parameter cleanups, wrap whole body into a
15034 try/finally with the cleanups.
15036 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15039 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15040 Use GET_MODE_CLASS for scalar floating point.
15042 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15046 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15047 Fix call of call_cgraph_insertion_hooks.
15049 2018-01-18 Martin Sebor <msebor@redhat.com>
15051 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15053 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15056 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15059 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15062 * common.opt: (-ffile-prefix-map): New option.
15063 * opts.c (common_handle_option): Defer it.
15064 * opts-global.c (handle_common_deferred_options): Handle it.
15065 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15066 * file-prefix-map.h: New file.
15067 (remap_debug_filename, add_debug_prefix_map): ...here.
15068 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15069 * final.c (debug_prefix_map, add_debug_prefix_map
15070 remap_debug_filename): Move to...
15071 * file-prefix-map.c: New file.
15072 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15073 generalize, get rid of alloca(), use strrchr() instead of strchr().
15074 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15075 Implement in terms of add_prefix_map().
15076 (remap_macro_filename, remap_debug_filename): Implement in term of
15078 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15079 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15080 * dbxout.c: Include file-prefix-map.h.
15081 * varasm.c: Likewise.
15082 * vmsdbgout.c: Likewise.
15083 * xcoffout.c: Likewise.
15084 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15085 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15086 * doc/invoke.texi (-ffile-prefix-map): Document.
15087 (-fdebug-prefix-map): Update description.
15089 2018-01-18 Martin Liska <mliska@suse.cz>
15091 * config/i386/i386.c (indirect_thunk_name): Document that also
15093 (output_indirect_thunk): Document why both instructions
15094 (pause and lfence) are generated.
15096 2018-01-18 Richard Biener <rguenther@suse.de>
15098 PR tree-optimization/83887
15099 * graphite-scop-detection.c
15100 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15101 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15102 (scop_detection::merge_sese): Re-implement with a flood-fill
15103 algorithm that properly finds a SESE region if it exists.
15105 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15108 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15109 pointer_diff optimizations use view_convert instead of convert.
15111 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15113 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15114 Generate different code for -mno-speculate-indirect-jumps.
15115 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15116 (*call_indirect_aix<mode>): Disable for
15117 -mno-speculate-indirect-jumps.
15118 (*call_indirect_aix<mode>_nospec): New define_insn.
15119 (*call_value_indirect_aix<mode>): Disable for
15120 -mno-speculate-indirect-jumps.
15121 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15122 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15123 -mno-speculate-indirect-jumps.
15124 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15126 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15128 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15129 long double type, set the flags for noting the default long double
15130 type, even if we don't pass or return a long double type.
15132 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15135 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15138 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15140 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15141 support for merge[hl].
15142 (fold_mergehl_helper): New helper function.
15143 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15144 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15145 (altivec_vmrglw_direct): Add xxmrglw insn.
15147 2018-01-17 Andrew Waterman <andrew@sifive.com>
15149 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15150 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15152 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15155 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15156 call the lto_location_cache before reading the
15157 DECL_SOURCE_LOCATION of the types.
15159 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15160 Richard Sandiford <richard.sandiford@linaro.org>
15162 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15163 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15164 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15165 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15166 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15168 * config/aarch64/constraints.md (aarch64_movti_operand):
15170 * config/aarch64/predicates.md (Uti): Add new constraint.
15172 2018-01-17 Carl Love <cel@us.ibm.com>
15174 * config/rs6000/vsx.md (define_expand xl_len_r,
15175 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15176 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15178 (define_expand, define_insn): Move the shift left from the
15179 define_insn to the define_expand for lxvl and stxvl instructions.
15180 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15181 and XL_LEN_R definitions to PURE.
15183 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15185 * config/i386/i386.c (indirect_thunk_name): Declare regno
15186 as unsigned int. Compare regno with INVALID_REGNUM.
15187 (output_indirect_thunk): Ditto.
15188 (output_indirect_thunk_function): Ditto.
15189 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15190 in the call to output_indirect_thunk_function.
15192 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15194 PR middle-end/83884
15195 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15196 rather than the size of inner_type to determine the stack slot size
15197 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15199 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15202 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15205 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15207 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15208 endian Linux systems to optionally enable multilibs for selecting
15209 the long double type if the user configured an explicit type.
15210 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15211 have no long double multilibs if not defined.
15212 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15213 warn if the user used -mabi={ieee,ibm}longdouble and we built
15214 multilibs for long double.
15215 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15216 appropriate multilib option.
15217 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15219 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15220 for building long double multilibs.
15221 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15223 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15225 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15228 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15230 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15233 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15236 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15239 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15241 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15242 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15244 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15246 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15247 different rtl trees depending on TARGET_64BIT.
15248 (rs6000_gen_lvx): Likewise.
15250 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15252 * config/visium/visium.md (nop): Tweak comment.
15253 (hazard_nop): Likewise.
15255 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15257 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15258 -mspeculate-indirect-jumps.
15259 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15260 for -mno-speculate-indirect-jumps.
15261 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15262 (*call_value_indirect_elfv2<mode>): Disable for
15263 -mno-speculate-indirect-jumps.
15264 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15265 (indirect_jump): Emit different RTL for
15266 -mno-speculate-indirect-jumps.
15267 (*indirect_jump<mode>): Disable for
15268 -mno-speculate-indirect-jumps.
15269 (*indirect_jump<mode>_nospec): New define_insn.
15270 (tablejump): Emit different RTL for
15271 -mno-speculate-indirect-jumps.
15272 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15273 (tablejumpsi_nospec): New define_expand.
15274 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15275 (tablejumpdi_nospec): New define_expand.
15276 (*tablejump<mode>_internal1): Disable for
15277 -mno-speculate-indirect-jumps.
15278 (*tablejump<mode>_internal1_nospec): New define_insn.
15279 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15282 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15284 * caller-save.c (insert_save): Drop unnecessary parameter. All
15287 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15288 Richard Biener <rguenth@suse.de>
15291 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15292 return early, inline manually is_gimple_sizepos. Make sure if we
15293 call gimplify_expr we don't end up with a gimple constant.
15294 * tree.c (variably_modified_type_p): Don't return true for
15295 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15296 * gimplify.h (is_gimple_sizepos): Remove.
15298 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15300 PR tree-optimization/83857
15301 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15302 vectorizable_live_operation for pure SLP statements.
15303 (vectorizable_live_operation): Handle PHIs.
15305 2018-01-16 Richard Biener <rguenther@suse.de>
15307 PR tree-optimization/83867
15308 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15309 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15311 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15314 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15315 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15316 If off is not INTEGER_CST, issue a may not be aligned warning
15317 rather than isn't aligned. Use isn%'t rather than isn't.
15318 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15320 <case MULT_EXPR>: Improve the case when bottom and one of the
15321 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15322 operand, in that case check if the other operand is multiple of
15323 bottom divided by the INTEGER_CST operand.
15325 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15328 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15329 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15330 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15331 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15332 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15333 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15334 (pa_function_arg_size): New function.
15336 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15338 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15339 in a separate statement.
15341 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15343 PR tree-optimization/83847
15344 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15345 group gathers and scatters.
15347 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15349 PR rtl-optimization/86620
15350 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15352 PR rtl-optimization/83213
15353 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15354 to last if both are JUMP_INSNs.
15356 PR tree-optimization/83843
15357 * gimple-ssa-store-merging.c
15358 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15359 store_immediate_info for bswap/nop orig_stores.
15361 2018-01-15 Andrew Waterman <andrew@sifive.com>
15363 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15365 <UDIV>: Increase cost if !TARGET_DIV.
15367 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15369 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15370 (define_attr "cr_logical_3op"): New.
15371 (cceq_ior_compare): Adjust.
15372 (cceq_ior_compare_complement): Adjust.
15373 (*cceq_rev_compare): Adjust.
15374 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15375 (is_cracked_insn): Adjust.
15376 (insn_must_be_first_in_group): Adjust.
15377 * config/rs6000/40x.md: Adjust.
15378 * config/rs6000/440.md: Adjust.
15379 * config/rs6000/476.md: Adjust.
15380 * config/rs6000/601.md: Adjust.
15381 * config/rs6000/603.md: Adjust.
15382 * config/rs6000/6xx.md: Adjust.
15383 * config/rs6000/7450.md: Adjust.
15384 * config/rs6000/7xx.md: Adjust.
15385 * config/rs6000/8540.md: Adjust.
15386 * config/rs6000/cell.md: Adjust.
15387 * config/rs6000/e300c2c3.md: Adjust.
15388 * config/rs6000/e500mc.md: Adjust.
15389 * config/rs6000/e500mc64.md: Adjust.
15390 * config/rs6000/e5500.md: Adjust.
15391 * config/rs6000/e6500.md: Adjust.
15392 * config/rs6000/mpc.md: Adjust.
15393 * config/rs6000/power4.md: Adjust.
15394 * config/rs6000/power5.md: Adjust.
15395 * config/rs6000/power6.md: Adjust.
15396 * config/rs6000/power7.md: Adjust.
15397 * config/rs6000/power8.md: Adjust.
15398 * config/rs6000/power9.md: Adjust.
15399 * config/rs6000/rs64.md: Adjust.
15400 * config/rs6000/titan.md: Adjust.
15402 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15404 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15405 ix86_indirect_branch_register logic.
15407 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15409 * config/i386/constraints.md (Bs): Update
15410 ix86_indirect_branch_register check. Don't check
15411 ix86_indirect_branch_register with GOT_memory_operand.
15413 * config/i386/predicates.md (GOT_memory_operand): Don't check
15414 ix86_indirect_branch_register here.
15415 (GOT32_symbol_operand): Likewise.
15417 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15419 * config/i386/predicates.md (constant_call_address_operand):
15420 Rewrite ix86_indirect_branch_register logic.
15421 (sibcall_insn_operand): Likewise.
15423 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15425 * config/i386/constraints.md (Bs): Replace
15426 ix86_indirect_branch_thunk_register with
15427 ix86_indirect_branch_register.
15429 * config/i386/i386.md (indirect_jump): Likewise.
15430 (tablejump): Likewise.
15431 (*sibcall_memory): Likewise.
15432 (*sibcall_value_memory): Likewise.
15433 Peepholes of indirect call and jump via memory: Likewise.
15434 * config/i386/i386.opt: Likewise.
15435 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15436 (GOT_memory_operand): Likewise.
15437 (call_insn_operand): Likewise.
15438 (sibcall_insn_operand): Likewise.
15439 (GOT32_symbol_operand): Likewise.
15441 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15443 PR middle-end/83837
15444 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15445 type rather than type addr's type points to.
15446 (expand_omp_atomic_mutex): Likewise.
15447 (expand_omp_atomic): Likewise.
15449 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15452 * config/i386/i386.c (output_indirect_thunk_function): Use
15453 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15454 for __x86_return_thunk.
15456 2018-01-15 Richard Biener <rguenther@suse.de>
15458 PR middle-end/83850
15459 * expmed.c (extract_bit_field_1): Fix typo.
15461 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15464 * config/arm/iterators.md (VF): New mode iterator.
15465 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15466 Remove integer-related logic from pattern.
15467 (neon_vabd<mode>_3): Likewise.
15469 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15471 PR middle-end/82694
15472 * common.opt (fstrict-overflow): No longer an alias.
15473 (fwrapv-pointer): New option.
15474 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15475 also for pointer types based on flag_wrapv_pointer.
15476 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15477 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15478 opts->x_flag_wrapv got set.
15479 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15480 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15481 POINTER_TYPE_OVERFLOW_UNDEFINED.
15482 * match.pd: Likewise in address comparison pattern.
15483 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15485 2018-01-15 Richard Biener <rguenther@suse.de>
15488 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15489 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15490 Reset type names to their identifier if their TYPE_DECL doesn't
15491 have linkage (and thus is used for ODR and devirt).
15492 (save_debug_info_for_decl): Remove.
15493 (save_debug_info_for_type): Likewise.
15494 (add_tree_to_fld_list): Adjust.
15495 * tree-pretty-print.c (dump_generic_node): Make dumping of
15496 type names more robust.
15498 2018-01-15 Richard Biener <rguenther@suse.de>
15500 * BASE-VER: Bump to 8.0.1.
15502 2018-01-14 Martin Sebor <msebor@redhat.com>
15505 * builtins.c (check_access): Avoid warning when the no-warning bit
15508 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15510 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15511 * ira-color (allocno_hard_regs_compare): Likewise.
15513 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15516 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15517 Use .pushsection/.popsection.
15519 2018-01-14 Martin Sebor <msebor@redhat.com>
15522 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15524 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15526 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15527 entry from extra_headers.
15528 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15529 extra_headers, make the list bitwise identical to the i?86-*-* one.
15531 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15533 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15534 -mcmodel=large with -mindirect-branch=thunk,
15535 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15536 -mfunction-return=thunk-extern.
15537 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15538 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15539 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15541 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15543 * config/i386/i386.c (print_reg): Print the name of the full
15544 integer register without '%'.
15545 (ix86_print_operand): Handle 'V'.
15546 * doc/extend.texi: Document 'V' modifier.
15548 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15550 * config/i386/constraints.md (Bs): Disallow memory operand for
15551 -mindirect-branch-register.
15553 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15554 (GOT_memory_operand): Likewise.
15555 (call_insn_operand): Likewise.
15556 (sibcall_insn_operand): Likewise.
15557 (GOT32_symbol_operand): Likewise.
15558 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15559 for -mindirect-branch-register.
15560 (tablejump): Likewise.
15561 (*sibcall_memory): Likewise.
15562 (*sibcall_value_memory): Likewise.
15563 Disallow peepholes of indirect call and jump via memory for
15564 -mindirect-branch-register.
15565 (*call_pop): Replace m with Bw.
15566 (*call_value_pop): Likewise.
15567 (*sibcall_pop_memory): Replace m with Bs.
15568 * config/i386/i386.opt (mindirect-branch-register): New option.
15569 * doc/invoke.texi: Document -mindirect-branch-register option.
15571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15573 * config/i386/i386-protos.h (ix86_output_function_return): New.
15574 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15575 set function_return_type.
15576 (indirect_thunk_name): Add ret_p to indicate thunk for function
15578 (output_indirect_thunk_function): Pass false to
15579 indirect_thunk_name.
15580 (ix86_output_indirect_branch_via_reg): Likewise.
15581 (ix86_output_indirect_branch_via_push): Likewise.
15582 (output_indirect_thunk_function): Create alias for function
15583 return thunk if regno < 0.
15584 (ix86_output_function_return): New function.
15585 (ix86_handle_fndecl_attribute): Handle function_return.
15586 (ix86_attribute_table): Add function_return.
15587 * config/i386/i386.h (machine_function): Add
15588 function_return_type.
15589 * config/i386/i386.md (simple_return_internal): Use
15590 ix86_output_function_return.
15591 (simple_return_internal_long): Likewise.
15592 * config/i386/i386.opt (mfunction-return=): New option.
15593 (indirect_branch): Mention -mfunction-return=.
15594 * doc/extend.texi: Document function_return function attribute.
15595 * doc/invoke.texi: Document -mfunction-return= option.
15597 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15599 * config/i386/i386-opts.h (indirect_branch): New.
15600 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15601 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15602 with local indirect jump when converting indirect call and jump.
15603 (ix86_set_indirect_branch_type): New.
15604 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15605 (indirectlabelno): New.
15606 (indirect_thunk_needed): Likewise.
15607 (indirect_thunk_bnd_needed): Likewise.
15608 (indirect_thunks_used): Likewise.
15609 (indirect_thunks_bnd_used): Likewise.
15610 (INDIRECT_LABEL): Likewise.
15611 (indirect_thunk_name): Likewise.
15612 (output_indirect_thunk): Likewise.
15613 (output_indirect_thunk_function): Likewise.
15614 (ix86_output_indirect_branch_via_reg): Likewise.
15615 (ix86_output_indirect_branch_via_push): Likewise.
15616 (ix86_output_indirect_branch): Likewise.
15617 (ix86_output_indirect_jmp): Likewise.
15618 (ix86_code_end): Call output_indirect_thunk_function if needed.
15619 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15621 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15622 (ix86_attribute_table): Add indirect_branch.
15623 * config/i386/i386.h (machine_function): Add indirect_branch_type
15624 and has_local_indirect_jump.
15625 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15627 (tablejump): Likewise.
15628 (*indirect_jump): Use ix86_output_indirect_jmp.
15629 (*tablejump_1): Likewise.
15630 (simple_return_indirect_internal): Likewise.
15631 * config/i386/i386.opt (mindirect-branch=): New option.
15632 (indirect_branch): New.
15635 (thunk-inline): Likewise.
15636 (thunk-extern): Likewise.
15637 * doc/extend.texi: Document indirect_branch function attribute.
15638 * doc/invoke.texi: Document -mindirect-branch= option.
15640 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15643 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15645 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15647 * ipa-inline.c (want_inline_small_function_p): Return false if
15648 inlining has already failed with CIF_FINAL_ERROR.
15649 (update_caller_keys): Call want_inline_small_function_p before
15651 (update_callee_keys): Likewise.
15653 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15655 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15657 (rs6000_quadword_masked_address_p): Likewise.
15658 (quad_aligned_load_p): Likewise.
15659 (quad_aligned_store_p): Likewise.
15660 (const_load_sequence_p): Add comment to describe the outer-most loop.
15661 (mimic_memory_attributes_and_flags): New function.
15662 (rs6000_gen_stvx): Likewise.
15663 (replace_swapped_aligned_store): Likewise.
15664 (rs6000_gen_lvx): Likewise.
15665 (replace_swapped_aligned_load): Likewise.
15666 (replace_swapped_load_constant): Capitalize argument name in
15667 comment describing this function.
15668 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15669 and stores that access quad-word aligned addresses and replace
15670 with stvx or lvx instructions when appropriate.
15671 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15672 New function prototype.
15673 (rs6000_quadword_masked_address_p): Likewise.
15674 (rs6000_gen_lvx): Likewise.
15675 (rs6000_gen_stvx): Likewise.
15676 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15677 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15678 when memory address is aligned.
15679 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15680 this split to select lvx instruction when memory address is aligned.
15681 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15682 instruction when memory address is aligned.
15683 (*vsx_le_perm_load_v16qi): Likewise.
15684 (four unnamed splitters): Modify to select the stvx instruction
15685 when memory is aligned.
15687 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15689 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15690 which appears in the queue multiple times.
15692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15693 Alan Hayward <alan.hayward@arm.com>
15694 David Sherwood <david.sherwood@arm.com>
15696 * tree-vectorizer.h (vec_lower_bound): New structure.
15697 (_loop_vec_info): Add check_nonzero and lower_bounds.
15698 (LOOP_VINFO_CHECK_NONZERO): New macro.
15699 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15700 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15701 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15702 fields. Make seg_len the distance travelled, not including the
15704 (dr_direction_indicator): Declare.
15705 (dr_zero_step_indicator): Likewise.
15706 (dr_known_forward_stride_p): Likewise.
15707 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15709 (runtime_alias_check_p): Allow runtime alias checks with
15711 (operator ==): Compare access_size and align.
15712 (prune_runtime_alias_test_list): Rework for new distinction between
15713 the access_size and seg_len.
15714 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15716 (get_segment_min_max): New function.
15717 (create_intersect_range_checks): Use it.
15718 (dr_step_indicator): New function.
15719 (dr_direction_indicator): Likewise.
15720 (dr_zero_step_indicator): Likewise.
15721 (dr_known_forward_stride_p): Likewise.
15722 * tree-loop-distribution.c (data_ref_segment_size): Return
15723 DR_STEP * (niters - 1).
15724 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15726 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15727 (vect_preserves_scalar_order_p): New function, split out from...
15728 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15729 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15730 (vect_vfa_access_size): New function.
15731 (vect_vfa_align): Likewise.
15732 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15733 (dump_lower_bound): New function.
15734 (vect_check_lower_bound): Likewise.
15735 (vect_small_gap_p): Likewise.
15736 (vectorizable_with_step_bound_p): Likewise.
15737 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15738 depencies if the vectorization factor is 1. Convert the checks
15739 for nonzero steps into checks on the bounds of DR_STEP. Try using
15740 a bunds check for variable steps if the minimum required step is
15741 relatively small. Update calls to the dr_with_seg_len
15742 constructor and to vect_compile_time_alias.
15743 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15745 (vect_loop_versioning): Call it.
15746 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15748 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15751 Alan Hayward <alan.hayward@arm.com>
15752 David Sherwood <david.sherwood@arm.com>
15754 * doc/sourcebuild.texi (vect_scatter_store): Document.
15755 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15757 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15759 * genopinit.c (main): Add supports_vec_scatter_store and
15760 supports_vec_scatter_store_cached to target_optabs.
15761 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15762 IFN_MASK_SCATTER_STORE.
15763 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15765 * internal-fn.h (internal_store_fn_p): Declare.
15766 (internal_fn_stored_value_index): Likewise.
15767 * internal-fn.c (scatter_store_direct): New macro.
15768 (expand_scatter_store_optab_fn): New function.
15769 (direct_scatter_store_optab_supported_p): New macro.
15770 (internal_store_fn_p): New function.
15771 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15772 IFN_MASK_SCATTER_STORE.
15773 (internal_fn_mask_index): Likewise.
15774 (internal_fn_stored_value_index): New function.
15775 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15776 for scatter stores.
15777 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15778 * optabs-query.c (supports_vec_scatter_store_p): New function.
15779 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15780 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15781 true for scatter stores.
15782 (vect_gather_scatter_fn_p): Handle scatter stores too.
15783 (vect_check_gather_scatter): Consider using scatter stores if
15784 supports_vec_scatter_store_p.
15785 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15786 scatter stores too.
15787 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15788 internal_fn_stored_value_index.
15789 (check_load_store_masking): Handle scatter stores too.
15790 (vect_get_store_rhs): Make public.
15791 (vectorizable_call): Use internal_store_fn_p.
15792 (vectorizable_store): Handle scatter store internal functions.
15793 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15794 when deciding whether the end of the group has been reached.
15795 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15796 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15797 (mask_scatter_store<mode>): New insns.
15799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15800 Alan Hayward <alan.hayward@arm.com>
15801 David Sherwood <david.sherwood@arm.com>
15803 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15804 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15805 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15807 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15808 Use vect_truncate_gather_scatter_offset if we can't treat the
15809 operation as a normal gather load or scatter store.
15810 (get_group_load_store_type): Take the gather_scatter_info
15811 as argument. Try using a gather load or scatter store for
15812 single-element groups.
15813 (get_load_store_type): Update calls to get_group_load_store_type
15814 and vect_use_strided_gather_scatters_p.
15816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15817 Alan Hayward <alan.hayward@arm.com>
15818 David Sherwood <david.sherwood@arm.com>
15820 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15821 optional tree argument.
15822 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15824 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15825 but continue to use the current value as a fallback.
15826 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15827 to compare the updates.
15828 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15829 (get_load_store_type): Use it when handling a strided access.
15830 (vect_get_strided_load_store_ops): New function.
15831 (vect_get_data_ptr_increment): Likewise.
15832 (vectorizable_load): Handle strided gather loads. Always pass
15833 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15836 Alan Hayward <alan.hayward@arm.com>
15837 David Sherwood <david.sherwood@arm.com>
15839 * doc/md.texi (gather_load@var{m}): Document.
15840 (mask_gather_load@var{m}): Likewise.
15841 * genopinit.c (main): Add supports_vec_gather_load and
15842 supports_vec_gather_load_cached to target_optabs.
15843 * optabs-tree.c (init_tree_optimization_optabs): Use
15844 ggc_cleared_alloc to allocate target_optabs.
15845 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15846 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15848 * internal-fn.h (internal_load_fn_p): Declare.
15849 (internal_gather_scatter_fn_p): Likewise.
15850 (internal_fn_mask_index): Likewise.
15851 (internal_gather_scatter_fn_supported_p): Likewise.
15852 * internal-fn.c (gather_load_direct): New macro.
15853 (expand_gather_load_optab_fn): New function.
15854 (direct_gather_load_optab_supported_p): New macro.
15855 (direct_internal_fn_optab): New function.
15856 (internal_load_fn_p): Likewise.
15857 (internal_gather_scatter_fn_p): Likewise.
15858 (internal_fn_mask_index): Likewise.
15859 (internal_gather_scatter_fn_supported_p): Likewise.
15860 * optabs-query.c (supports_at_least_one_mode_p): New function.
15861 (supports_vec_gather_load_p): Likewise.
15862 * optabs-query.h (supports_vec_gather_load_p): Declare.
15863 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15864 and memory_type field.
15865 (NUM_PATTERNS): Bump to 15.
15866 * tree-vect-data-refs.c: Include internal-fn.h.
15867 (vect_gather_scatter_fn_p): New function.
15868 (vect_describe_gather_scatter_call): Likewise.
15869 (vect_check_gather_scatter): Try using internal functions for
15870 gather loads. Recognize existing calls to a gather load function.
15871 (vect_analyze_data_refs): Consider using gather loads if
15872 supports_vec_gather_load_p.
15873 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15874 (vect_get_gather_scatter_offset_type): Likewise.
15875 (vect_convert_mask_for_vectype): Likewise.
15876 (vect_add_conversion_to_patterm): Likewise.
15877 (vect_try_gather_scatter_pattern): Likewise.
15878 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15879 (vect_vect_recog_func_ptrs): Add it.
15880 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15881 internal_fn_mask_index and internal_gather_scatter_fn_p.
15882 (check_load_store_masking): Take the gather_scatter_info as an
15883 argument and handle gather loads.
15884 (vect_get_gather_scatter_ops): New function.
15885 (vectorizable_call): Check internal_load_fn_p.
15886 (vectorizable_load): Likewise. Handle gather load internal
15888 (vectorizable_store): Update call to check_load_store_masking.
15889 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15890 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15891 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15892 (aarch64_gather_scale_operand_d): New predicates.
15893 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15894 (mask_gather_load<mode>): New insns.
15896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15897 Alan Hayward <alan.hayward@arm.com>
15898 David Sherwood <david.sherwood@arm.com>
15900 * optabs.def (fold_left_plus_optab): New optab.
15901 * doc/md.texi (fold_left_plus_@var{m}): Document.
15902 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15903 * internal-fn.c (fold_left_direct): Define.
15904 (expand_fold_left_optab_fn): Likewise.
15905 (direct_fold_left_optab_supported_p): Likewise.
15906 * fold-const-call.c (fold_const_fold_left): New function.
15907 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15908 * tree-parloops.c (valid_reduction_p): New function.
15909 (gather_scalar_reductions): Use it.
15910 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15911 (vect_finish_replace_stmt): Declare.
15912 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15913 (needs_fold_left_reduction_p): New function, split out from...
15914 (vect_is_simple_reduction): ...here. Accept reductions that
15915 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15916 (vect_force_simple_reduction): Also store the reduction type in
15917 the assignment's STMT_VINFO_REDUC_TYPE.
15918 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15919 (merge_with_identity): New function.
15920 (vect_expand_fold_left): Likewise.
15921 (vectorize_fold_left_reduction): Likewise.
15922 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15923 scalar phi in place for it. Check for target support and reject
15924 cases that would reassociate the operation. Defer the transform
15925 phase to vectorize_fold_left_reduction.
15926 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15927 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15928 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15932 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15933 call to ifc_temp_var.
15935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15936 Alan Hayward <alan.hayward@arm.com>
15937 David Sherwood <david.sherwood@arm.com>
15939 * target.def (legitimize_address_displacement): Take the original
15940 offset as a poly_int.
15941 * targhooks.h (default_legitimize_address_displacement): Update
15943 * targhooks.c (default_legitimize_address_displacement): Likewise.
15944 * doc/tm.texi: Regenerate.
15945 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15946 as an argument, moving assert of ad->disp == ad->disp_term to...
15947 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15948 Try calling targetm.legitimize_address_displacement before expanding
15949 the address rather than afterwards, and adjust for the new interface.
15950 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15951 Match the new hook interface. Handle SVE addresses.
15952 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15953 new hook interface.
15955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15957 * Makefile.in (OBJS): Add early-remat.o.
15958 * target.def (select_early_remat_modes): New hook.
15959 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15960 * doc/tm.texi: Regenerate.
15961 * targhooks.h (default_select_early_remat_modes): Declare.
15962 * targhooks.c (default_select_early_remat_modes): New function.
15963 * timevar.def (TV_EARLY_REMAT): New timevar.
15964 * passes.def (pass_early_remat): New pass.
15965 * tree-pass.h (make_pass_early_remat): Declare.
15966 * early-remat.c: New file.
15967 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15969 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15972 Alan Hayward <alan.hayward@arm.com>
15973 David Sherwood <david.sherwood@arm.com>
15975 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15976 vfm1 with a bound_epilog parameter.
15977 (vect_do_peeling): Update calls accordingly, and move the prologue
15978 call earlier in the function. Treat the base bound_epilog as 0 for
15979 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15980 this base when peeling for gaps.
15981 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15982 with fully-masked loops.
15983 (vect_estimate_min_profitable_iters): Handle the single peeled
15984 iteration in that case.
15986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15987 Alan Hayward <alan.hayward@arm.com>
15988 David Sherwood <david.sherwood@arm.com>
15990 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15991 single-element interleaving even if the size is not a power of 2.
15992 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15993 accesses for single-element interleaving if the group size is
15996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15997 Alan Hayward <alan.hayward@arm.com>
15998 David Sherwood <david.sherwood@arm.com>
16000 * doc/md.texi (fold_extract_last_@var{m}): Document.
16001 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16002 * optabs.def (fold_extract_last_optab): New optab.
16003 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16004 * internal-fn.c (fold_extract_direct): New macro.
16005 (expand_fold_extract_optab_fn): Likewise.
16006 (direct_fold_extract_optab_supported_p): Likewise.
16007 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16008 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16009 EXTRACT_LAST_REDUCTION.
16010 (get_initial_def_for_reduction): Do not create an initial vector
16011 for EXTRACT_LAST_REDUCTION reductions.
16012 (vectorizable_reduction): Leave the scalar phi in place for
16013 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16014 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16015 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16016 transform phase to vectorizable_condition.
16017 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16019 (vect_finish_stmt_generation): ...here.
16020 (vect_finish_replace_stmt): New function.
16021 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16022 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16024 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16027 Alan Hayward <alan.hayward@arm.com>
16028 David Sherwood <david.sherwood@arm.com>
16030 * doc/md.texi (extract_last_@var{m}): Document.
16031 * optabs.def (extract_last_optab): New optab.
16032 * internal-fn.def (EXTRACT_LAST): New internal function.
16033 * internal-fn.c (cond_unary_direct): New macro.
16034 (expand_cond_unary_optab_fn): Likewise.
16035 (direct_cond_unary_optab_supported_p): Likewise.
16036 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16037 loops using EXTRACT_LAST.
16038 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16039 (extract_last_<mode>): ...this optab.
16040 (vec_extract<mode><Vel>): Update accordingly.
16042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16043 Alan Hayward <alan.hayward@arm.com>
16044 David Sherwood <david.sherwood@arm.com>
16046 * target.def (empty_mask_is_expensive): New hook.
16047 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16048 * doc/tm.texi: Regenerate.
16049 * targhooks.h (default_empty_mask_is_expensive): Declare.
16050 * targhooks.c (default_empty_mask_is_expensive): New function.
16051 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16052 if the target says that empty masks are expensive.
16053 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16055 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16058 Alan Hayward <alan.hayward@arm.com>
16059 David Sherwood <david.sherwood@arm.com>
16061 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16062 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16063 (vect_use_loop_mask_for_alignment_p): New function.
16064 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16065 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16066 niters_skip argument. Make sure that the first niters_skip elements
16067 of the first iteration are inactive.
16068 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16069 Update call to vect_set_loop_masks_directly.
16070 (get_misalign_in_elems): New function, split out from...
16071 (vect_gen_prolog_loop_niters): ...here.
16072 (vect_update_init_of_dr): Take a code argument that specifies whether
16073 the adjustment should be added or subtracted.
16074 (vect_update_init_of_drs): Likewise.
16075 (vect_prepare_for_masked_peels): New function.
16076 (vect_do_peeling): Skip prologue peeling if we're using a mask
16077 instead. Update call to vect_update_inits_of_drs.
16078 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16080 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16081 alignment. Do not include the number of peeled iterations in
16082 the minimum threshold in that case.
16083 (vectorizable_induction): Adjust the start value down by
16084 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16085 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16086 Take the number of skipped iterations into account when calculating
16088 * tree-vect-stmts.c (vect_gen_while_not): New function.
16090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16091 Alan Hayward <alan.hayward@arm.com>
16092 David Sherwood <david.sherwood@arm.com>
16094 * doc/sourcebuild.texi (vect_fully_masked): Document.
16095 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16096 default value to 0.
16097 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16099 (vect_analyze_loop_2): ...here. Don't check the vectorization
16100 factor against the number of loop iterations if the loop is
16103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16104 Alan Hayward <alan.hayward@arm.com>
16105 David Sherwood <david.sherwood@arm.com>
16107 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16108 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16109 (dump_groups): Update accordingly.
16110 (iv_use::mem_type): New member variable.
16111 (address_p): New function.
16112 (record_use): Add a mem_type argument and initialize the new
16114 (record_group_use): Add a mem_type argument. Use address_p.
16115 Remove obsolete null checks of base_object. Update call to record_use.
16116 (find_interesting_uses_op): Update call to record_group_use.
16117 (find_interesting_uses_cond): Likewise.
16118 (find_interesting_uses_address): Likewise.
16119 (get_mem_type_for_internal_fn): New function.
16120 (find_address_like_use): Likewise.
16121 (find_interesting_uses_stmt): Try find_address_like_use before
16122 calling find_interesting_uses_op.
16123 (addr_offset_valid_p): Use the iv mem_type field as the type
16124 of the addressed memory.
16125 (add_autoinc_candidates): Likewise.
16126 (get_address_cost): Likewise.
16127 (split_small_address_groups_p): Use address_p.
16128 (split_address_groups): Likewise.
16129 (add_iv_candidate_for_use): Likewise.
16130 (autoinc_possible_for_pair): Likewise.
16131 (rewrite_groups): Likewise.
16132 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16133 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16134 (get_alias_ptr_type_for_ptr_address): New function.
16135 (rewrite_use_address): Rewrite address uses in calls that were
16136 identified by find_address_like_use.
16138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16139 Alan Hayward <alan.hayward@arm.com>
16140 David Sherwood <david.sherwood@arm.com>
16142 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16144 * gimple-expr.h (is_gimple_addressable: Likewise.
16145 * gimple-expr.c (is_gimple_address): Likewise.
16146 * internal-fn.c (expand_call_mem_ref): New function.
16147 (expand_mask_load_optab_fn): Use it.
16148 (expand_mask_store_optab_fn): Likewise.
16150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16151 Alan Hayward <alan.hayward@arm.com>
16152 David Sherwood <david.sherwood@arm.com>
16154 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16155 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16156 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16157 (cond_umax@var{mode}): Document.
16158 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16159 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16160 (cond_umin_optab, cond_umax_optab): New optabs.
16161 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16162 (COND_IOR, COND_XOR): New internal functions.
16163 * internal-fn.h (get_conditional_internal_fn): Declare.
16164 * internal-fn.c (cond_binary_direct): New macro.
16165 (expand_cond_binary_optab_fn): Likewise.
16166 (direct_cond_binary_optab_supported_p): Likewise.
16167 (get_conditional_internal_fn): New function.
16168 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16169 Cope with reduction statements that are vectorized as calls rather
16171 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16172 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16173 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16174 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16175 (UNSPEC_COND_EOR): New unspecs.
16176 (optab): Add mappings for them.
16177 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16178 (sve_int_op, sve_fp_op): New int attributes.
16180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16181 Alan Hayward <alan.hayward@arm.com>
16182 David Sherwood <david.sherwood@arm.com>
16184 * optabs.def (while_ult_optab): New optab.
16185 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16186 * internal-fn.def (WHILE_ULT): New internal function.
16187 * internal-fn.h (direct_internal_fn_supported_p): New override
16188 that takes two types as argument.
16189 * internal-fn.c (while_direct): New macro.
16190 (expand_while_optab_fn): New function.
16191 (convert_optab_supported_p): Likewise.
16192 (direct_while_optab_supported_p): New macro.
16193 * wide-int.h (wi::udiv_ceil): New function.
16194 * tree-vectorizer.h (rgroup_masks): New structure.
16195 (vec_loop_masks): New typedef.
16196 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16197 and fully_masked_p.
16198 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16199 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16200 (vect_max_vf): New function.
16201 (slpeel_make_loop_iterate_ntimes): Delete.
16202 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16203 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16204 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16205 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16206 internal-fn.h, stor-layout.h and optabs-query.h.
16207 (vect_set_loop_mask): New function.
16208 (add_preheader_seq): Likewise.
16209 (add_header_seq): Likewise.
16210 (interleave_supported_p): Likewise.
16211 (vect_maybe_permute_loop_masks): Likewise.
16212 (vect_set_loop_masks_directly): Likewise.
16213 (vect_set_loop_condition_masked): Likewise.
16214 (vect_set_loop_condition_unmasked): New function, split out from
16215 slpeel_make_loop_iterate_ntimes.
16216 (slpeel_make_loop_iterate_ntimes): Rename to..
16217 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16218 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16219 (vect_do_peeling): Update call accordingly.
16220 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16222 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16223 mask_compare_type, can_fully_mask_p and fully_masked_p.
16224 (release_vec_loop_masks): New function.
16225 (_loop_vec_info): Use it to free the loop masks.
16226 (can_produce_all_loop_masks_p): New function.
16227 (vect_get_max_nscalars_per_iter): Likewise.
16228 (vect_verify_full_masking): Likewise.
16229 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16230 retries, and free the mask rgroups before retrying. Check loop-wide
16231 reasons for disallowing fully-masked loops. Make the final decision
16232 about whether use a fully-masked loop or not.
16233 (vect_estimate_min_profitable_iters): Do not assume that peeling
16234 for the number of iterations will be needed for fully-masked loops.
16235 (vectorizable_reduction): Disable fully-masked loops.
16236 (vectorizable_live_operation): Likewise.
16237 (vect_halve_mask_nunits): New function.
16238 (vect_double_mask_nunits): Likewise.
16239 (vect_record_loop_mask): Likewise.
16240 (vect_get_loop_mask): Likewise.
16241 (vect_transform_loop): Handle the case in which the final loop
16242 iteration might handle a partial vector. Call vect_set_loop_condition
16243 instead of slpeel_make_loop_iterate_ntimes.
16244 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16245 (check_load_store_masking): New function.
16246 (prepare_load_store_mask): Likewise.
16247 (vectorizable_store): Handle fully-masked loops.
16248 (vectorizable_load): Likewise.
16249 (supportable_widening_operation): Use vect_halve_mask_nunits for
16251 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16252 (vect_gen_while): New function.
16253 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16254 (aarch64_uqdec<mode>): New insn.
16256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16257 Alan Hayward <alan.hayward@arm.com>
16258 David Sherwood <david.sherwood@arm.com>
16260 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16261 (reduc_xor_scal_optab): New optabs.
16262 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16263 (reduc_xor_scal_@var{m}): Document.
16264 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16265 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16266 internal functions.
16267 * fold-const-call.c (fold_const_call): Handle them.
16268 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16269 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16270 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16271 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16272 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16273 (UNSPEC_XORV): New unspecs.
16274 (optab): Add entries for them.
16275 (BITWISEV): New int iterator.
16276 (bit_reduc_op): New int attributes.
16278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16279 Alan Hayward <alan.hayward@arm.com>
16280 David Sherwood <david.sherwood@arm.com>
16282 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16283 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16284 * optabs.def (vec_shl_insert_optab): New optab.
16285 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16286 (duplicate_and_interleave): Likewise.
16287 * tree-vect-loop.c: Include internal-fn.h.
16288 (neutral_op_for_slp_reduction): New function, split out from
16289 get_initial_defs_for_reduction.
16290 (get_initial_def_for_reduction): Handle option 2 for variable-length
16291 vectors by loading the neutral value into a vector and then shifting
16292 the initial value into element 0.
16293 (get_initial_defs_for_reduction): Replace the code argument with
16294 the neutral value calculated by neutral_op_for_slp_reduction.
16295 Use gimple_build_vector for constant-length vectors.
16296 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16297 but the first group_size elements have a neutral value.
16298 Use duplicate_and_interleave otherwise.
16299 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16300 Update call to get_initial_defs_for_reduction. Handle SLP
16301 reductions for variable-length vectors by creating one vector
16302 result for each scalar result, with the elements associated
16303 with other scalar results stubbed out with the neutral value.
16304 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16305 Require IFN_VEC_SHL_INSERT for double reductions on
16306 variable-length vectors, or SLP reductions that have
16307 a neutral value. Require can_duplicate_and_interleave_p
16308 support for variable-length unchained SLP reductions if there
16309 is no neutral value, such as for MIN/MAX reductions. Also require
16310 the number of vector elements to be a multiple of the number of
16311 SLP statements when doing variable-length unchained SLP reductions.
16312 Update call to vect_create_epilog_for_reduction.
16313 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16314 and remove initial values.
16315 (duplicate_and_interleave): Make public.
16316 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16317 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16320 Alan Hayward <alan.hayward@arm.com>
16321 David Sherwood <david.sherwood@arm.com>
16323 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16324 (can_duplicate_and_interleave_p): New function.
16325 (vect_get_and_check_slp_defs): Take the vector of statements
16326 rather than just the current one. Remove excess parentheses.
16327 Restriction rejectinon of vect_constant_def and vect_external_def
16328 for variable-length vectors to boolean types, or types for which
16329 can_duplicate_and_interleave_p is false.
16330 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16331 (duplicate_and_interleave): New function.
16332 (vect_get_constant_vectors): Use gimple_build_vector for
16333 constant-length vectors and suitable variable-length constant
16334 vectors. Use duplicate_and_interleave for other variable-length
16335 vectors. Don't defer the update when inserting new statements.
16337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16338 Alan Hayward <alan.hayward@arm.com>
16339 David Sherwood <david.sherwood@arm.com>
16341 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16342 min_profitable_iters doesn't go negative.
16344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16345 Alan Hayward <alan.hayward@arm.com>
16346 David Sherwood <david.sherwood@arm.com>
16348 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16349 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16350 * optabs.def (vec_mask_load_lanes_optab): New optab.
16351 (vec_mask_store_lanes_optab): Likewise.
16352 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16353 (MASK_STORE_LANES): Likewise.
16354 * internal-fn.c (mask_load_lanes_direct): New macro.
16355 (mask_store_lanes_direct): Likewise.
16356 (expand_mask_load_optab_fn): Handle masked operations.
16357 (expand_mask_load_lanes_optab_fn): New macro.
16358 (expand_mask_store_optab_fn): Handle masked operations.
16359 (expand_mask_store_lanes_optab_fn): New macro.
16360 (direct_mask_load_lanes_optab_supported_p): Likewise.
16361 (direct_mask_store_lanes_optab_supported_p): Likewise.
16362 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16364 (vect_load_lanes_supported): Likewise.
16365 * tree-vect-data-refs.c (strip_conversion): New function.
16366 (can_group_stmts_p): Likewise.
16367 (vect_analyze_data_ref_accesses): Use it instead of checking
16368 for a pair of assignments.
16369 (vect_store_lanes_supported): Take a masked_p parameter.
16370 (vect_load_lanes_supported): Likewise.
16371 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16372 vect_store_lanes_supported and vect_load_lanes_supported.
16373 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16374 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16375 parameter. Don't allow gaps for masked accesses.
16376 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16377 and vect_load_lanes_supported.
16378 (get_load_store_type): Take a masked_p parameter and update
16379 call to get_group_load_store_type.
16380 (vectorizable_store): Update call to get_load_store_type.
16381 Handle IFN_MASK_STORE_LANES.
16382 (vectorizable_load): Update call to get_load_store_type.
16383 Handle IFN_MASK_LOAD_LANES.
16385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16386 Alan Hayward <alan.hayward@arm.com>
16387 David Sherwood <david.sherwood@arm.com>
16389 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16391 * config/aarch64/aarch64-protos.h
16392 (aarch64_sve_struct_memory_operand_p): Declare.
16393 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16394 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16395 (VPRED, vpred): Handle SVE structure modes.
16396 * config/aarch64/constraints.md (Utx): New constraint.
16397 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16398 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16399 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16400 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16401 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16402 structure modes. Split into pieces after RA.
16403 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16404 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16406 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16407 SVE structure modes.
16408 (aarch64_classify_address): Likewise.
16409 (sizetochar): Move earlier in file.
16410 (aarch64_print_operand): Handle SVE register lists.
16411 (aarch64_array_mode): New function.
16412 (aarch64_sve_struct_memory_operand_p): Likewise.
16413 (TARGET_ARRAY_MODE): Redefine.
16415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16416 Alan Hayward <alan.hayward@arm.com>
16417 David Sherwood <david.sherwood@arm.com>
16419 * target.def (array_mode): New target hook.
16420 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16421 * doc/tm.texi: Regenerate.
16422 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16423 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16424 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16425 targetm.array_mode.
16426 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16430 Alan Hayward <alan.hayward@arm.com>
16431 David Sherwood <david.sherwood@arm.com>
16433 * fold-const.c (fold_binary_loc): Check the argument types
16434 rather than the result type when testing for a vector operation.
16436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16438 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16439 * doc/tm.texi: Regenerate.
16441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16442 Alan Hayward <alan.hayward@arm.com>
16443 David Sherwood <david.sherwood@arm.com>
16445 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16446 (sve): Document new AArch64 extension.
16447 * doc/md.texi (w): Extend the description of the AArch64
16448 constraint to include SVE vectors.
16449 (Upl, Upa): Document new AArch64 predicate constraints.
16450 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16452 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16453 (msve-vector-bits=): New option.
16454 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16455 SVE when these are disabled.
16456 (sve): New extension.
16457 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16458 modes. Adjust their number of units based on aarch64_sve_vg.
16459 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16460 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16461 aarch64_addr_query_type.
16462 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16463 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16464 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16465 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16466 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16467 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16468 (aarch64_simd_imm_zero_p): Delete.
16469 (aarch64_check_zero_based_sve_index_immediate): Declare.
16470 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16471 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16472 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16473 (aarch64_sve_float_mul_immediate_p): Likewise.
16474 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16475 rather than an rtx.
16476 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16477 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16478 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16479 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16480 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16481 (aarch64_regmode_natural_size): Likewise.
16482 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16483 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16485 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16486 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16487 for VG and the SVE predicate registers.
16488 (V_ALIASES): Add a "z"-prefixed alias.
16489 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16490 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16491 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16492 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16493 (REG_CLASS_NAMES): Add entries for them.
16494 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16495 and the predicate registers.
16496 (aarch64_sve_vg): Declare.
16497 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16498 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16499 (REGMODE_NATURAL_SIZE): Define.
16500 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16502 * config/aarch64/aarch64.c: Include cfgrtl.h.
16503 (simd_immediate_info): Add a constructor for series vectors,
16504 and an associated step field.
16505 (aarch64_sve_vg): New variable.
16506 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16507 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16508 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16509 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16510 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16511 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16512 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16513 (aarch64_get_mask_mode): New functions.
16514 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16515 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16516 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16517 predicate modes and predicate registers. Explicitly restrict
16518 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16519 to store a vector mode if it is recognized by
16520 aarch64_classify_vector_mode.
16521 (aarch64_regmode_natural_size): New function.
16522 (aarch64_hard_regno_caller_save_mode): Return the original mode
16524 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16525 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16526 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16527 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16529 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16530 does not overlap dest if the function is frame-related. Handle
16532 (aarch64_split_add_offset): New function.
16533 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16534 them aarch64_add_offset.
16535 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16536 and update call to aarch64_sub_sp.
16537 (aarch64_add_cfa_expression): New function.
16538 (aarch64_expand_prologue): Pass extra temporary registers to the
16539 functions above. Handle the case in which we need to emit new
16540 DW_CFA_expressions for registers that were originally saved
16541 relative to the stack pointer, but now have to be expressed
16542 relative to the frame pointer.
16543 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16545 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16546 IP0 and IP1 values for SVE frames.
16547 (aarch64_expand_vec_series): New function.
16548 (aarch64_expand_sve_widened_duplicate): Likewise.
16549 (aarch64_expand_sve_const_vector): Likewise.
16550 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16551 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16552 into the register, rather than emitting a SET directly.
16553 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16554 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16555 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16556 (offset_9bit_signed_scaled_p): New functions.
16557 (aarch64_replicate_bitmask_imm): New function.
16558 (aarch64_bitmask_imm): Use it.
16559 (aarch64_cannot_force_const_mem): Reject expressions involving
16560 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16561 (aarch64_classify_index): Handle SVE indices, by requiring
16562 a plain register index with a scale that matches the element size.
16563 (aarch64_classify_address): Handle SVE addresses. Assert that
16564 the mode of the address is VOIDmode or an integer mode.
16565 Update call to aarch64_classify_symbol.
16566 (aarch64_classify_symbolic_expression): Update call to
16567 aarch64_classify_symbol.
16568 (aarch64_const_vec_all_in_range_p): New function.
16569 (aarch64_print_vector_float_operand): Likewise.
16570 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16571 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16572 and the FP immediates 1.0 and 0.5.
16573 (aarch64_print_address_internal): Handle SVE addresses.
16574 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16575 (aarch64_regno_regclass): Handle predicate registers.
16576 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16578 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16579 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16580 (aarch64_convert_sve_vector_bits): New function.
16581 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16582 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16583 rather than an rtx.
16584 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16585 Handle SVE vector and predicate modes. Accept VL-based constants
16586 that need only one temporary register, and VL offsets that require
16587 no temporary registers.
16588 (aarch64_conditional_register_usage): Mark the predicate registers
16589 as fixed if SVE isn't available.
16590 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16591 Return true for SVE vector and predicate modes.
16592 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16593 rather than an unsigned int. Handle SVE modes.
16594 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16596 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16598 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16599 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16600 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16601 (aarch64_sve_float_mul_immediate_p): New functions.
16602 (aarch64_sve_valid_immediate): New function.
16603 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16604 Explicitly reject structure modes. Check for INDEX constants.
16605 Handle PTRUE and PFALSE constants.
16606 (aarch64_check_zero_based_sve_index_immediate): New function.
16607 (aarch64_simd_imm_zero_p): Delete.
16608 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16609 vector modes. Accept constants in the range of CNT[BHWD].
16610 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16611 ask for an Advanced SIMD mode.
16612 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16613 (aarch64_simd_vector_alignment): Handle SVE predicates.
16614 (aarch64_vectorize_preferred_vector_alignment): New function.
16615 (aarch64_simd_vector_alignment_reachable): Use it instead of
16617 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16618 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16620 (MAX_VECT_LEN): Delete.
16621 (expand_vec_perm_d): Add a vec_flags field.
16622 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16623 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16624 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16626 (aarch64_evpc_rev): Rename to...
16627 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16628 (aarch64_evpc_rev_global): New function.
16629 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16630 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16632 (aarch64_evpc_sve_tbl): New function.
16633 (aarch64_expand_vec_perm_const_1): Update after rename of
16634 aarch64_evpc_rev. Handle SVE permutes too, trying
16635 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16636 than aarch64_evpc_tbl.
16637 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16638 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16639 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16640 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16641 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16642 (aarch64_expand_sve_vcond): New functions.
16643 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16644 of aarch64_vector_mode_p.
16645 (aarch64_dwarf_poly_indeterminate_value): New function.
16646 (aarch64_compute_pressure_classes): Likewise.
16647 (aarch64_can_change_mode_class): Likewise.
16648 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16649 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16650 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16651 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16652 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16653 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16654 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16655 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16657 (Dn, Dl, Dr): Accept const as well as const_vector.
16658 (Dz): Likewise. Compare against CONST0_RTX.
16659 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16660 of "vector" where appropriate.
16661 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16662 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16663 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16664 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16665 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16666 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16667 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16668 (v_int_equiv): Extend to SVE modes.
16669 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16671 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16672 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16673 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16674 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16675 (SVE_COND_FP_CMP): New int iterators.
16676 (perm_hilo): Handle the new unpack unspecs.
16677 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16679 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16680 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16681 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16682 (aarch64_equality_operator, aarch64_constant_vector_operand)
16683 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16684 (aarch64_sve_nonimmediate_operand): Likewise.
16685 (aarch64_sve_general_operand): Likewise.
16686 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16687 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16688 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16689 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16690 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16691 (aarch64_sve_float_arith_immediate): Likewise.
16692 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16693 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16694 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16695 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16696 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16697 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16698 (aarch64_sve_float_arith_operand): Likewise.
16699 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16700 (aarch64_sve_float_mul_operand): Likewise.
16701 (aarch64_sve_vec_perm_operand): Likewise.
16702 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16703 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16704 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16705 as well as const_vector.
16706 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16707 in file. Use CONST0_RTX and CONSTM1_RTX.
16708 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16709 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16710 Use aarch64_simd_imm_zero.
16711 * config/aarch64/aarch64-sve.md: New file.
16712 * config/aarch64/aarch64.md: Include it.
16713 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16714 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16715 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16716 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16717 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16718 (sve): New attribute.
16719 (enabled): Disable instructions with the sve attribute unless
16721 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16722 aarch64_expand_mov_immediate.
16723 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16724 CNT[BHSD] immediates.
16725 (movti): Split CONST_POLY_INT moves into two halves.
16726 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16727 Split additions that need a temporary here if the destination
16728 is the stack pointer.
16729 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16730 (*add<mode>3_poly_1): New instruction.
16731 (set_clobber_cc): New expander.
16733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16735 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16736 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16737 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16738 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16739 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16740 Change innermode from fixed_mode_size to machine_mode.
16741 (simplify_subreg): Update call accordingly. Handle a constant-sized
16742 subreg of a variable-length CONST_VECTOR.
16744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16745 Alan Hayward <alan.hayward@arm.com>
16746 David Sherwood <david.sherwood@arm.com>
16748 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16749 (add_offset_to_base): New function, split out from...
16750 (create_mem_ref): ...here. When handling a scale other than 1,
16751 check first whether the address is valid without the offset.
16752 Add it into the base if so, leaving the index and scale as-is.
16754 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16757 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16758 fold_for_warn before checking if arg2 is INTEGER_CST.
16760 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16762 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16763 (store_multiple_operation): Delete.
16764 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16765 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16766 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16767 guarded by TARGET_STRING.
16768 (rs6000_output_load_multiple): Delete.
16769 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16770 OPTION_MASK_STRING / TARGET_STRING handling.
16771 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16772 (const rs6000_opt_masks) <"string">: Change mask to 0.
16773 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16774 (MASK_STRING): Delete.
16775 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16777 (load_multiple): Delete.
16784 (store_multiple): Delete.
16791 (movmemsi_8reg): Delete.
16792 (corresponding unnamed define_insn): Delete.
16793 (movmemsi_6reg): Delete.
16794 (corresponding unnamed define_insn): Delete.
16795 (movmemsi_4reg): Delete.
16796 (corresponding unnamed define_insn): Delete.
16797 (movmemsi_2reg): Delete.
16798 (corresponding unnamed define_insn): Delete.
16799 (movmemsi_1reg): Delete.
16800 (corresponding unnamed define_insn): Delete.
16801 * config/rs6000/rs6000.opt (mno-string): New.
16802 (mstring): Replace by deprecation warning stub.
16803 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16805 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16807 * regrename.c (regrename_do_replace): If replacing the same
16808 reg multiple times, try to reuse last created gen_raw_REG.
16811 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16812 main to workaround a bug in GDB.
16814 2018-01-12 Tom de Vries <tom@codesourcery.com>
16817 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16819 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16821 PR rtl-optimization/80481
16822 * ira-color.c (get_cap_member): New function.
16823 (allocnos_conflict_by_live_ranges_p): Use it.
16824 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16825 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16827 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16830 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16831 (*saddl_se_1): Ditto.
16832 (*ssubsi_1): Ditto.
16833 (*ssubl_se_1): Ditto.
16835 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16837 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16838 rather than wi::to_widest for DR_INITs.
16839 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16840 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16841 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16843 (vect_analyze_group_access_1): Note that here.
16845 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16847 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16848 polynomial type sizes.
16850 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16852 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16853 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16854 (gimple_add_tmp_var): Likewise.
16856 2018-01-12 Martin Liska <mliska@suse.cz>
16858 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16859 (gimple_alloc_sizes): Likewise.
16860 (dump_gimple_statistics): Use PRIu64 in printf format.
16861 * gimple.h: Change uint64_t to int.
16863 2018-01-12 Martin Liska <mliska@suse.cz>
16865 * tree-core.h: Use uint64_t instead of int.
16866 * tree.c (tree_node_counts): Likewise.
16867 (tree_node_sizes): Likewise.
16868 (dump_tree_statistics): Use PRIu64 in printf format.
16870 2018-01-12 Martin Liska <mliska@suse.cz>
16872 * Makefile.in: As qsort_chk is implemented in vec.c, add
16873 vec.o to linkage of gencfn-macros.
16874 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16875 passing the info to record_node_allocation_statistics.
16876 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16878 * ggc-common.c (struct ggc_usage): Add operator== and use
16879 it in operator< and compare function.
16880 * mem-stats.h (struct mem_usage): Likewise.
16881 * vec.c (struct vec_usage): Remove operator< and compare
16882 function. Can be simply inherited.
16884 2018-01-12 Martin Jambor <mjambor@suse.cz>
16887 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16888 * tree-ssa-math-opts.c: Include domwalk.h.
16889 (convert_mult_to_fma_1): New function.
16890 (fma_transformation_info): New type.
16891 (fma_deferring_state): Likewise.
16892 (cancel_fma_deferring): New function.
16893 (result_of_phi): Likewise.
16894 (last_fma_candidate_feeds_initial_phi): Likewise.
16895 (convert_mult_to_fma): Added deferring logic, split actual
16896 transformation to convert_mult_to_fma_1.
16897 (math_opts_dom_walker): New type.
16898 (math_opts_dom_walker::after_dom_children): New method, body moved
16899 here from pass_optimize_widening_mul::execute, added deferring logic
16901 (pass_optimize_widening_mul::execute): Moved most of code to
16902 math_opts_dom_walker::after_dom_children.
16903 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16904 * config/i386/i386.c (ix86_option_override_internal): Added
16905 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16907 2018-01-12 Richard Biener <rguenther@suse.de>
16910 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16911 inline instance vars.
16913 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16916 * config/rx/rx.c (rx_is_restricted_memory_address):
16917 Handle SUBREG case.
16919 2018-01-12 Richard Biener <rguenther@suse.de>
16921 PR tree-optimization/80846
16922 * target.def (split_reduction): New target hook.
16923 * targhooks.c (default_split_reduction): New function.
16924 * targhooks.h (default_split_reduction): Declare.
16925 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16926 target requests first reduce vectors by combining low and high
16928 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16929 (get_vectype_for_scalar_type_and_size): Export.
16930 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16931 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16932 * doc/tm.texi: Regenerate.
16933 * config/i386/i386.c (ix86_split_reduction): Implement
16934 TARGET_VECTORIZE_SPLIT_REDUCTION.
16936 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16939 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16940 in PIC mode except for TARGET_VXWORKS_RTP.
16941 * config/sparc/sparc.c: Include cfgrtl.h.
16942 (TARGET_INIT_PIC_REG): Define.
16943 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16944 (sparc_pic_register_p): New predicate.
16945 (sparc_legitimate_address_p): Use it.
16946 (sparc_legitimize_pic_address): Likewise.
16947 (sparc_delegitimize_address): Likewise.
16948 (sparc_mode_dependent_address_p): Likewise.
16949 (gen_load_pcrel_sym): Remove 4th parameter.
16950 (load_got_register): Adjust call to above. Remove obsolete stuff.
16951 (sparc_expand_prologue): Do not call load_got_register here.
16952 (sparc_flat_expand_prologue): Likewise.
16953 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16954 (sparc_use_pseudo_pic_reg): New function.
16955 (sparc_init_pic_reg): Likewise.
16956 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16957 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16959 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16961 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16962 Add item for branch_cost.
16964 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16966 PR rtl-optimization/83565
16967 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16968 not extend the result to a larger mode for rotate operations.
16969 (num_sign_bit_copies1): Likewise.
16971 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16974 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16976 Use values-Xc.o for -pedantic.
16977 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16979 2018-01-12 Martin Liska <mliska@suse.cz>
16982 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16984 (possible_polymorphic_call_targets): Use it.
16985 (ipa_devirt): Likewise.
16987 2018-01-12 Martin Liska <mliska@suse.cz>
16989 * profile-count.h (enum profile_quality): Use 0 as invalid
16990 enum value of profile_quality.
16992 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16994 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16995 -mext-string options.
16997 2018-01-12 Richard Biener <rguenther@suse.de>
16999 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17000 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17001 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17003 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17005 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17007 * configure.ac (--with-long-double-format): Add support for the
17008 configuration option to change the default long double format on
17010 * config.gcc (powerpc*-linux*-*): Likewise.
17011 * configure: Regenerate.
17012 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17013 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17014 used without modification.
17016 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17018 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17019 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17020 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17021 MISC_BUILTIN_SPEC_BARRIER.
17022 (rs6000_init_builtins): Likewise.
17023 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17025 (speculation_barrier): New define_insn.
17026 * doc/extend.texi: Document __builtin_speculation_barrier.
17028 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17031 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17032 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17033 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17035 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17036 integral modes instead of "ss" and "sd".
17037 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17038 vectors with 32-bit and 64-bit elements.
17039 (vecdupssescalarmodesuffix): New mode attribute.
17040 (vec_dup<mode>): Use it.
17042 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17045 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17046 frame if argument is passed on stack.
17048 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17051 * ree.c (combine_reaching_defs): Optimize also
17052 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17053 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17055 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17057 PR middle-end/83189
17058 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17060 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17062 PR middle-end/83718
17063 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17064 after they are computed.
17066 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17068 PR tree-optimization/83695
17069 * gimple-loop-linterchange.cc
17070 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17071 reset cached scev information after interchange.
17072 (pass_linterchange::execute): Remove call to scev_reset_htab.
17074 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17076 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17077 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17078 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17079 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17080 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17081 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17082 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17083 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17084 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17085 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17086 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17087 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17088 (V_lane_reg): Likewise.
17089 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17091 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17092 (vfmal_lane_low<mode>_intrinsic,
17093 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17094 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17095 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17096 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17097 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17098 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17100 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17102 * config/arm/arm-cpus.in (fp16fml): New feature.
17103 (ALL_SIMD): Add fp16fml.
17104 (armv8.2-a): Add fp16fml as an option.
17105 (armv8.3-a): Likewise.
17106 (armv8.4-a): Add fp16fml as part of fp16.
17107 * config/arm/arm.h (TARGET_FP16FML): Define.
17108 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17110 * config/arm/arm-modes.def (V2HF): Define.
17111 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17112 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17113 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17114 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17115 vfmsl_low, vfmsl_high): New set of builtins.
17116 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17117 (vfml_op): New code attribute.
17118 (VFMLHALVES): New int iterator.
17119 (VFML, VFMLSEL): New mode attributes.
17120 (V_reg): Define mapping for V2HF.
17121 (V_hi, V_lo): New mode attributes.
17122 (VF_constraint): Likewise.
17123 (vfml_half, vfml_half_selector): New int attributes.
17124 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17126 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17127 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17129 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17130 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17131 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17132 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17134 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17135 Document new effective target and option set.
17137 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17139 * config/arm/arm-cpus.in (armv8_4): New feature.
17140 (ARMv8_4a): New fgroup.
17141 (armv8.4-a): New arch.
17142 * config/arm/arm-tables.opt: Regenerate.
17143 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17144 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17145 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17146 Add matching rules for -march=armv8.4-a and extensions.
17147 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17149 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17152 * config/rx/rx.md (BW): New mode attribute.
17153 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17155 2018-01-11 Richard Biener <rguenther@suse.de>
17157 PR tree-optimization/83435
17158 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17159 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17160 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17162 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17163 Alan Hayward <alan.hayward@arm.com>
17164 David Sherwood <david.sherwood@arm.com>
17166 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17168 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17169 (aarch64_print_address_internal): Use it to check for a zero offset.
17171 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17172 Alan Hayward <alan.hayward@arm.com>
17173 David Sherwood <david.sherwood@arm.com>
17175 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17176 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17177 Return a poly_int64 rather than a HOST_WIDE_INT.
17178 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17179 rather than a HOST_WIDE_INT.
17180 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17181 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17182 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17183 final_offset from HOST_WIDE_INT to poly_int64.
17184 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17185 to_constant when getting the number of units in an Advanced SIMD
17187 (aarch64_builtin_vectorized_function): Check for a constant number
17189 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17191 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17192 attribute instead of GET_MODE_NUNITS.
17193 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17194 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17195 GET_MODE_SIZE for fixed-size registers.
17196 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17197 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17198 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17199 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17200 (aarch64_print_operand, aarch64_print_address_internal)
17201 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17202 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17203 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17204 Handle polynomial GET_MODE_SIZE.
17205 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17206 wider than SImode without modification.
17207 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17208 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17209 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17210 passing and returning SVE modes.
17211 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17212 rather than GEN_INT.
17213 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17214 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17215 (aarch64_allocate_and_probe_stack_space): Likewise.
17216 (aarch64_layout_frame): Cope with polynomial offsets.
17217 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17218 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17219 polynomial offsets.
17220 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17221 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17222 poly_int64 rather than a HOST_WIDE_INT.
17223 (aarch64_get_separate_components, aarch64_process_components)
17224 (aarch64_expand_prologue, aarch64_expand_epilogue)
17225 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17226 (aarch64_anchor_offset): New function, split out from...
17227 (aarch64_legitimize_address): ...here.
17228 (aarch64_builtin_vectorization_cost): Handle polynomial
17229 TYPE_VECTOR_SUBPARTS.
17230 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17232 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17233 number of elements from the PARALLEL rather than the mode.
17234 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17235 rather than GET_MODE_BITSIZE.
17236 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17237 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17238 (aarch64_expand_vec_perm_const_1): Handle polynomial
17239 d->perm.length () and d->perm elements.
17240 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17241 Apply to_constant to d->perm elements.
17242 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17243 polynomial CONST_VECTOR_NUNITS.
17244 (aarch64_move_pointer): Take amount as a poly_int64 rather
17246 (aarch64_progress_pointer): Avoid temporary variable.
17247 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17248 the mode attribute instead of GET_MODE.
17250 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17251 Alan Hayward <alan.hayward@arm.com>
17252 David Sherwood <david.sherwood@arm.com>
17254 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17255 x exists before using it.
17256 (aarch64_add_constant_internal): Rename to...
17257 (aarch64_add_offset_1): ...this. Replace regnum with separate
17258 src and dest rtxes. Handle the case in which they're different,
17259 including when the offset is zero. Replace scratchreg with an rtx.
17260 Use 2 additions if there is no spare register into which we can
17261 move a 16-bit constant.
17262 (aarch64_add_constant): Delete.
17263 (aarch64_add_offset): Replace reg with separate src and dest
17264 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17265 Use aarch64_add_offset_1.
17266 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17267 an rtx rather than an int. Take the delta as a poly_int64
17268 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17269 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17270 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17271 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17272 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17273 and aarch64_add_sp.
17274 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17275 aarch64_add_constant.
17277 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17279 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17280 Use scalar_float_mode.
17282 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17284 * config/aarch64/aarch64-simd.md
17285 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17286 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17287 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17288 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17289 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17290 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17291 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17292 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17293 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17294 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17296 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17299 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17300 targ_options->x_arm_arch_string is non NULL.
17302 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17304 * config/aarch64/aarch64.h
17305 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17307 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17310 * expmed.c (emit_store_flag_force): Swap if const op0
17311 and change VOIDmode to mode of op0.
17313 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17315 PR rtl-optimization/83761
17316 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17317 than bytes to mode_for_size.
17319 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17321 PR middle-end/83189
17322 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17323 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17326 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17328 PR middle-end/83575
17329 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17330 when in layout mode.
17331 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17332 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17335 2018-01-10 Michael Collison <michael.collison@arm.com>
17337 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17338 * config/aarch64/aarch64-option-extension.def: Add
17339 AARCH64_OPT_EXTENSION of 'fp16fml'.
17340 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17341 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17342 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17343 * config/aarch64/constraints.md (Ui7): New constraint.
17344 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17345 (VFMLA_SEL_W): Ditto.
17348 (VFMLA16_LOW): New int iterator.
17349 (VFMLA16_HIGH): Ditto.
17350 (UNSPEC_FMLAL): New unspec.
17351 (UNSPEC_FMLSL): Ditto.
17352 (UNSPEC_FMLAL2): Ditto.
17353 (UNSPEC_FMLSL2): Ditto.
17354 (f16mac): New code attribute.
17355 * config/aarch64/aarch64-simd-builtins.def
17356 (aarch64_fmlal_lowv2sf): Ditto.
17357 (aarch64_fmlsl_lowv2sf): Ditto.
17358 (aarch64_fmlalq_lowv4sf): Ditto.
17359 (aarch64_fmlslq_lowv4sf): Ditto.
17360 (aarch64_fmlal_highv2sf): Ditto.
17361 (aarch64_fmlsl_highv2sf): Ditto.
17362 (aarch64_fmlalq_highv4sf): Ditto.
17363 (aarch64_fmlslq_highv4sf): Ditto.
17364 (aarch64_fmlal_lane_lowv2sf): Ditto.
17365 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17366 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17367 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17368 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17369 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17370 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17371 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17372 (aarch64_fmlal_lane_highv2sf): Ditto.
17373 (aarch64_fmlsl_lane_highv2sf): Ditto.
17374 (aarch64_fmlal_laneq_highv2sf): Ditto.
17375 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17376 (aarch64_fmlalq_lane_highv4sf): Ditto.
17377 (aarch64_fmlsl_lane_highv4sf): Ditto.
17378 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17379 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17380 * config/aarch64/aarch64-simd.md:
17381 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17382 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17383 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17384 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17385 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17386 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17387 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17388 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17389 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17390 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17391 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17392 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17393 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17394 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17395 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17396 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17397 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17398 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17399 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17400 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17401 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17402 (vfmlsl_low_u32): Ditto.
17403 (vfmlalq_low_u32): Ditto.
17404 (vfmlslq_low_u32): Ditto.
17405 (vfmlal_high_u32): Ditto.
17406 (vfmlsl_high_u32): Ditto.
17407 (vfmlalq_high_u32): Ditto.
17408 (vfmlslq_high_u32): Ditto.
17409 (vfmlal_lane_low_u32): Ditto.
17410 (vfmlsl_lane_low_u32): Ditto.
17411 (vfmlal_laneq_low_u32): Ditto.
17412 (vfmlsl_laneq_low_u32): Ditto.
17413 (vfmlalq_lane_low_u32): Ditto.
17414 (vfmlslq_lane_low_u32): Ditto.
17415 (vfmlalq_laneq_low_u32): Ditto.
17416 (vfmlslq_laneq_low_u32): Ditto.
17417 (vfmlal_lane_high_u32): Ditto.
17418 (vfmlsl_lane_high_u32): Ditto.
17419 (vfmlal_laneq_high_u32): Ditto.
17420 (vfmlsl_laneq_high_u32): Ditto.
17421 (vfmlalq_lane_high_u32): Ditto.
17422 (vfmlslq_lane_high_u32): Ditto.
17423 (vfmlalq_laneq_high_u32): Ditto.
17424 (vfmlslq_laneq_high_u32): Ditto.
17425 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17426 (AARCH64_FL_FOR_ARCH8_4): New.
17427 (AARCH64_ISA_F16FML): New ISA flag.
17428 (TARGET_F16FML): New feature flag for fp16fml.
17429 (doc/invoke.texi): Document new fp16fml option.
17431 2018-01-10 Michael Collison <michael.collison@arm.com>
17433 * config/aarch64/aarch64-builtins.c:
17434 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17435 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17436 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17437 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17438 (AARCH64_ISA_SHA3): New ISA flag.
17439 (TARGET_SHA3): New feature flag for sha3.
17440 * config/aarch64/iterators.md (sha512_op): New int attribute.
17441 (CRYPTO_SHA512): New int iterator.
17442 (UNSPEC_SHA512H): New unspec.
17443 (UNSPEC_SHA512H2): Ditto.
17444 (UNSPEC_SHA512SU0): Ditto.
17445 (UNSPEC_SHA512SU1): Ditto.
17446 * config/aarch64/aarch64-simd-builtins.def
17447 (aarch64_crypto_sha512hqv2di): New builtin.
17448 (aarch64_crypto_sha512h2qv2di): Ditto.
17449 (aarch64_crypto_sha512su0qv2di): Ditto.
17450 (aarch64_crypto_sha512su1qv2di): Ditto.
17451 (aarch64_eor3qv8hi): Ditto.
17452 (aarch64_rax1qv2di): Ditto.
17453 (aarch64_xarqv2di): Ditto.
17454 (aarch64_bcaxqv8hi): Ditto.
17455 * config/aarch64/aarch64-simd.md:
17456 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17457 (aarch64_crypto_sha512su0qv2di): Ditto.
17458 (aarch64_crypto_sha512su1qv2di): Ditto.
17459 (aarch64_eor3qv8hi): Ditto.
17460 (aarch64_rax1qv2di): Ditto.
17461 (aarch64_xarqv2di): Ditto.
17462 (aarch64_bcaxqv8hi): Ditto.
17463 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17464 (vsha512h2q_u64): Ditto.
17465 (vsha512su0q_u64): Ditto.
17466 (vsha512su1q_u64): Ditto.
17467 (veor3q_u16): Ditto.
17468 (vrax1q_u64): Ditto.
17469 (vxarq_u64): Ditto.
17470 (vbcaxq_u16): Ditto.
17471 * config/arm/types.md (crypto_sha512): New type attribute.
17472 (crypto_sha3): Ditto.
17473 (doc/invoke.texi): Document new sha3 option.
17475 2018-01-10 Michael Collison <michael.collison@arm.com>
17477 * config/aarch64/aarch64-builtins.c:
17478 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17479 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17480 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17481 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17482 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17483 (AARCH64_ISA_SM4): New ISA flag.
17484 (TARGET_SM4): New feature flag for sm4.
17485 * config/aarch64/aarch64-simd-builtins.def
17486 (aarch64_sm3ss1qv4si): Ditto.
17487 (aarch64_sm3tt1aq4si): Ditto.
17488 (aarch64_sm3tt1bq4si): Ditto.
17489 (aarch64_sm3tt2aq4si): Ditto.
17490 (aarch64_sm3tt2bq4si): Ditto.
17491 (aarch64_sm3partw1qv4si): Ditto.
17492 (aarch64_sm3partw2qv4si): Ditto.
17493 (aarch64_sm4eqv4si): Ditto.
17494 (aarch64_sm4ekeyqv4si): Ditto.
17495 * config/aarch64/aarch64-simd.md:
17496 (aarch64_sm3ss1qv4si): Ditto.
17497 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17498 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17499 (aarch64_sm4eqv4si): Ditto.
17500 (aarch64_sm4ekeyqv4si): Ditto.
17501 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17502 (sm3part_op): Ditto.
17503 (CRYPTO_SM3TT): Ditto.
17504 (CRYPTO_SM3PART): Ditto.
17505 (UNSPEC_SM3SS1): New unspec.
17506 (UNSPEC_SM3TT1A): Ditto.
17507 (UNSPEC_SM3TT1B): Ditto.
17508 (UNSPEC_SM3TT2A): Ditto.
17509 (UNSPEC_SM3TT2B): Ditto.
17510 (UNSPEC_SM3PARTW1): Ditto.
17511 (UNSPEC_SM3PARTW2): Ditto.
17512 (UNSPEC_SM4E): Ditto.
17513 (UNSPEC_SM4EKEY): Ditto.
17514 * config/aarch64/constraints.md (Ui2): New constraint.
17515 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17516 * config/arm/types.md (crypto_sm3): New type attribute.
17517 (crypto_sm4): Ditto.
17518 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17519 (vsm3tt1aq_u32): Ditto.
17520 (vsm3tt1bq_u32): Ditto.
17521 (vsm3tt2aq_u32): Ditto.
17522 (vsm3tt2bq_u32): Ditto.
17523 (vsm3partw1q_u32): Ditto.
17524 (vsm3partw2q_u32): Ditto.
17525 (vsm4eq_u32): Ditto.
17526 (vsm4ekeyq_u32): Ditto.
17527 (doc/invoke.texi): Document new sm4 option.
17529 2018-01-10 Michael Collison <michael.collison@arm.com>
17531 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17532 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17533 (AARCH64_FL_FOR_ARCH8_4): New.
17534 (AARCH64_FL_V8_4): New flag.
17535 (doc/invoke.texi): Document new armv8.4-a option.
17537 2018-01-10 Michael Collison <michael.collison@arm.com>
17539 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17540 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17541 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17542 * config/aarch64/aarch64-option-extension.def: Add
17543 AARCH64_OPT_EXTENSION of 'sha2'.
17544 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17545 (crypto): Disable sha2 and aes if crypto disabled.
17546 (crypto): Enable aes and sha2 if enabled.
17547 (simd): Disable sha2 and aes if simd disabled.
17548 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17550 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17551 (TARGET_SHA2): New feature flag for sha2.
17552 (TARGET_AES): New feature flag for aes.
17553 * config/aarch64/aarch64-simd.md:
17554 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17555 conditional on TARGET_AES.
17556 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17557 (aarch64_crypto_sha1hsi): Make pattern conditional
17559 (aarch64_crypto_sha1hv4si): Ditto.
17560 (aarch64_be_crypto_sha1hv4si): Ditto.
17561 (aarch64_crypto_sha1su1v4si): Ditto.
17562 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17563 (aarch64_crypto_sha1su0v4si): Ditto.
17564 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17565 (aarch64_crypto_sha256su0v4si): Ditto.
17566 (aarch64_crypto_sha256su1v4si): Ditto.
17567 (doc/invoke.texi): Document new aes and sha2 options.
17569 2018-01-10 Martin Sebor <msebor@redhat.com>
17571 PR tree-optimization/83781
17572 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17575 2018-01-11 Martin Sebor <msebor@gmail.com>
17576 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17578 PR tree-optimization/83501
17579 PR tree-optimization/81703
17581 * tree-ssa-strlen.c (get_string_cst): Rename...
17582 (get_string_len): ...to this. Handle global constants.
17583 (handle_char_store): Adjust.
17585 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17586 Jim Wilson <jimw@sifive.com>
17588 * config/riscv/riscv-protos.h (riscv_output_return): New.
17589 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17590 (riscv_attribute_table, riscv_output_return),
17591 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17592 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17593 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17595 (riscv_expand_prologue): Add early return for naked function.
17596 (riscv_expand_epilogue): Likewise.
17597 (riscv_function_ok_for_sibcall): Return false for naked function.
17598 (riscv_set_current_function): New.
17599 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17600 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17601 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17602 * doc/extend.texi (RISC-V Function Attributes): New.
17604 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17606 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17607 check for 128-bit long double before checking TCmode.
17608 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17609 128-bit long doubles before checking TFmode or TCmode.
17610 (FLOAT128_IBM_P): Likewise.
17612 2018-01-10 Martin Sebor <msebor@redhat.com>
17614 PR tree-optimization/83671
17615 * builtins.c (c_strlen): Unconditionally return zero for the empty
17617 Use -Warray-bounds for warnings.
17618 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17619 for non-constant array indices with COMPONENT_REF, arrays of
17620 arrays, and pointers to arrays.
17621 (gimple_fold_builtin_strlen): Determine and set length range for
17622 non-constant character arrays.
17624 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17626 PR middle-end/81897
17627 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17630 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17632 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17634 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17637 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17638 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17639 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17640 indexed_or_indirect_operand predicate.
17641 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17642 (*vsx_le_perm_load_v8hi): Likewise.
17643 (*vsx_le_perm_load_v16qi): Likewise.
17644 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17645 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17646 (*vsx_le_perm_store_v8hi): Likewise.
17647 (*vsx_le_perm_store_v16qi): Likewise.
17648 (eight unnamed splitters): Likewise.
17650 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17652 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17653 * config/rs6000/emmintrin.h: Likewise.
17654 * config/rs6000/mmintrin.h: Likewise.
17655 * config/rs6000/xmmintrin.h: Likewise.
17657 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17660 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17662 * tree.c (tree_nop_conversion): Return true for location wrapper
17664 (maybe_wrap_with_location): New function.
17665 (selftest::check_strip_nops): New function.
17666 (selftest::test_location_wrappers): New function.
17667 (selftest::tree_c_tests): Call it.
17668 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17669 (maybe_wrap_with_location): New decl.
17670 (EXPR_LOCATION_WRAPPER_P): New macro.
17671 (location_wrapper_p): New inline function.
17672 (tree_strip_any_location_wrapper): New inline function.
17674 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17677 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17678 stack_realign_offset for the largest alignment of stack slot
17680 (ix86_find_max_used_stack_alignment): New function.
17681 (ix86_finalize_stack_frame_flags): Use it. Set
17682 max_used_stack_alignment if we don't realign stack.
17683 * config/i386/i386.h (machine_function): Add
17684 max_used_stack_alignment.
17686 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17688 * config/arm/arm.opt (-mbranch-cost): New option.
17689 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17692 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17695 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17696 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17698 2018-01-10 Richard Biener <rguenther@suse.de>
17701 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17702 early out so it also covers the case where we have a non-NULL
17705 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17707 PR tree-optimization/83753
17708 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17709 for non-strided grouped accesses if the number of elements is 1.
17711 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17714 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17715 * i386.h (TARGET_USE_GATHER): Define.
17716 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17718 2018-01-10 Martin Liska <mliska@suse.cz>
17721 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17722 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17724 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17725 CLEANUP_NO_PARTITIONING is not set.
17727 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17729 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17730 for vectors, as a partial revert of r254296.
17731 * rtl.h (const_vec_p): Delete.
17732 (const_vec_duplicate_p): Don't test for vector CONSTs.
17733 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17734 * expmed.c (make_tree): Likewise.
17737 * common.md (E, F): Use CONSTANT_P instead of checking for
17739 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17740 checking for CONST_VECTOR.
17742 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17744 PR middle-end/83575
17745 * predict.c (force_edge_cold): Handle in more sane way edges
17746 with no prediction.
17748 2018-01-09 Carl Love <cel@us.ibm.com>
17750 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17752 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17753 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17754 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17755 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17756 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17757 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17758 * config/rs6000/rs6000-protos.h: Add extern defition for
17759 rs6000_generate_float2_double_code.
17760 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17762 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17763 (float2_v2df): Add define_expand.
17765 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17768 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17769 op_mode in the force_to_mode call.
17771 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17773 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17774 instead of checking each element individually.
17775 (aarch64_evpc_uzp): Likewise.
17776 (aarch64_evpc_zip): Likewise.
17777 (aarch64_evpc_ext): Likewise.
17778 (aarch64_evpc_rev): Likewise.
17779 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17780 instead of checking each element individually. Return true without
17782 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17783 whether all selected elements come from the same input, instead of
17784 checking each element individually. Remove calls to gen_rtx_REG,
17785 start_sequence and end_sequence and instead assert that no rtl is
17788 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17790 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17791 order of HIGH and CONST checks.
17793 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17795 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17796 if the destination isn't an SSA_NAME.
17798 2018-01-09 Richard Biener <rguenther@suse.de>
17800 PR tree-optimization/83668
17801 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17803 (canonicalize_loop_form): ... here, renamed from ...
17804 (canonicalize_loop_closed_ssa_form): ... this and amended to
17805 swap successor edges for loop exit blocks to make us use
17806 the RPO order we need for initial schedule generation.
17808 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17810 PR tree-optimization/64811
17811 * match.pd: When optimizing comparisons with Inf, avoid
17812 introducing or losing exceptions from comparisons with NaN.
17814 2018-01-09 Martin Liska <mliska@suse.cz>
17817 * asan.c (shadow_mem_size): Add gcc_assert.
17819 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17821 Don't save registers in main().
17824 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17825 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17826 * config/avr/avr.c (avr_set_current_function): Don't error if
17827 naked, OS_task or OS_main are specified at the same time.
17828 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17830 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17832 * common/config/avr/avr-common.c (avr_option_optimization_table):
17833 Switch on -mmain-is-OS_task for optimizing compilations.
17835 2018-01-09 Richard Biener <rguenther@suse.de>
17837 PR tree-optimization/83572
17838 * graphite.c: Include cfganal.h.
17839 (graphite_transform_loops): Connect infinite loops to exit
17840 and remove fake edges at the end.
17842 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17844 * ipa-inline.c (edge_badness): Revert accidental checkin.
17846 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17849 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17850 symbols; not inline clones.
17852 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17855 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17856 hard registers. Formatting fixes.
17858 PR preprocessor/83722
17859 * gcc.c (try_generate_repro): Pass
17860 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17861 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17864 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17865 Kito Cheng <kito.cheng@gmail.com>
17867 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17868 (riscv_leaf_function_p): Delete.
17869 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17871 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17873 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17875 (do_ifelse): New function.
17876 (do_isel): New function.
17877 (do_sub3): New function.
17878 (do_add3): New function.
17879 (do_load_mask_compare): New function.
17880 (do_overlap_load_compare): New function.
17881 (expand_compare_loop): New function.
17882 (expand_block_compare): Call expand_compare_loop() when appropriate.
17883 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17884 option description.
17885 (-mblock-compare-inline-loop-limit): New option.
17887 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17890 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17891 Reverse order of second and third operands in first alternative.
17892 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17893 of first and second elements in UNSPEC_VPERMR vector.
17894 (altivec_expand_vec_perm_le): Likewise.
17896 2018-01-08 Jeff Law <law@redhat.com>
17898 PR rtl-optimizatin/81308
17899 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17900 (process_switch): If group_case_labels makes a change, then set
17902 (pass_convert_switch::execute): If a switch is converted, then
17903 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17905 PR rtl-optimization/81308
17906 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17909 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17911 PR target/83663 - Revert r255946
17912 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17913 generation for cases where splatting a value is not useful.
17914 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17915 across a vec_duplicate and a paradoxical subreg forming a vector
17916 mode to a vec_concat.
17918 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17920 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17921 -march=armv8.3-a variants.
17922 * config/arm/t-multilib: Likewise.
17923 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17925 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17927 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17929 (cceq_ior_compare_complement): Give it a name so I can use it, and
17930 change boolean_or_operator predicate to boolean_operator so it can
17931 be used to generate a crand.
17932 (eqne): New code iterator.
17933 (bd/bd_neg): New code_attrs.
17934 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17935 a single define_insn.
17936 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17937 decrement (bdnzt/bdnzf/bdzt/bdzf).
17938 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17939 with the new names of the branch decrement patterns, and added the
17940 names of the branch decrement conditional patterns.
17942 2018-01-08 Richard Biener <rguenther@suse.de>
17944 PR tree-optimization/83563
17945 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17948 2018-01-08 Richard Biener <rguenther@suse.de>
17950 PR middle-end/83713
17951 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17953 2018-01-08 Richard Biener <rguenther@suse.de>
17955 PR tree-optimization/83685
17956 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17957 references to abnormals.
17959 2018-01-08 Richard Biener <rguenther@suse.de>
17962 * dwarf2out.c (output_indirect_strings): Handle empty
17963 skeleton_debug_str_hash.
17964 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17966 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17968 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17969 (emit_store_direct): Likewise.
17970 (arc_trampoline_adjust_address): Likewise.
17971 (arc_asm_trampoline_template): New function.
17972 (arc_initialize_trampoline): Use asm_trampoline_template.
17973 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17974 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17975 * config/arc/arc.md (flush_icache): Delete pattern.
17977 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17979 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17980 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17983 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17986 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17987 by not USED_FOR_TARGET.
17988 (make_pass_resolve_sw_modes): Likewise.
17990 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17992 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17995 2018-01-08 Richard Biener <rguenther@suse.de>
17997 PR middle-end/83580
17998 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18000 2018-01-08 Richard Biener <rguenther@suse.de>
18002 PR middle-end/83517
18003 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18005 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18007 PR middle-end/81897
18008 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18009 basic blocks with a small number of successors.
18010 (convert_control_dep_chain_into_preds): Improve handling of
18012 (dump_predicates): Split apart into...
18013 (dump_pred_chain): ...here...
18014 (dump_pred_info): ...and here.
18015 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18016 (can_chain_union_be_invalidated_p): Improve check for invalidation
18018 (uninit_uses_cannot_happen): Avoid unnecessary if
18019 convert_control_dep_chain_into_preds yielded nothing.
18021 2018-01-06 Martin Sebor <msebor@redhat.com>
18023 PR tree-optimization/83640
18024 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18025 subtracting negative offset from size.
18026 (builtin_access::overlap): Adjust offset bounds of the access to fall
18027 within the size of the object if possible.
18029 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18031 PR rtl-optimization/83699
18032 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18033 extract_bit_field_as_subreg to cases in which the extracted
18034 value is also a vector.
18036 * lra-constraints.c (process_alt_operands): Test for the equivalence
18037 substitutions when detecting a possible reload cycle.
18039 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18042 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18043 by default if flag_selective_schedling{,2}. Formatting fixes.
18045 PR rtl-optimization/83682
18046 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18047 if it has non-VECTOR_MODE element mode.
18048 (vec_duplicate_p): Likewise.
18050 PR middle-end/83694
18051 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18052 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18054 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18057 * config/i386/i386-builtin.def
18058 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18059 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18060 Require also OPTION_MASK_ISA_AVX512F in addition to
18061 OPTION_MASK_ISA_GFNI.
18062 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18063 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18064 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18065 to OPTION_MASK_ISA_GFNI.
18066 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18067 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18068 OPTION_MASK_ISA_AVX512BW.
18069 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18070 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18071 addition to OPTION_MASK_ISA_GFNI.
18072 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18073 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18074 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18075 to OPTION_MASK_ISA_GFNI.
18076 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18077 a requirement for all ISAs rather than any of them with a few
18079 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18081 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18082 bitmasks to be enabled with 3 exceptions, instead of requiring any
18083 enabled ISA with lots of exceptions.
18084 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18085 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18086 Change avx512bw in isa attribute to avx512f.
18087 * config/i386/sgxintrin.h: Add license boilerplate.
18088 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18089 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18090 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18091 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18093 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18094 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18095 temporarily sse2 rather than sse if not enabled already.
18098 * config/i386/sse.md (VI248_VLBW): Rename to ...
18099 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18100 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18101 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18102 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18103 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18104 mode iterator instead of VI248_VLBW.
18106 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18108 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18109 (record_modified): Skip clobbers; add debug output.
18110 (param_change_prob): Use sreal frequencies.
18112 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18114 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18115 punt for user-aligned variables.
18117 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18119 * tree-chrec.c (chrec_contains_symbols): Return true for
18122 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18125 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18126 of (x|y) == x for BICS pattern.
18128 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18130 PR tree-optimization/83605
18131 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18132 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18135 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18137 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18138 * config/epiphany/rtems.h: New file.
18140 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18141 Uros Bizjak <ubizjak@gmail.com>
18144 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18145 QIreg_operand instead of register_operand predicate.
18146 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18147 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18148 comments instead of -fmitigate[-_]rop.
18150 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18153 * cgraphunit.c (symbol_table::compile): Switch to text_section
18154 before calling assembly_start debug hook.
18155 * run-rtl-passes.c (run_rtl_passes): Likewise.
18158 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18160 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18161 range_int_cst_p rather than !symbolic_range_p before calling
18162 extract_range_from_multiplicative_op_1.
18164 2018-01-04 Jeff Law <law@redhat.com>
18166 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18167 redundant test in assertion.
18169 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18171 * doc/rtl.texi: Document machine_mode wrapper classes.
18173 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18175 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18176 using tree_to_uhwi.
18178 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18180 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18181 the VEC_PERM_EXPR fold to fail.
18183 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18186 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18187 to switched_sections.
18189 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18192 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18193 test for d.testing.
18195 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18198 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18199 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18201 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18204 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18205 is BLKmode and bitpos not zero or mode change is needed.
18207 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18210 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18213 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18216 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18217 instead of MULT rtx. Update all corresponding splitters.
18218 (*saddl_se): Ditto.
18219 (*ssub<modesuffix>): Ditto.
18220 (*ssubl_se): Ditto.
18221 (*cmp_sadd_di): Update split patterns.
18222 (*cmp_sadd_si): Ditto.
18223 (*cmp_sadd_sidi): Ditto.
18224 (*cmp_ssub_di): Ditto.
18225 (*cmp_ssub_si): Ditto.
18226 (*cmp_ssub_sidi): Ditto.
18227 * config/alpha/predicates.md (const23_operand): New predicate.
18228 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18229 Look for ASHIFT, not MULT inner operand.
18230 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18232 2018-01-04 Martin Liska <mliska@suse.cz>
18234 PR gcov-profile/83669
18235 * gcov.c (output_intermediate_file): Add version to intermediate
18237 * doc/gcov.texi: Document new field 'version' in intermediate
18238 file format. Fix location of '-k' option of gcov command.
18240 2018-01-04 Martin Liska <mliska@suse.cz>
18243 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18245 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18247 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18249 2018-01-03 Martin Sebor <msebor@redhat.com>
18251 PR tree-optimization/83655
18252 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18253 checking calls with invalid arguments.
18255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18257 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18258 (vectorizable_mask_load_store): Delete.
18259 (vectorizable_call): Return false for masked loads and stores.
18260 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18261 instead of gimple_assign_rhs1.
18262 (vectorizable_load): Handle IFN_MASK_LOAD.
18263 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18267 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18269 (vectorizable_mask_load_store): ...here.
18270 (vectorizable_load): ...and here.
18272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18274 * tree-vect-stmts.c (vect_build_all_ones_mask)
18275 (vect_build_zero_merge_argument): New functions, split out from...
18276 (vectorizable_load): ...here.
18278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18280 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18282 (vectorizable_mask_load_store): ...here.
18283 (vectorizable_store): ...and here.
18285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18287 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18289 (vectorizable_mask_load_store): ...here.
18291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18293 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18294 (vect_model_store_cost): Take a vec_load_store_type instead of a
18296 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18297 (vect_model_store_cost): Take a vec_load_store_type instead of a
18299 (vectorizable_mask_load_store): Update accordingly.
18300 (vectorizable_store): Likewise.
18301 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18305 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18306 IFN_MASK_LOAD calls here rather than...
18307 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18310 Alan Hayward <alan.hayward@arm.com>
18311 David Sherwood <david.sherwood@arm.com>
18313 * expmed.c (extract_bit_field_1): For vector extracts,
18314 fall back to extract_bit_field_as_subreg if vec_extract
18317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18318 Alan Hayward <alan.hayward@arm.com>
18319 David Sherwood <david.sherwood@arm.com>
18321 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18322 they are variable or constant sized.
18323 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18324 slots for constant-sized data.
18326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18327 Alan Hayward <alan.hayward@arm.com>
18328 David Sherwood <david.sherwood@arm.com>
18330 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18331 handling COND_EXPRs with boolean comparisons, try to find a better
18332 basis for the mask type than the boolean itself.
18334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18336 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18337 is calculated and how it can be overridden.
18338 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18339 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18341 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18345 Alan Hayward <alan.hayward@arm.com>
18346 David Sherwood <david.sherwood@arm.com>
18348 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18349 Remove the mode argument.
18350 (aarch64_simd_valid_immediate): Remove the mode and inverse
18352 * config/aarch64/iterators.md (bitsize): New iterator.
18353 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18354 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18355 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18356 aarch64_simd_valid_immediate.
18357 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18358 (aarch64_reg_or_bic_imm): Likewise.
18359 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18360 with an insn_type enum and msl with a modifier_type enum.
18361 Replace element_width with a scalar_mode. Change the shift
18362 to unsigned int. Add constructors for scalar_float_mode and
18363 scalar_int_mode elements.
18364 (aarch64_vect_float_const_representable_p): Delete.
18365 (aarch64_can_const_movi_rtx_p)
18366 (aarch64_simd_scalar_immediate_valid_for_move)
18367 (aarch64_simd_make_constant): Update call to
18368 aarch64_simd_valid_immediate.
18369 (aarch64_advsimd_valid_immediate_hs): New function.
18370 (aarch64_advsimd_valid_immediate): Likewise.
18371 (aarch64_simd_valid_immediate): Remove mode and inverse
18372 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18373 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18374 and aarch64_float_const_representable_p on the result.
18375 (aarch64_output_simd_mov_immediate): Remove mode argument.
18376 Update call to aarch64_simd_valid_immediate and use of
18377 simd_immediate_info.
18378 (aarch64_output_scalar_simd_mov_immediate): Update call
18381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18382 Alan Hayward <alan.hayward@arm.com>
18383 David Sherwood <david.sherwood@arm.com>
18385 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18386 (mode_nunits): Likewise CONST_MODE_NUNITS.
18387 * machmode.def (ADJUST_NUNITS): Document.
18388 * genmodes.c (mode_data::need_nunits_adj): New field.
18389 (blank_mode): Update accordingly.
18390 (adj_nunits): New variable.
18391 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18393 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18394 listed in adj_nunits.
18395 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18396 listed in adj_nunits. Don't emit case statements for such modes.
18397 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18398 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18399 nothing if adj_nunits is nonnull.
18400 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18401 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18402 (emit_mode_fbit): Update use of print_maybe_const_decl.
18403 (emit_move_size): Likewise. Treat the array as non-const
18405 (emit_mode_adjustments): Handle adj_nunits.
18407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18409 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18410 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18411 (VECTOR_MODES): Use it.
18412 (make_vector_modes): Take the prefix as an argument.
18414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18415 Alan Hayward <alan.hayward@arm.com>
18416 David Sherwood <david.sherwood@arm.com>
18418 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18419 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18420 for MODE_VECTOR_BOOL.
18421 * machmode.def (VECTOR_BOOL_MODE): Document.
18422 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18423 (make_vector_bool_mode): New function.
18424 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18426 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18427 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18429 * stor-layout.c (int_mode_for_mode): Likewise.
18430 * tree.c (build_vector_type_for_mode): Likewise.
18431 * varasm.c (output_constant_pool_2): Likewise.
18432 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18433 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18434 for MODE_VECTOR_BOOL.
18435 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18436 of mode class checks.
18437 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18438 instead of a list of mode class checks.
18439 (expand_vector_scalar_condition): Likewise.
18440 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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 * machmode.h (mode_size): Change from unsigned short to
18448 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18449 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18450 or if measurement_type is not polynomial.
18451 (fixed_size_mode::includes_p): Check for constant-sized modes.
18452 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18453 return a poly_uint16 rather than an unsigned short.
18454 (emit_mode_size): Change the type of mode_size from unsigned short
18455 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18456 (emit_mode_adjustments): Cope with polynomial vector sizes.
18457 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18459 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18461 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18462 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18463 * caller-save.c (setup_save_areas): Likewise.
18464 (replace_reg_with_saved_mem): Likewise.
18465 * calls.c (emit_library_call_value_1): Likewise.
18466 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18467 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18468 (gen_lowpart_for_combine): Likewise.
18469 * convert.c (convert_to_integer_1): Likewise.
18470 * cse.c (equiv_constant, cse_insn): Likewise.
18471 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18472 (cselib_subst_to_values): Likewise.
18473 * dce.c (word_dce_process_block): Likewise.
18474 * df-problems.c (df_word_lr_mark_ref): Likewise.
18475 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18476 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18477 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18478 (rtl_for_decl_location): Likewise.
18479 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18480 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18481 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18482 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18483 (expand_expr_real_1): Likewise.
18484 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18485 (pad_below): Likewise.
18486 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18487 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18488 * ira.c (get_subreg_tracking_sizes): Likewise.
18489 * ira-build.c (ira_create_allocno_objects): Likewise.
18490 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18491 (ira_sort_regnos_for_alter_reg): Likewise.
18492 * ira-costs.c (record_operand_costs): Likewise.
18493 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18494 (resolve_simple_move): Likewise.
18495 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18496 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18497 (lra_constraints): Likewise.
18498 (CONST_POOL_OK_P): Reject variable-sized modes.
18499 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18500 (add_pseudo_to_slot, lra_spill): Likewise.
18501 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18502 * optabs-query.c (get_best_extraction_insn): Likewise.
18503 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18504 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18505 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18506 * recog.c (offsettable_address_addr_space_p): Likewise.
18507 * regcprop.c (maybe_mode_change): Likewise.
18508 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18509 * regrename.c (build_def_use): Likewise.
18510 * regstat.c (dump_reg_info): Likewise.
18511 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18512 (find_reloads, find_reloads_subreg_address): Likewise.
18513 * reload1.c (eliminate_regs_1): Likewise.
18514 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18515 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18516 (simplify_binary_operation_1, simplify_subreg): Likewise.
18517 * targhooks.c (default_function_arg_padding): Likewise.
18518 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18519 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18520 (verify_gimple_assign_ternary): Likewise.
18521 * tree-inline.c (estimate_move_cost): Likewise.
18522 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18523 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18524 (get_address_cost_ainc): Likewise.
18525 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18526 (vect_supportable_dr_alignment): Likewise.
18527 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18528 (vectorizable_reduction): Likewise.
18529 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18530 (vectorizable_operation, vectorizable_load): Likewise.
18531 * tree.c (build_same_sized_truth_vector_type): Likewise.
18532 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18533 * var-tracking.c (emit_note_insn_var_location): Likewise.
18534 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18535 (ADDR_VEC_ALIGN): Likewise.
18537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18538 Alan Hayward <alan.hayward@arm.com>
18539 David Sherwood <david.sherwood@arm.com>
18541 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18543 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18544 or if measurement_type is polynomial.
18545 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18546 * combine.c (make_extraction): Likewise.
18547 * dse.c (find_shift_sequence): Likewise.
18548 * dwarf2out.c (mem_loc_descriptor): Likewise.
18549 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18550 (extract_bit_field, extract_low_bits): Likewise.
18551 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18552 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18553 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18554 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18555 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18556 * reload.c (find_reloads): Likewise.
18557 * reload1.c (alter_reg): Likewise.
18558 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18559 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18560 * tree-if-conv.c (predicate_mem_writes): Likewise.
18561 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18562 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18563 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18564 * valtrack.c (dead_debug_insert_temp): Likewise.
18565 * varasm.c (mergeable_constant_section): Likewise.
18566 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18569 Alan Hayward <alan.hayward@arm.com>
18570 David Sherwood <david.sherwood@arm.com>
18572 * expr.c (expand_assignment): Cope with polynomial mode sizes
18573 when assigning to a CONCAT.
18575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18576 Alan Hayward <alan.hayward@arm.com>
18577 David Sherwood <david.sherwood@arm.com>
18579 * machmode.h (mode_precision): Change from unsigned short to
18581 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18583 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18584 or if measurement_type is not polynomial.
18585 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18586 in which the mode is already known to be a scalar_int_mode.
18587 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18588 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18590 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18591 for GET_MODE_PRECISION.
18592 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18593 for GET_MODE_PRECISION.
18594 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18596 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18597 (expand_field_assignment, make_extraction): Likewise.
18598 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18599 (get_last_value): Likewise.
18600 * convert.c (convert_to_integer_1): Likewise.
18601 * cse.c (cse_insn): Likewise.
18602 * expr.c (expand_expr_real_1): Likewise.
18603 * lra-constraints.c (simplify_operand_subreg): Likewise.
18604 * optabs-query.c (can_atomic_load_p): Likewise.
18605 * optabs.c (expand_atomic_load): Likewise.
18606 (expand_atomic_store): Likewise.
18607 * ree.c (combine_reaching_defs): Likewise.
18608 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18609 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18610 * tree.h (type_has_mode_precision_p): Likewise.
18611 * ubsan.c (instrument_si_overflow): Likewise.
18613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18614 Alan Hayward <alan.hayward@arm.com>
18615 David Sherwood <david.sherwood@arm.com>
18617 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18618 polynomial numbers of units.
18619 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18620 (valid_vector_subparts_p): New function.
18621 (build_vector_type): Remove temporary shim and take the number
18622 of units as a poly_uint64 rather than an int.
18623 (build_opaque_vector_type): Take the number of units as a
18624 poly_uint64 rather than an int.
18625 * tree.c (build_vector_from_ctor): Handle polynomial
18626 TYPE_VECTOR_SUBPARTS.
18627 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18628 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18629 (build_vector_from_val): If the number of units is variable,
18630 use build_vec_duplicate_cst for constant operands and
18631 VEC_DUPLICATE_EXPR otherwise.
18632 (make_vector_type): Remove temporary is_constant ().
18633 (build_vector_type, build_opaque_vector_type): Take the number of
18634 units as a poly_uint64 rather than an int.
18635 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18637 * cfgexpand.c (expand_debug_expr): Likewise.
18638 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18639 (store_constructor, expand_expr_real_1): Likewise.
18640 (const_scalar_mask_from_tree): Likewise.
18641 * fold-const-call.c (fold_const_reduction): Likewise.
18642 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18643 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18644 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18645 (fold_relational_const): Likewise.
18646 (native_interpret_vector): Likewise. Change the size from an
18647 int to an unsigned int.
18648 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18649 TYPE_VECTOR_SUBPARTS.
18650 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18651 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18652 duplicating a non-constant operand into a variable-length vector.
18653 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18654 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18655 * ipa-icf.c (sem_variable::equals): Likewise.
18656 * match.pd: Likewise.
18657 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18658 * print-tree.c (print_node): Likewise.
18659 * stor-layout.c (layout_type): Likewise.
18660 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18661 * tree-cfg.c (verify_gimple_comparison): Likewise.
18662 (verify_gimple_assign_binary): Likewise.
18663 (verify_gimple_assign_ternary): Likewise.
18664 (verify_gimple_assign_single): Likewise.
18665 * tree-pretty-print.c (dump_generic_node): Likewise.
18666 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18667 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18668 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18669 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18670 (vect_shift_permute_load_chain): Likewise.
18671 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18672 (expand_vector_condition, optimize_vector_constructor): Likewise.
18673 (lower_vec_perm, get_compute_type): Likewise.
18674 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18675 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18676 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18677 (vect_recog_mask_conversion_pattern): Likewise.
18678 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18679 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18680 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18681 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18682 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18683 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18684 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18685 (supportable_widening_operation): Likewise.
18686 (supportable_narrowing_operation): Likewise.
18687 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18689 * varasm.c (output_constant): Likewise.
18691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18692 Alan Hayward <alan.hayward@arm.com>
18693 David Sherwood <david.sherwood@arm.com>
18695 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18696 so that both the length == 3 and length != 3 cases set up their
18697 own permute vectors. Add comments explaining why we know the
18698 number of elements is constant.
18699 (vect_permute_load_chain): Likewise.
18701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18702 Alan Hayward <alan.hayward@arm.com>
18703 David Sherwood <david.sherwood@arm.com>
18705 * machmode.h (mode_nunits): Change from unsigned char to
18707 (ONLY_FIXED_SIZE_MODES): New macro.
18708 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18709 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18710 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18712 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18713 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18714 or if measurement_type is not polynomial.
18715 * genmodes.c (ZERO_COEFFS): New macro.
18716 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18718 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18719 Use ZERO_COEFFS when emitting initializers.
18720 * data-streamer.h (bp_pack_poly_value): New function.
18721 (bp_unpack_poly_value): Likewise.
18722 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18723 for GET_MODE_NUNITS.
18724 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18725 for GET_MODE_NUNITS.
18726 * tree.c (make_vector_type): Remove temporary shim and make
18727 the real function take the number of units as a poly_uint64
18728 rather than an int.
18729 (build_vector_type_for_mode): Handle polynomial nunits.
18730 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18731 * emit-rtl.c (const_vec_series_p_1): Likewise.
18732 (gen_rtx_CONST_VECTOR): Likewise.
18733 * fold-const.c (test_vec_duplicate_folding): Likewise.
18734 * genrecog.c (validate_pattern): Likewise.
18735 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18736 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18737 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18738 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18739 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18740 * rtlanal.c (subreg_get_info): Likewise.
18741 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18742 (vect_grouped_load_supported): Likewise.
18743 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18744 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18745 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18746 (simplify_const_unary_operation, simplify_binary_operation_1)
18747 (simplify_const_binary_operation, simplify_ternary_operation)
18748 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18749 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18750 instead of CONST_VECTOR_NUNITS.
18751 * varasm.c (output_constant_pool_2): Likewise.
18752 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18753 explicit-encoded elements in the XVEC for variable-length vectors.
18755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18757 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18760 Alan Hayward <alan.hayward@arm.com>
18761 David Sherwood <david.sherwood@arm.com>
18763 * coretypes.h (fixed_size_mode): Declare.
18764 (fixed_size_mode_pod): New typedef.
18765 * builtins.h (target_builtins::x_apply_args_mode)
18766 (target_builtins::x_apply_result_mode): Change type to
18767 fixed_size_mode_pod.
18768 * builtins.c (apply_args_size, apply_result_size, result_vector)
18769 (expand_builtin_apply_args_1, expand_builtin_apply)
18770 (expand_builtin_return): Update accordingly.
18772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18774 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18775 * cselib.c (cselib_hash_rtx): Likewise.
18776 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18777 CONST_VECTOR encoding.
18779 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18780 Jeff Law <law@redhat.com>
18783 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18784 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18785 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18786 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18789 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18790 explicitly probe *sp in a noreturn function if there were any callee
18791 register saves or frame pointer is needed.
18793 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18796 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18797 BLKmode for ternary, binary or unary expressions.
18800 * var-tracking.c (delete_vta_debug_insn): New inline function.
18801 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18802 insns from get_insns () to NULL instead of each bb separately.
18803 Use delete_vta_debug_insn. No longer static.
18804 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18805 delete_vta_debug_insns callers.
18806 * rtl.h (delete_vta_debug_insns): Declare.
18807 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18808 instead of variable_tracking_main.
18810 2018-01-03 Martin Sebor <msebor@redhat.com>
18812 PR tree-optimization/83603
18813 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18814 arguments past the endof the argument list in functions declared
18815 without a prototype.
18816 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18817 Avoid checking when arguments are null.
18819 2018-01-03 Martin Sebor <msebor@redhat.com>
18822 * doc/extend.texi (attribute const): Fix a typo.
18823 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18824 issuing -Wsuggest-attribute for void functions.
18826 2018-01-03 Martin Sebor <msebor@redhat.com>
18828 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18829 offset_int::from instead of wide_int::to_shwi.
18830 (maybe_diag_overlap): Remove assertion.
18831 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18832 * gimple-ssa-sprintf.c (format_directive): Same.
18833 (parse_directive): Same.
18834 (sprintf_dom_walker::compute_format_length): Same.
18835 (try_substitute_return_value): Same.
18837 2018-01-03 Jeff Law <law@redhat.com>
18839 PR middle-end/83654
18840 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18841 non-constant residual for zero at runtime and avoid probing in
18842 that case. Reorganize code for trailing problem to mirror handling
18845 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18847 PR tree-optimization/83501
18848 * tree-ssa-strlen.c (get_string_cst): New.
18849 (handle_char_store): Call get_string_cst.
18851 2018-01-03 Martin Liska <mliska@suse.cz>
18853 PR tree-optimization/83593
18854 * tree-ssa-strlen.c: Include tree-cfg.h.
18855 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18856 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18857 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18859 (strlen_dom_walker::before_dom_children): Call
18860 gimple_purge_dead_eh_edges. Dump tranformation with details
18862 (strlen_dom_walker::before_dom_children): Update call by adding
18863 new argument cleanup_eh.
18864 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18866 2018-01-03 Martin Liska <mliska@suse.cz>
18869 * cif-code.def (VARIADIC_THUNK): New enum value.
18870 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18873 2018-01-03 Jan Beulich <jbeulich@suse.com>
18875 * sse.md (mov<mode>_internal): Tighten condition for when to use
18876 vmovdqu<ssescalarsize> for TI and OI modes.
18878 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18880 Update copyright years.
18882 2018-01-03 Martin Liska <mliska@suse.cz>
18885 * ipa-visibility.c (function_and_variable_visibility): Skip
18886 functions with noipa attribure.
18888 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18890 * gcc.c (process_command): Update copyright notice dates.
18891 * gcov-dump.c (print_version): Ditto.
18892 * gcov.c (print_version): Ditto.
18893 * gcov-tool.c (print_version): Ditto.
18894 * gengtype.c (create_file): Ditto.
18895 * doc/cpp.texi: Bump @copying's copyright year.
18896 * doc/cppinternals.texi: Ditto.
18897 * doc/gcc.texi: Ditto.
18898 * doc/gccint.texi: Ditto.
18899 * doc/gcov.texi: Ditto.
18900 * doc/install.texi: Ditto.
18901 * doc/invoke.texi: Ditto.
18903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18905 * vector-builder.h (vector_builder::m_full_nelts): Change from
18906 unsigned int to poly_uint64.
18907 (vector_builder::full_nelts): Update prototype accordingly.
18908 (vector_builder::new_vector): Likewise.
18909 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18910 (vector_builder::operator ==): Likewise.
18911 (vector_builder::finalize): Likewise.
18912 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18913 Take the number of elements as a poly_uint64 rather than an
18915 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18916 from unsigned int to poly_uint64.
18917 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18918 (vec_perm_indices::new_vector): Likewise.
18919 (vec_perm_indices::length): Likewise.
18920 (vec_perm_indices::nelts_per_input): Likewise.
18921 (vec_perm_indices::input_nelts): Likewise.
18922 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18923 number of elements per input as a poly_uint64 rather than an
18924 unsigned int. Use the original encoding for variable-length
18925 vectors, rather than clamping each individual element.
18926 For the second and subsequent elements in each pattern,
18927 clamp the step and base before clamping their sum.
18928 (vec_perm_indices::series_p): Handle polynomial element counts.
18929 (vec_perm_indices::all_in_range_p): Likewise.
18930 (vec_perm_indices_to_tree): Likewise.
18931 (vec_perm_indices_to_rtx): Likewise.
18932 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18933 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18934 (tree_vector_builder::new_binary_operation): Handle polynomial
18935 element counts. Return false if we need to know the number
18936 of elements at compile time.
18937 * fold-const.c (fold_vec_perm): Punt if the number of elements
18938 isn't known at compile time.
18940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18942 * vec-perm-indices.h (vec_perm_builder): Change element type
18943 from HOST_WIDE_INT to poly_int64.
18944 (vec_perm_indices::element_type): Update accordingly.
18945 (vec_perm_indices::clamp): Handle polynomial element_types.
18946 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18947 (vec_perm_indices::all_in_range_p): Likewise.
18948 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18950 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18951 polynomial vec_perm_indices element types.
18952 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18953 * fold-const.c (fold_vec_perm): Likewise.
18954 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18955 * tree-vect-generic.c (lower_vec_perm): Likewise.
18956 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18957 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18958 element type to HOST_WIDE_INT.
18960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18961 Alan Hayward <alan.hayward@arm.com>
18962 David Sherwood <david.sherwood@arm.com>
18964 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18965 rather than an int. Use plus_constant.
18966 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18967 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18970 Alan Hayward <alan.hayward@arm.com>
18971 David Sherwood <david.sherwood@arm.com>
18973 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18974 a HOST_WIDE_INT to a poly_int64.
18976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18977 Alan Hayward <alan.hayward@arm.com>
18978 David Sherwood <david.sherwood@arm.com>
18980 * calls.c (load_register_parameters): Cope with polynomial
18981 mode sizes. Require a constant size for BLKmode parameters
18982 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18983 forces a parameter to be padded at the lsb end in order to
18984 fill a complete number of words, require the parameter size
18985 to be ordered wrt UNITS_PER_WORD.
18987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18988 Alan Hayward <alan.hayward@arm.com>
18989 David Sherwood <david.sherwood@arm.com>
18991 * reload1.c (spill_stack_slot_width): Change element type
18992 from unsigned int to poly_uint64_pod.
18993 (alter_reg): Treat mode sizes as polynomial.
18995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18996 Alan Hayward <alan.hayward@arm.com>
18997 David Sherwood <david.sherwood@arm.com>
18999 * reload.c (complex_word_subreg_p): New function.
19000 (reload_inner_reg_of_subreg, push_reload): Use it.
19002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19003 Alan Hayward <alan.hayward@arm.com>
19004 David Sherwood <david.sherwood@arm.com>
19006 * lra-constraints.c (process_alt_operands): Reject matched
19007 operands whose sizes aren't ordered.
19008 (match_reload): Refer to this check here.
19010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19011 Alan Hayward <alan.hayward@arm.com>
19012 David Sherwood <david.sherwood@arm.com>
19014 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19015 that the mode size is in the set {1, 2, 4, 8, 16}.
19017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19018 Alan Hayward <alan.hayward@arm.com>
19019 David Sherwood <david.sherwood@arm.com>
19021 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19022 Use plus_constant instead of gen_rtx_PLUS.
19024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19025 Alan Hayward <alan.hayward@arm.com>
19026 David Sherwood <david.sherwood@arm.com>
19028 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19029 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19030 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19031 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19032 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19033 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19034 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19035 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19036 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19037 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19039 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19040 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19041 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19042 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19043 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19044 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19045 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19046 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19047 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19048 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19050 * expr.c (emit_move_resolve_push): Treat the input and result
19051 of PUSH_ROUNDING as a poly_int64.
19052 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19053 (emit_push_insn): Likewise.
19054 * lra-eliminations.c (mark_not_eliminable): Likewise.
19055 * recog.c (push_operand): Likewise.
19056 * reload1.c (elimination_effects): Likewise.
19057 * rtlanal.c (nonzero_bits1): Likewise.
19058 * calls.c (store_one_arg): Likewise. Require the padding to be
19059 known at compile time.
19061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19062 Alan Hayward <alan.hayward@arm.com>
19063 David Sherwood <david.sherwood@arm.com>
19065 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19066 Use plus_constant instead of gen_rtx_PLUS.
19068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19069 Alan Hayward <alan.hayward@arm.com>
19070 David Sherwood <david.sherwood@arm.com>
19072 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19073 rather than an int.
19075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19076 Alan Hayward <alan.hayward@arm.com>
19077 David Sherwood <david.sherwood@arm.com>
19079 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19080 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19081 via stack temporaries. Treat the mode size as polynomial too.
19083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19084 Alan Hayward <alan.hayward@arm.com>
19085 David Sherwood <david.sherwood@arm.com>
19087 * expr.c (expand_expr_real_2): When handling conversions involving
19088 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19089 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19090 as a poly_uint64 too.
19092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19093 Alan Hayward <alan.hayward@arm.com>
19094 David Sherwood <david.sherwood@arm.com>
19096 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19099 Alan Hayward <alan.hayward@arm.com>
19100 David Sherwood <david.sherwood@arm.com>
19102 * combine.c (can_change_dest_mode): Handle polynomial
19103 REGMODE_NATURAL_SIZE.
19104 * expmed.c (store_bit_field_1): Likewise.
19105 * expr.c (store_constructor): Likewise.
19106 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19107 and polynomial REGMODE_NATURAL_SIZE.
19108 (gen_lowpart_common): Likewise.
19109 * reginfo.c (record_subregs_of_mode): Likewise.
19110 * rtlanal.c (read_modify_subreg_p): Likewise.
19112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19113 Alan Hayward <alan.hayward@arm.com>
19114 David Sherwood <david.sherwood@arm.com>
19116 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19117 numbers of elements.
19119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19120 Alan Hayward <alan.hayward@arm.com>
19121 David Sherwood <david.sherwood@arm.com>
19123 * match.pd: Cope with polynomial numbers of vector elements.
19125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19126 Alan Hayward <alan.hayward@arm.com>
19127 David Sherwood <david.sherwood@arm.com>
19129 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19130 in a POINTER_PLUS_EXPR.
19132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19133 Alan Hayward <alan.hayward@arm.com>
19134 David Sherwood <david.sherwood@arm.com>
19136 * omp-simd-clone.c (simd_clone_subparts): New function.
19137 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19138 (ipa_simd_modify_function_body): 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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19145 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19146 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19147 (expand_vector_condition, vector_element): Likewise.
19148 (subparts_gt): New function.
19149 (get_compute_type): Use subparts_gt.
19150 (count_type_subparts): Delete.
19151 (expand_vector_operations_1): Use subparts_gt instead of
19152 count_type_subparts.
19154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19155 Alan Hayward <alan.hayward@arm.com>
19156 David Sherwood <david.sherwood@arm.com>
19158 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19159 (vect_compile_time_alias): ...this new function. Do the calculation
19160 on poly_ints rather than trees.
19161 (vect_prune_runtime_alias_test_list): Update call accordingly.
19163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19164 Alan Hayward <alan.hayward@arm.com>
19165 David Sherwood <david.sherwood@arm.com>
19167 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19169 (vect_schedule_slp_instance): Likewise.
19171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19172 Alan Hayward <alan.hayward@arm.com>
19173 David Sherwood <david.sherwood@arm.com>
19175 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19176 constant and extern definitions for variable-length vectors.
19177 (vect_get_constant_vectors): Note that the number of units
19178 is known to be constant.
19180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19181 Alan Hayward <alan.hayward@arm.com>
19182 David Sherwood <david.sherwood@arm.com>
19184 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19185 of units as polynomial. Choose between WIDE and NARROW based
19188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19189 Alan Hayward <alan.hayward@arm.com>
19190 David Sherwood <david.sherwood@arm.com>
19192 * tree-vect-stmts.c (simd_clone_subparts): New function.
19193 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19196 Alan Hayward <alan.hayward@arm.com>
19197 David Sherwood <david.sherwood@arm.com>
19199 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19200 vectors as polynomial. Use build_index_vector for
19201 IFN_GOMP_SIMD_LANE.
19203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19204 Alan Hayward <alan.hayward@arm.com>
19205 David Sherwood <david.sherwood@arm.com>
19207 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19208 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19209 for variable-length vectors.
19210 (vectorizable_mask_load_store): Treat the number of units as
19211 polynomial, asserting that it is constant if the condition has
19212 already been enforced.
19213 (vectorizable_store, vectorizable_load): Likewise.
19215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19216 Alan Hayward <alan.hayward@arm.com>
19217 David Sherwood <david.sherwood@arm.com>
19219 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19220 of units as polynomial. Punt if we can't tell at compile time
19221 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
19228 of units as polynomial. Punt on SLP inductions. Use an integer
19229 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19230 cast of such a series for variable-length floating-point
19233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19234 Alan Hayward <alan.hayward@arm.com>
19235 David Sherwood <david.sherwood@arm.com>
19237 * tree.h (build_index_vector): Declare.
19238 * tree.c (build_index_vector): New function.
19239 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19240 of units as polynomial, forcibly converting it to a constant if
19241 vectorizable_reduction has already enforced the condition.
19242 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19243 to create a {1,2,3,...} vector.
19244 (vectorizable_reduction): Treat the number of units as polynomial.
19245 Choose vectype_in based on the largest scalar element size rather
19246 than the smallest number of units. Enforce the restrictions
19249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19250 Alan Hayward <alan.hayward@arm.com>
19251 David Sherwood <david.sherwood@arm.com>
19253 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19254 number of units as polynomial.
19256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19257 Alan Hayward <alan.hayward@arm.com>
19258 David Sherwood <david.sherwood@arm.com>
19260 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19261 * target.def (autovectorize_vector_sizes): Return the vector sizes
19262 by pointer, using vector_sizes rather than a bitmask.
19263 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19264 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19265 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19267 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19268 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19269 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19270 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19271 * omp-general.c (omp_max_vf): Likewise.
19272 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19273 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19274 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19275 * tree-vect-slp.c (vect_slp_bb): Likewise.
19276 * doc/tm.texi: Regenerate.
19277 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19279 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19280 the vector size as a poly_uint64 rather than an unsigned int.
19281 (current_vector_size): Change from an unsigned int to a poly_uint64.
19282 (get_vectype_for_scalar_type): Update accordingly.
19283 * tree.h (build_truth_vector_type): Take the size and number of
19284 units as a poly_uint64 rather than an unsigned int.
19285 (build_vector_type): Add a temporary overload that takes
19286 the number of units as a poly_uint64 rather than an unsigned int.
19287 * tree.c (make_vector_type): Likewise.
19288 (build_truth_vector_type): Take the number of units as a poly_uint64
19289 rather than an unsigned int.
19291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19292 Alan Hayward <alan.hayward@arm.com>
19293 David Sherwood <david.sherwood@arm.com>
19295 * target.def (get_mask_mode): Take the number of units and length
19296 as poly_uint64s rather than unsigned ints.
19297 * targhooks.h (default_get_mask_mode): Update accordingly.
19298 * targhooks.c (default_get_mask_mode): Likewise.
19299 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19300 * doc/tm.texi: Regenerate.
19302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19303 Alan Hayward <alan.hayward@arm.com>
19304 David Sherwood <david.sherwood@arm.com>
19306 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19307 * omp-general.c (omp_max_vf): Likewise.
19308 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19309 (expand_omp_simd): Handle polynomial safelen.
19310 * omp-low.c (omplow_simd_context): Add a default constructor.
19311 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19312 (lower_rec_simd_input_clauses): Update accordingly.
19313 (lower_rec_input_clauses): Likewise.
19315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19316 Alan Hayward <alan.hayward@arm.com>
19317 David Sherwood <david.sherwood@arm.com>
19319 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19320 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19321 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19322 (vect_analyze_slp_cost): Likewise.
19323 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19324 (vect_model_load_cost): Likewise.
19326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19327 Alan Hayward <alan.hayward@arm.com>
19328 David Sherwood <david.sherwood@arm.com>
19330 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19331 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19332 from an unsigned int * to a poly_uint64_pod *.
19333 (calculate_unrolling_factor): New function.
19334 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19337 Alan Hayward <alan.hayward@arm.com>
19338 David Sherwood <david.sherwood@arm.com>
19340 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19341 from an unsigned int to a poly_uint64.
19342 (_loop_vec_info::slp_unrolling_factor): Likewise.
19343 (_loop_vec_info::vectorization_factor): Change from an int
19345 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19346 (vect_get_num_vectors): New function.
19347 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19348 (vect_get_num_copies): Use vect_get_num_vectors.
19349 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19350 to an unsigned int *.
19351 (vect_analyze_data_refs): Change min_vf from an int * to a
19353 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19354 than an unsigned HOST_WIDE_INT.
19355 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19356 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19357 to an unsigned int *.
19358 (vect_analyze_data_ref_dependences): Likewise.
19359 (vect_compute_data_ref_alignment): Handle polynomial vf.
19360 (vect_enhance_data_refs_alignment): Likewise.
19361 (vect_prune_runtime_alias_test_list): Likewise.
19362 (vect_shift_permute_load_chain): Likewise.
19363 (vect_supportable_dr_alignment): Likewise.
19364 (dependence_distance_ge_vf): Take the vectorization factor as a
19365 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19366 (vect_analyze_data_refs): Change min_vf from an int * to a
19368 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19369 vfm1 as a poly_uint64 rather than an int. Make the same change
19370 for the returned bound_scalar.
19371 (vect_gen_vector_loop_niters): Handle polynomial vf.
19372 (vect_do_peeling): Likewise. Update call to
19373 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19374 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19376 * tree-vect-loop.c (vect_determine_vectorization_factor)
19377 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19378 (vect_get_known_peeling_cost): Likewise.
19379 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19380 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19381 (vect_transform_loop): Likewise. Use the lowest possible VF when
19382 updating the upper bounds of the loop.
19383 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19384 rather than an int.
19385 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19386 polynomial unroll factors.
19387 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19388 (vect_make_slp_decision): Likewise.
19389 (vect_supported_load_permutation_p): Likewise, and polynomial
19391 (vect_analyze_slp_cost): Handle polynomial vf.
19392 (vect_slp_analyze_node_operations): Likewise.
19393 (vect_slp_analyze_bb_1): Likewise.
19394 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19395 than an unsigned HOST_WIDE_INT.
19396 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19397 (vectorizable_load): Handle polynomial vf.
19398 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19400 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19403 Alan Hayward <alan.hayward@arm.com>
19404 David Sherwood <david.sherwood@arm.com>
19406 * match.pd: Handle bit operations involving three constants
19407 and try to fold one pair.
19409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19411 * tree-vect-loop-manip.c: Include gimple-fold.h.
19412 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19413 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19414 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19415 Add a path that uses a step of VF instead of 1, but disable it
19417 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19418 and niters_no_overflow parameters. Update calls to
19419 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19420 Create a new SSA name if the latter choses to use a ste other
19421 than zero, and return it via niters_vector_mult_vf_var.
19422 * tree-vect-loop.c (vect_transform_loop): Update calls to
19423 vect_do_peeling, vect_gen_vector_loop_niters and
19424 slpeel_make_loop_iterate_ntimes.
19425 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19426 (vect_gen_vector_loop_niters): Update declarations after above changes.
19428 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19430 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19431 128-bit round to integer instructions.
19432 (ceil<mode>2): Likewise.
19433 (btrunc<mode>2): Likewise.
19434 (round<mode>2): Likewise.
19436 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19438 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19439 unaligned VSX load/store on P8/P9.
19440 (expand_block_clear): Allow the use of unaligned VSX
19441 load/store on P8/P9.
19443 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19445 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19447 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19448 swap associated with both a load and a store.
19450 2018-01-02 Andrew Waterman <andrew@sifive.com>
19452 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19453 * config/riscv/riscv.md (clear_cache): Use it.
19455 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19457 * web.c: Remove out-of-date comment.
19459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19461 * expr.c (fixup_args_size_notes): Check that any existing
19462 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19463 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19464 (emit_single_push_insn): ...here.
19466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19468 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19469 (const_vector_encoded_nelts): New function.
19470 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19471 (const_vector_int_elt, const_vector_elt): Declare.
19472 * emit-rtl.c (const_vector_int_elt_1): New function.
19473 (const_vector_elt): Likewise.
19474 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19475 of CONST_VECTOR_ELT.
19477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19479 * expr.c: Include rtx-vector-builder.h.
19480 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19481 directly on the tree encoding.
19482 (const_vector_from_tree): Likewise.
19483 * optabs.c: Include rtx-vector-builder.h.
19484 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19485 sequence of "u" values.
19486 * vec-perm-indices.c: Include rtx-vector-builder.h.
19487 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19488 directly on the vec_perm_indices encoding.
19490 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19492 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19493 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19494 * rtx-vector-builder.h: New file.
19495 * rtx-vector-builder.c: Likewise.
19496 * rtl.h (rtx_def::u2): Add a const_vector field.
19497 (CONST_VECTOR_NPATTERNS): New macro.
19498 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19499 (CONST_VECTOR_DUPLICATE_P): Likewise.
19500 (CONST_VECTOR_STEPPED_P): Likewise.
19501 (CONST_VECTOR_ENCODED_ELT): Likewise.
19502 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19503 (unwrap_const_vec_duplicate): Likewise.
19504 (const_vec_series_p): Check for a non-duplicated vector encoding.
19505 Say that the function only returns true for integer vectors.
19506 * emit-rtl.c: Include rtx-vector-builder.h.
19507 (gen_const_vec_duplicate_1): Delete.
19508 (gen_const_vector): Call gen_const_vec_duplicate instead of
19509 gen_const_vec_duplicate_1.
19510 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19511 (gen_const_vec_duplicate): Use rtx_vector_builder.
19512 (gen_const_vec_series): Likewise.
19513 (gen_rtx_CONST_VECTOR): Likewise.
19514 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19515 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19516 Build a new vector rather than modifying a CONST_VECTOR in-place.
19517 (handle_special_swappables): Update call accordingly.
19518 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19519 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19520 Build a new vector rather than modifying a CONST_VECTOR in-place.
19521 (handle_special_swappables): Update call accordingly.
19523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19525 * simplify-rtx.c (simplify_const_binary_operation): Use
19526 CONST_VECTOR_ELT instead of XVECEXP.
19528 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19530 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19531 the selector elements to be different from the data elements
19532 if the selector is a VECTOR_CST.
19533 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19534 ssizetype for the selector.
19536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19538 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19539 before testing each element individually.
19540 * tree-vect-generic.c (lower_vec_perm): Likewise.
19542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19544 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19545 * selftest-run-tests.c (selftest::run_tests): Call it.
19546 * vector-builder.h (vector_builder::operator ==): New function.
19547 (vector_builder::operator !=): Likewise.
19548 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19549 (vec_perm_indices::all_from_input_p): New function.
19550 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19551 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19552 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19553 instead of reading the VECTOR_CST directly. Detect whether both
19554 vector inputs are the same before constructing the vec_perm_indices,
19555 and update the number of inputs argument accordingly. Use the
19556 utility functions added above. Only construct sel2 if we need to.
19558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19560 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19561 the broadcast of the low byte.
19562 (expand_mult_highpart): Use an explicit encoding for the permutes.
19563 * optabs-query.c (can_mult_highpart_p): Likewise.
19564 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19565 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19566 (vectorizable_bswap): Likewise.
19567 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19568 explicit encoding for the power-of-2 permutes.
19569 (vect_permute_store_chain): Likewise.
19570 (vect_grouped_load_supported): Likewise.
19571 (vect_permute_load_chain): Likewise.
19573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19575 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19576 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19577 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19578 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19579 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19580 (vect_gen_perm_mask_any): Likewise.
19582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19584 * int-vector-builder.h: New file.
19585 * vec-perm-indices.h: Include int-vector-builder.h.
19586 (vec_perm_indices): Redefine as an int_vector_builder.
19587 (auto_vec_perm_indices): Delete.
19588 (vec_perm_builder): Redefine as a stand-alone class.
19589 (vec_perm_indices::vec_perm_indices): New function.
19590 (vec_perm_indices::clamp): Likewise.
19591 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19592 (vec_perm_indices::new_vector): New function.
19593 (vec_perm_indices::new_expanded_vector): Update for new
19594 vec_perm_indices class.
19595 (vec_perm_indices::rotate_inputs): New function.
19596 (vec_perm_indices::all_in_range_p): Operate directly on the
19597 encoded form, without computing elided elements.
19598 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19599 encoding. Update for new vec_perm_indices class.
19600 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19601 the given vec_perm_builder.
19602 (expand_vec_perm_var): Update vec_perm_builder constructor.
19603 (expand_mult_highpart): Use vec_perm_builder instead of
19604 auto_vec_perm_indices.
19605 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19606 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19607 or double series encoding as appropriate.
19608 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19609 vec_perm_indices instead of auto_vec_perm_indices.
19610 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19611 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19612 (vect_permute_store_chain): Likewise.
19613 (vect_grouped_load_supported): Likewise.
19614 (vect_permute_load_chain): Likewise.
19615 (vect_shift_permute_load_chain): Likewise.
19616 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19617 (vect_transform_slp_perm_load): Likewise.
19618 (vect_schedule_slp_instance): Likewise.
19619 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19620 (vectorizable_mask_load_store): Likewise.
19621 (vectorizable_bswap): Likewise.
19622 (vectorizable_store): Likewise.
19623 (vectorizable_load): Likewise.
19624 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19625 vec_perm_indices instead of auto_vec_perm_indices. Use
19626 tree_to_vec_perm_builder to read the vector from a tree.
19627 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19628 vec_perm_builder instead of a vec_perm_indices.
19629 (have_whole_vector_shift): Use vec_perm_builder and
19630 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19631 truncation to calc_vec_perm_mask_for_shift.
19632 (vect_create_epilog_for_reduction): Likewise.
19633 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19634 from auto_vec_perm_indices to vec_perm_indices.
19635 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19636 instead of changing individual elements.
19637 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19638 the vector in d.perm.
19639 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19640 from auto_vec_perm_indices to vec_perm_indices.
19641 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19642 instead of changing individual elements.
19643 (arm_vectorize_vec_perm_const): Use new_vector to install
19644 the vector in d.perm.
19645 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19646 Update vec_perm_builder constructor.
19647 (rs6000_expand_interleave): Likewise.
19648 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19649 (rs6000_expand_interleave): Likewise.
19651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19653 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19654 to qimode could truncate the indices.
19655 * optabs.c (expand_vec_perm_var): Likewise.
19657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19659 * Makefile.in (OBJS): Add vec-perm-indices.o.
19660 * vec-perm-indices.h: New file.
19661 * vec-perm-indices.c: Likewise.
19662 * target.h (vec_perm_indices): Replace with a forward class
19664 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19665 * optabs.h: Include vec-perm-indices.h.
19666 (expand_vec_perm): Delete.
19667 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19668 (expand_vec_perm_const): Declare.
19669 * target.def (vec_perm_const_ok): Replace with...
19670 (vec_perm_const): ...this new hook.
19671 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19672 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19673 * doc/tm.texi: Regenerate.
19674 * optabs.def (vec_perm_const): Delete.
19675 * doc/md.texi (vec_perm_const): Likewise.
19676 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19677 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19678 expand_vec_perm for constant permutation vectors. Assert that
19679 the mode of variable permutation vectors is the integer equivalent
19680 of the mode that is being permuted.
19681 * optabs-query.h (selector_fits_mode_p): Declare.
19682 * optabs-query.c: Include vec-perm-indices.h.
19683 (selector_fits_mode_p): New function.
19684 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19685 is defined, instead of checking whether the vec_perm_const_optab
19686 exists. Use targetm.vectorize.vec_perm_const instead of
19687 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19688 fit in the vector mode before using a variable permute.
19689 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19690 vec_perm_indices instead of an rtx.
19691 (expand_vec_perm): Replace with...
19692 (expand_vec_perm_const): ...this new function. Take the selector
19693 as a vec_perm_indices rather than an rtx. Also take the mode of
19694 the selector. Update call to shift_amt_for_vec_perm_mask.
19695 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19696 Use vec_perm_indices::new_expanded_vector to expand the original
19697 selector into bytes. Check whether the indices fit in the vector
19698 mode before using a variable permute.
19699 (expand_vec_perm_var): Make global.
19700 (expand_mult_highpart): Use expand_vec_perm_const.
19701 * fold-const.c: Includes vec-perm-indices.h.
19702 * tree-ssa-forwprop.c: Likewise.
19703 * tree-vect-data-refs.c: Likewise.
19704 * tree-vect-generic.c: Likewise.
19705 * tree-vect-loop.c: Likewise.
19706 * tree-vect-slp.c: Likewise.
19707 * tree-vect-stmts.c: Likewise.
19708 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19710 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19711 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19712 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19713 (aarch64_vectorize_vec_perm_const): ...this new function.
19714 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19715 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19716 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19717 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19718 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19719 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19720 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19722 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19723 check for NEON modes.
19724 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19725 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19726 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19727 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19729 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19730 the old VEC_PERM_CONST conditions.
19731 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19732 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19733 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19734 (ia64_vectorize_vec_perm_const_ok): Merge into...
19735 (ia64_vectorize_vec_perm_const): ...this new function.
19736 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19737 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19738 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19739 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19740 * config/mips/mips.c (mips_expand_vec_perm_const)
19741 (mips_vectorize_vec_perm_const_ok): Merge into...
19742 (mips_vectorize_vec_perm_const): ...this new function.
19743 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19744 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19745 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19746 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19747 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19748 (rs6000_expand_vec_perm_const): Delete.
19749 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19751 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19752 (altivec_expand_vec_perm_const_le): Take each operand individually.
19753 Operate on constant selectors rather than rtxes.
19754 (altivec_expand_vec_perm_const): Likewise. Update call to
19755 altivec_expand_vec_perm_const_le.
19756 (rs6000_expand_vec_perm_const): Delete.
19757 (rs6000_vectorize_vec_perm_const_ok): Delete.
19758 (rs6000_vectorize_vec_perm_const): New function.
19759 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19760 an element count and rtx array.
19761 (rs6000_expand_extract_even): Update call accordingly.
19762 (rs6000_expand_interleave): Likewise.
19763 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19764 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19765 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19766 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19767 (rs6000_expand_vec_perm_const): Delete.
19768 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19769 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19770 (altivec_expand_vec_perm_const_le): Take each operand individually.
19771 Operate on constant selectors rather than rtxes.
19772 (altivec_expand_vec_perm_const): Likewise. Update call to
19773 altivec_expand_vec_perm_const_le.
19774 (rs6000_expand_vec_perm_const): Delete.
19775 (rs6000_vectorize_vec_perm_const_ok): Delete.
19776 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19777 reference to the SPE evmerge intructions.
19778 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19779 an element count and rtx array.
19780 (rs6000_expand_extract_even): Update call accordingly.
19781 (rs6000_expand_interleave): Likewise.
19782 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19783 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19785 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19789 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19790 vector mode and that that mode matches the mode of the data
19792 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19793 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19794 directly using expand_vec_perm_1 when forcing selectors into
19796 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19800 * optabs-query.h (can_vec_perm_p): Delete.
19801 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19802 * optabs-query.c (can_vec_perm_p): Split into...
19803 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19804 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19805 particular selector is valid.
19806 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19807 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19808 (vect_grouped_load_supported): Likewise.
19809 (vect_shift_permute_load_chain): Likewise.
19810 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19811 (vect_transform_slp_perm_load): Likewise.
19812 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19813 (vectorizable_bswap): Likewise.
19814 (vect_gen_perm_mask_checked): Likewise.
19815 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19816 implementations of variable permutation vectors into account
19817 when deciding which selector to use.
19818 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19819 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19820 with a false third argument.
19821 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19822 to test whether the constant selector is valid and can_vec_perm_var_p
19823 to test whether a variable selector is valid.
19825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19827 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19828 * optabs-query.c (can_vec_perm_p): Likewise.
19829 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19830 instead of vec_perm_indices.
19831 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19832 (vect_gen_perm_mask_checked): Likewise,
19833 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19834 (vect_gen_perm_mask_checked): Likewise,
19836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19838 * optabs-query.h (qimode_for_vec_perm): Declare.
19839 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19840 (qimode_for_vec_perm): ...this new function.
19841 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19843 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19845 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19846 does not have a conditional at the top.
19848 2018-01-02 Richard Biener <rguenther@suse.de>
19850 * ipa-inline.c (big_speedup_p): Fix expression.
19852 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19855 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19858 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19862 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19863 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19864 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19865 cond_taken_branch_cost 3->4.
19867 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19869 PR tree-optimization/83581
19870 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19871 TODO_cleanup_cfg if any changes have been made.
19873 PR middle-end/83608
19874 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19875 convert_modes if target mode has the right side, but different mode
19878 PR middle-end/83609
19879 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19880 last argument when extracting from CONCAT. If either from_real or
19881 from_imag is NULL, use expansion through memory. If result is not
19882 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19883 the parts directly to inner mode, if even that fails, use expansion
19886 PR middle-end/83623
19887 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19888 check for bswap in mode rather than HImode and use that in expand_unop
19891 Copyright (C) 2018 Free Software Foundation, Inc.
19893 Copying and distribution of this file, with or without modification,
19894 are permitted in any medium without royalty provided the copyright
19895 notice and this notice are preserved.