1 2018-07-13 Richard Biener <rguenther@suse.de>
4 * match.pd (addr1 - addr2): Allow either of the operand to
7 2018-07-13 Tom de Vries <tdevries@suse.de>
9 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
14 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
15 arrays instead of numbered variables.
17 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
19 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
20 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
22 2018-07-13 Richard Biener <rguenther@suse.de>
25 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
26 instead of get_context_die.
28 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
29 Richard Biener <rguenther@suse.de>
32 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
33 that the loop latch destination where phi is defined.
35 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
37 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
38 (riscv_expand_epilogue): Add assertion to check interrupt mode.
39 (riscv_set_current_function): Extract getting interrupt type to new
41 (riscv_get_interrupt_type): New function.
42 (riscv_merge_decl_attributes): New function, checking interrupt type is
44 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
46 2018-07-12 Paul Koning <ni1d@arrl.net>
48 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
51 2018-07-12 Paul Koning <ni1d@arrl.net>
53 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
54 zero reference, add doloop_end instead.
55 * doc/md.texi (decrement_and_branch_until_zero): Remove.
56 (Looping patterns): Remove decrement_and_branch_until_zero. Add
57 detail for doloop_end.
59 2018-07-12 Martin Sebor <msebor@redhat.com>
62 * attribs.c (decl_attributes): Reject conflicting attributes before
63 calling attribute handlers.
65 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
67 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
69 (gcc::dump_manager::get_dump_file_name): likewise.
70 (dump_begin): Likewise.
71 * dumpfile.h (dump_begin): Update prototype.
72 (gcc::dump_manager::get_dump_file_name,
73 gcc::dump_manager::get_dump_file_name): Update prototype.
75 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
77 * internal-fn.h (vectorizable_internal_fn_p): New function.
78 * tree-vect-slp.c (compatible_calls_p): Likewise.
79 (vect_build_slp_tree_1): Remove nops argument. Handle calls
80 to internal functions.
81 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
83 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
85 * fold-const.h (inverse_conditions_p): Declare.
86 * fold-const.c (inverse_conditions_p): New function.
87 * match.pd: Use inverse_conditions_p. Add folds of view_converts
88 that test the inverse condition of a conditional internal function.
89 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
90 * internal-fn.c (internal_fn_mask_index): Handle conditional
92 (vectorized_internal_fn_supported_p): New function.
93 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
94 (any_pred_load_store): Replace with...
95 (need_to_predicate): ...this new variable.
96 (redundant_ssa_names): New variable.
97 (ifcvt_can_use_mask_load_store): Move initial checks to...
98 (ifcvt_can_predicate): ...this new function. Handle tree codes
99 for which a conditional internal function exists.
100 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
101 instead of ifcvt_can_use_mask_load_store. Update after variable
103 (predicate_load_or_store): New function, split out from
104 predicate_mem_writes.
105 (check_redundant_cond_expr): New function.
106 (value_available_p): Likewise.
107 (predicate_rhs_code): Likewise.
108 (predicate_mem_writes): Rename to...
109 (predicate_statements): ...this. Use predicate_load_or_store
110 and predicate_rhs_code.
111 (combine_blocks, tree_if_conversion): Update after above name changes.
112 (ifcvt_local_dce): Handle redundant_ssa_names.
113 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
114 general conditional functions.
115 * tree-vect-stmts.c (vectorizable_call): Likewise.
117 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
118 Alan Hayward <alan.hayward@arm.com>
119 David Sherwood <david.sherwood@arm.com>
121 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
122 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
123 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
124 plus and minus and convert them into IFN_COND_FMA-based sequences.
125 (convert_mult_to_fma): Handle conditional plus and minus.
127 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
129 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
130 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
131 (cond_fnms_optab): New optabs.
132 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
134 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
135 * internal-fn.h (get_conditional_internal_fn): Declare.
136 (get_unconditional_internal_fn): Likewise.
137 * internal-fn.c (cond_ternary_direct): New macro.
138 (expand_cond_ternary_optab_fn): Likewise.
139 (direct_cond_ternary_optab_supported_p): Likewise.
140 (FOR_EACH_COND_FN_PAIR): Likewise.
141 (get_conditional_internal_fn): New function.
142 (get_unconditional_internal_fn): Likewise.
143 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
144 (gimple_match_op::gimple_match_op): Add a new overload for 5
146 (gimple_match_op::set_op): Likewise.
147 (gimple_resimplify5): Declare.
148 * genmatch.c (decision_tree::gen): Generate simplifications for
150 * gimple-match-head.c (gimple_simplify): Define an overload for
151 5 operands. Handle calls with 5 arguments in the top-level overload.
152 (convert_conditional_op): Handle conversions from unconditional
153 internal functions to conditional ones.
154 (gimple_resimplify5): New function.
155 (build_call_internal): Pass a fifth operand.
156 (maybe_push_res_to_seq): Likewise.
157 (try_conditional_simplification): Try converting conditional
158 internal functions to unconditional internal functions.
159 Handle 3-operand unconditional forms.
160 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
161 Define ternary equivalents of the current rules for binary conditional
163 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
165 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
166 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
167 (optab): Handle them.
168 (SVE_COND_FP_TERNARY): New int iterator.
169 (sve_fmla_op, sve_fmad_op): New int attributes.
170 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
171 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
172 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
174 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
176 * target.def (preferred_else_value): New target hook.
177 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
178 * doc/tm.texi: Regenerate.
179 * targhooks.h (default_preferred_else_value): Declare.
180 * targhooks.c (default_preferred_else_value): New function.
181 * internal-fn.h (conditional_internal_fn_code): Declare.
182 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
183 (get_conditional_internal_fn): Use it.
184 (conditional_internal_fn_code): New function.
185 * gimple-match.h (gimple_match_cond): New struct.
186 (gimple_match_op): Add a cond member function.
187 (gimple_match_op::gimple_match_op): Update all forms to take a
189 * genmatch.c (expr::gen_transform): Use the same condition as res_op
190 for the suboperation, but don't specify a particular else_value.
191 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
192 (visit_nary_op, visit_reference_op_load): Pass
193 gimple_match_cond::UNCOND to the gimple_match_op constructor.
194 * gimple-match-head.c: Include tree-eh.h
195 (convert_conditional_op): New function.
196 (maybe_resimplify_conditional_op): Likewise.
197 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
198 (gimple_resimplify2): Likewise.
199 (gimple_resimplify3): Likewise.
200 (gimple_resimplify4): Likewise.
201 (maybe_push_res_to_seq): Return null for conditional operations.
202 (try_conditional_simplification): New function.
203 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
205 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
207 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
209 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
211 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
213 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
215 (hash_tree): Do not hash DECL_FCONTEXT
216 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
217 Do not stream DECL_FCONTEXT.
218 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
219 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
221 2018-07-12 Richard Biener <rguenther@suse.de>
224 * dwarf2out.c (gen_block_die): Only output blocks when they have
225 at least one !DECL_IGNORED_P variable.
227 2018-07-12 Richard Biener <rguenther@suse.de>
230 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
231 Remove -mieee-fp handling.
233 2018-07-12 Richard Biener <rguenther@suse.de>
235 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
236 left-over from last patch.
238 2018-07-12 Jakub Jelinek <jakub@redhat.com>
240 PR tree-optimization/86492
241 * gimple-ssa-store-merging.c
242 (imm_store_chain_info::coalesce_immediate_stores): Call
243 check_no_overlap even for the merge_overlapping case. Formatting fix.
245 2018-07-12 Richard Biener <rguenther@suse.de>
248 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
249 move possibly trapping operations into the conditional.
251 2018-07-12 Richard Biener <rguenther@suse.de>
253 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
254 (vn_lookup_simplify_result): Remove recursion limit applied
256 (vn_nary_build_or_lookup_1): Adjust.
257 (try_to_simplify): Likewise.
258 * gimple-match-head.c (gimple_resimplify1): Instead apply one
260 (gimple_resimplify2): Likewise.
261 (gimple_resimplify3): Likewise.
262 (gimple_resimplify4): Likewise.
264 2018-07-11 Jakub Jelinek <jakub@redhat.com>
266 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
267 Use __mmask64 type instead of __mmask8 for __M argument.
268 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
269 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
271 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
272 __mmask16 for __M argument.
273 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
274 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
275 to __mmask16 instead of __mmask8.
276 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
277 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
278 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
279 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
280 instead of __mmask16 for __U argument.
281 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
282 __mmask16 instead of __mmask8 for __U argument.
283 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
285 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
287 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
289 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
291 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
293 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
295 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
297 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
299 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
301 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
303 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
304 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
305 return type as well as __M argument type and all casts from __mmask8
307 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
308 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
309 return type as well as __M argument type and all casts from __mmask8
311 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
312 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
313 return type as well as __M argument type and all casts from __mmask8
315 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
316 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
317 return type as well as __M argument type and all casts from __mmask8
319 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
320 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
323 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
325 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
326 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
329 2018-07-11 Paul Koning <ni1d@arrl.net>
331 * doc/md.texi (define_subst): Document how multiple occurrences of
332 the same argument in the replacement pattern are handled.
334 2018-07-11 Paul Koning <ni1d@arrl.net>
336 * doc/extend.texi (Common Variable Attributes): Move "mode" into
338 (Common Type Attributes): Add "mode" attribute.
340 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
342 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
343 stream DECL_ORIGINAL_TYPE.
344 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
345 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
346 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
347 Do not walk original type.
348 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
350 (write_ts_decl_non_common_tree_pointers): Do not stream
352 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
353 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
355 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
357 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
358 thread twice from the same starting edge.
360 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
362 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
363 * gimple.c (gimple_call_nonnull_result_p): ...here...
364 (gimple_call_nonnull_arg): ...and here.
365 * gimple.h (gimple_call_nonnull_result_p): New.
366 (gimple_call_nonnull_arg): New.
368 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
370 * config/arm/arm-cpus.in: Move information from fpu field of each
371 cpu definition to the isa field.
372 * config/arm/parsecpu.awk (fpu): Delete match rule.
373 (gen_comm_data): Don't add bits from the CPU's FPU entry.
375 2018-07-11 Richard Biener <rguenther@suse.de>
378 * dwarf2out.c (init_sections_and_labels): Use
379 output_asm_line_debug_info consistently.
380 (dwarf2out_early_finish): Likewise.
381 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
384 2018-07-11 Richard Biener <rguenther@suse.de>
386 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
387 Rework father_bb setting in a way to avoid propagating constants
388 multiple times on a loop body.
390 2018-07-10 Mark Wielaard <mark@klomp.org>
393 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
395 2018-07-10 Richard Biener <rguenther@suse.de>
397 * hash-map.h (hash_map::iterator::operator*): Return
398 references to key and value.
400 2018-07-10 Jakub Jelinek <jakub@redhat.com>
403 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
404 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
405 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
406 (gimplify_omp_for): For composite loops, move outer
407 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
408 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
409 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
410 TREE_LIST for both the original class iterator and the "last" helper
411 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
412 loop, remember has_decl_expr from outer composite loops for the
413 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
415 2018-07-09 Martin Sebor <msebor@redhat.com>
419 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
420 accessing implicitly initialized array elements.
421 * expr.c (string_constant): Handle string initializers of
422 character arrays within aggregates.
423 * gimple-fold.c (fold_array_ctor_reference): Add argument.
424 Store element offset. As a special case, handle zero size.
425 (fold_nonarray_ctor_reference): Same.
426 (fold_ctor_reference): Add argument. Store subobject offset.
427 * gimple-fold.h (fold_ctor_reference): Add argument.
429 2018-07-09 Paul Koning <ni1d@arrl.net>
431 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
432 (pdp11_insn_cost): New function.
433 (pdp11_md_asm_adjust): New function.
434 (TARGET_INVALID_WITHIN_DOLOOP): Define.
435 (pdp11_rtx_costs): Update to match machine better.
436 (output_addr_const_pdp11): Correct format mismatch warnings.
437 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
438 * config/pdp11/pdp11.md: General change to add base_cost and/or
439 length attributes for use by new pdp11_insn_cost function.
440 (MIN_BRANCH): Correct definition.
442 (doloop_end): Use standard pattern name for looping pattern.
443 (doloop_end_nocc): New.
444 (movsf): Add another constraint alternative.
445 (zero_extendqihi2): Add constraint alternatives for not in place
447 (zero_extendhisi2): Remove.
448 (shift patterns): Add CC handling variants.
452 (define_peephole2): New peephole to recognize mov that sets CC for
455 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
457 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
458 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
461 2018-07-09 Jakub Jelinek <jakub@redhat.com>
464 * real.c (real_nextafter): Return true if result is denormal.
466 2018-07-09 Martin Liska <mliska@suse.cz>
468 * common.opt: Add back wrongly removed attribute.
470 2018-07-09 Richard Biener <rguenther@suse.de>
473 * dwarf2out.c (gen_block_die): For an early generated DIE
474 always output high/low PC attributes.
476 2018-07-09 Tom de Vries <tdevries@suse.de>
478 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
479 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
480 onto VAR_DECL with abstract origin.
482 2018-07-07 Jim Wilson <jimw@sifive.com>
484 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
486 2018-07-07 Tom de Vries <tdevries@suse.de>
488 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
490 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
492 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
494 (combine_bound): Use wide-int overflow calculation instead of
496 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
498 * fold-const.c (int_const_binop_2): Same.
499 (extract_muldiv_1): Same.
500 (fold_div_compare): Same.
501 (fold_abs_const): Same.
503 * poly-int.h (add): Same.
507 * predict.c (predict_iv_comparison): Same.
508 * profile-count.c (slow_safe_scale_64bit): Same.
509 * simplify-rtx.c (simplify_const_binary_operation): Same.
510 * tree-chrec.c (tree_fold_binomial): Same.
511 * tree-data-ref.c (split_constant_offset_1): Same.
512 * tree-if-conv.c (idx_within_array_bound): Same.
513 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
514 * tree-ssa-phiopt.c (minmax_replacement): Same.
515 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
516 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
517 * vr-values.c (vr_values::adjust_range_with_scev): Same.
518 * wide-int.cc (wi::add_large): Same.
519 (wi::mul_internal): Same.
520 (wi::sub_large): Same.
521 (wi::divmod_internal): Same.
522 * wide-int.h: Change overflow type to overflow_type for neg, add,
523 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
524 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
525 mul_internal, divmod_internal.
526 (overflow_type): New enum.
527 (accumulate_overflow): New.
529 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
531 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
532 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
534 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
536 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
537 argument is checked for zero before entering loop, avoid checking again.
539 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
541 * gimplify.h (generic_expr_could_trap_p): Set as global function.
542 * gimplify.h (generic_expr_could_trap_p): Likwise.
543 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
545 2018-07-06 Jakub Jelinek <jakub@redhat.com>
547 PR tree-optimization/86401
548 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
549 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
550 (fold_bit_and_mask): ... here. New helper function for match.pd.
551 * fold-const.h (fold_bit_and_mask): Declare.
552 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
554 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
557 * target.def (translate_mode_attribute): New hook.
558 * targhooks.h (default_translate_mode_attribute): Declare.
559 * targhooks.c (default_translate_mode_attribute): New function.
560 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
561 * doc/tm.texi: Regenerate.
562 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
563 (rs6000_translate_mode_attribute): New function.
565 2018-07-06 Paul Koning <ni1d@arrl.net>
567 * doc/md.texi (define_split): Document DONE and FAIL.
568 (define_peephole2): Ditto.
570 2018-07-05 Jeff Law <law@redhat.com>
572 PR tree-optimization/86010
573 * tree-ssa-dse.c (compute_trims): More aggressively trim at
574 both the head and tail of mem* and str* calls.
576 2018-07-05 Jim Wilson <jimw@sifive.com>
578 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
580 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
582 * config/aarch64/aarch64-simd.md: correct flags text for
583 MIN_EXPR replacement.
585 2018-07-05 James Clarke <jrtc27@jrtc27.com>
587 * configure: Regenerated.
589 2018-07-05 Carl Love <cel@us.ibm.com>
591 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
592 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
593 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
594 VSX_BUILTIN_DOUBLEL_V4SF.
596 2018-07-05 Martin Sebor <msebor@redhat.com>
599 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
600 than its domain to compute its the upper bound of a char array.
602 2018-07-05 Nathan Sidwell <nathan@acm.org>
604 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
605 * doc/cpp.texi: Update comment.
606 * doc/tm.texi: Rebuilt.
607 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
608 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
609 * doc/extend.texi (Backwards Compatibility): Clarify it is system
610 headers affected by extern "C".
611 * system.h: Poison NO_IMPLICIT_EXTERN_C.
612 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
613 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
614 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
615 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
616 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
617 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
618 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
619 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
620 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
621 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
622 config/sparc/sp64-elf.h, config/spu/spu.h,
623 config/stormy16/stormy16.h, config/v850/v850.h,
624 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
625 define NO_IMPLICIT_EXTERN_C.
626 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
628 2018-07-05 Tamar Christina <tamar.christina@arm.com>
631 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
632 instead of GET_MODE_SIZE when comparing Units.
634 2018-07-05 Tamar Christina <tamar.christina@arm.com>
637 * rtlanal.c (set_noop_p): Constrain on mode change,
638 include hard-reg-set.h
640 2018-07-05 Tamar Christina <tamar.christina@arm.com>
642 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
644 2018-07-05 Jakub Jelinek <jakub@redhat.com>
647 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
650 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
652 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
654 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
657 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
659 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
661 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
664 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
666 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
668 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
671 2018-07-04 Martin Liska <mliska@suse.cz>
673 * tree-switch-conversion.c: Define
674 max_ratio_for_speed and max_ratio_for_size constants.
676 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
677 Martin Liska <mliska@suse.cz>
682 * common.opt: Rename align options with 'str_' prefix.
683 * common/config/i386/i386-common.c (set_malign_value): New
685 (ix86_handle_option): Use it to set -falign-* options/
686 * config/aarch64/aarch64-protos.h (struct tune_params): Change
687 type from int to string.
688 * config/aarch64/aarch64.c: Update default values from int
690 * config/alpha/alpha.c (alpha_override_options_after_change):
692 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
693 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
694 max skip conditionally.
695 * config/i386/freebsd.h (SUBALIGN_LOG): New.
696 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
697 max skip conditionally.
698 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
699 max skip conditionally.
700 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
701 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
702 max skip conditionally.
703 * config/i386/i386.c (struct ptt): Change type from int to
705 (ix86_default_align): Set default values.
706 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
707 max skip conditionally.
708 * config/i386/iamcu.h (SUBALIGN_LOG): New.
709 (ASM_OUTPUT_MAX_SKIP_ALIGN):
710 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
711 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
712 max skip conditionally.
713 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
714 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
715 * config/i386/x86-64.h (SUBALIGN_LOG): New.
716 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
717 max skip conditionally.
718 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
719 * config/ia64/ia64.c (ia64_option_override): Set default values
720 for alignment options.
721 * config/m68k/m68k.c: Handle new str_align_* options.
722 * config/mips/mips.c (mips_set_compression_mode): Change
724 (mips_option_override): Set default values for options.
725 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
727 * config/rs6000/rs6000.c (rs6000_option_override_internal):
729 * config/rx/rx.c (rx_option_override): Likewise.
730 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
731 (LABEL_ALIGN): Use align_labels_log.
732 (LOOP_ALIGN): Use align_loops_align.
733 * config/s390/s390.c (s390_asm_output_function_label): Use new
735 * config/sh/sh.c (sh_override_options_after_change):
736 Change type of constants.
737 * config/spu/spu.c (spu_sched_init): Likewise.
738 * config/sparc/sparc.c (sparc_option_override): Set default
740 * config/visium/visium.c (visium_option_override): Likewise.
741 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
742 emit p2align format with last argument if it's not needed.
743 * doc/invoke.texi: Document extended format of -falign-*.
744 * final.c: Use align_labels alignment.
745 * flags.h (struct target_flag_state): Change type to use
747 (struct align_flags_tuple): New.
748 (struct align_flags): Likewise.
749 (align_loops_log): Redefine macro to use new types.
750 (align_loops_max_skip): Redefine macro to use new types.
751 (align_jumps_log): Redefine macro to use new types.
752 (align_jumps_max_skip): Redefine macro to use new types.
753 (align_labels_log): Redefine macro to use new types.
754 (align_labels_max_skip): Redefine macro to use new types.
755 (align_functions_log): Redefine macro to use new types.
756 (align_loops): Redefine macro to use new types.
757 (align_jumps): Redefine macro to use new types.
758 (align_labels): Redefine macro to use new types.
759 (align_functions): Redefine macro to use new types.
760 (align_functions_max_skip): Redefine macro to use new types.
761 (align_loops_value): New macro.
762 (align_jumps_value): New macro.
763 (align_labels_value): New macro.
764 (align_functions_value): New macro.
765 * function.c (invoke_set_current_function_hook): Propagate
766 alignment values from flags to global variables default in
768 * ipa-icf.c (sem_function::equals_wpa): Use
769 cl_optimization_option_eq instead of memcmp.
770 * lto-streamer.h (cl_optimization_stream_out): Support streaming
772 (cl_optimization_stream_in): Likewise.
773 * optc-save-gen.awk: Support strings in cl_optimization.
774 * opth-gen.awk: Likewise.
775 * opts.c (finish_options): Remove error checking of invalid
777 (MAX_CODE_ALIGN): Remove.
778 (MAX_CODE_ALIGN_VALUE): Likewise.
779 (parse_and_check_align_values): New function.
780 (check_alignment_argument): Likewise.
781 (common_handle_option): Use check_alignment_argument.
782 * opts.h (parse_and_check_align_values): Declare.
783 * toplev.c (init_alignments): Remove.
784 (read_log_maxskip): New.
785 (parse_N_M): Likewise.
786 (parse_alignment_opts): Likewise.
787 (backend_init_target): Remove usage of init_alignments.
788 * toplev.h (parse_alignment_opts): Declare.
789 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
791 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
792 * tree.c (cl_option_hasher::equal): New.
793 * varasm.c: Use new global macros.
795 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
797 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
798 Use a simpler align directive also if MAXSKIP = ALIGN-1.
799 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
800 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
801 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
802 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
803 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
804 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
805 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
806 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
807 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
808 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
809 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
811 2018-07-04 Martin Liska <mliska@suse.cz>
812 Jonathan Wakely <jwakely@redhat.com>
814 * coverage.c: Use correct type.
815 * doc/invoke.texi: Language correction.
817 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
820 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
821 ENDBRANCH for non-tail call which may return via indirect branch.
822 * doc/extend.texi: Document indirect_return attribute.
824 2018-07-03 Martin Sebor <msebor@redhat.com>
826 PR tree-optimization/86274
827 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
829 (format_floating): Correct handling of infinities and NaNs.
831 2018-07-03 Martin Sebor <msebor@redhat.com>
833 * print-tree.c (print_real_cst): New function.
834 (print_node_brief): Call it.
837 2018-07-03 Jeff Law <law@redhat.com>
839 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
840 into a single pattern.
842 * config/h8300/h8300.md (ors code_iterator): New.
843 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
844 a single pattern and single splitter.
845 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
846 (iorqi3_1, xorqi3_1): Likewise.
847 (iorqi3, xorqi3 expanders): Similarly.
849 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
850 (movmd_internal) into a single pattern using the P mode iterator.
851 (movmd splitters): Similarly.
852 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
853 (movsd splitters): Similarly.
855 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
856 ADDB, ADDW and ADDL into a single ADD attribute which selects the
857 right table based on the size of the operand.
858 * config/h8300/h8300.md (length_table): Corresponding changes. All
859 references to "addb", "addw" and "addl" changed to "add".
860 (btst patterns): Merge two variants into a single pattern.
861 (tstqi, tsthi): Likewise.
862 (addhi3_incdec, addsi3_incdec): Likewise.
863 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
864 (mulhi3, mulsi3): Likewise.
865 (udivhi3, udivsi3): Likewise.
866 (divhi3, divsi3): Likewise.
867 (andorqi3, andorhi3, andorsi3): Likewise.
869 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
872 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
873 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
875 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
877 PR tree-optimization/85694
878 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
879 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
881 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
882 (<u>avg<mode>3_ceil): New patterns.
884 2018-07-03 David Malcolm <dmalcolm@redhat.com>
886 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
887 scan-tree-dump directive.
888 * gcc.dg/vect/slp-perm-2.c: Likewise.
889 * gcc.dg/vect/slp-perm-3.c: Likewise.
890 * gcc.dg/vect/slp-perm-5.c: Likewise.
891 * gcc.dg/vect/slp-perm-6.c: Likewise.
892 * gcc.dg/vect/slp-perm-7.c: Likewise.
893 * gcc.dg/vect/slp-perm-8.c: Likewise.
895 2018-07-03 Marek Polacek <polacek@redhat.com>
898 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
900 2018-07-03 Richard Biener <rguenther@suse.de>
903 * tree-ssa-structalias.c (find_func_clobbers): Properly
904 handle indirect calls.
906 2018-07-03 Jeff Law <law@redhat.com>
908 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
909 (shifts): New code iterator.
910 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
911 expander. Fix HImode handling on H8/SX.
912 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
913 (subqi3, subhi3, subsi3 expanders): Likewise.
914 (andqi3, andhi3, andsi3 expanders): Likewise.
915 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
916 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
917 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
918 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
919 (zero_extendqihi2, zero_extendqisi2): Likewise.
920 (extendqihi2, extendqisi2): Likewise.
921 (rotlqi3, rotlhi3, rotlsi3): Likewise.
922 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
923 (rotlqi3_1, rotlhi3_1): Likewise.
924 (logicalhi3_sn, logicalsi3_sn): Likewise.
925 (logicalhi3, logicalsi3): Likewise.
927 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
929 * tree-vect-patterns.c (vect_recog_rotate_pattern)
930 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
931 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
932 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
933 type to append_pattern_def_seq instead of creating a stmt_vec_info
935 (build_mask_conversion): Likewise. Remove vinfo argument.
936 (vect_add_conversion_to_patterm): Likewise, renaming to...
937 (vect_add_conversion_to_pattern): ...this.
938 (vect_recog_mask_conversion_pattern): Update call to
939 build_mask_conversion. Pass the vector type to
940 append_pattern_def_seq here too.
941 (vect_recog_gather_scatter_pattern): Update call to
942 vect_add_conversion_to_pattern.
944 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
946 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
947 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
948 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
949 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
950 STMT_VINFO_PATTERN_DEF_SEQ to null here.
951 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
952 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
953 append_pattern_def_seq instead of new_pattern_def_seq.
954 (vect_recog_divmod_pattern): Do both of the above.
955 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
958 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
960 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
961 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
962 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
963 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
964 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
965 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
966 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
967 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
968 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
969 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
970 parameter with a single stmt_vec_info.
971 (vect_recog_func_ptr): Likewise.
972 (vect_recog_gather_scatter_pattern): Likewise, folding in...
973 (vect_try_gather_scatter_pattern): ...this.
974 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
975 the stmt_vec_info of the statement to be matched. Don't clear
976 STMT_VINFO_RELATED_STMT.
977 (vect_pattern_recog): Update call accordingly.
979 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
981 PR tree-optimization/85694
982 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
983 (uavgM3_ceil): Document new optabs.
984 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
985 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
987 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
988 (savg_ceil_optab): New optabs.
989 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
990 (vect_vect_recog_func_ptrs): Add it.
991 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
992 constant directly from the associated lhs.
994 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
996 * tree-vect-patterns.c (vect_split_statement): New function.
997 (vect_convert_input): Use it to try to split an existing cast.
999 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1001 * poly-int.h (print_hex): New function.
1002 * dumpfile.h (dump_dec, dump_hex): Declare.
1003 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1004 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1005 min_input_precision, operation_precision and operation_sign.
1006 * tree-vect-patterns.c (vect_get_range_info): New function.
1007 (vect_same_loop_or_bb_p, vect_single_imm_use)
1008 (vect_operation_fits_smaller_type): Delete.
1009 (vect_look_through_possible_promotion): Add an optional
1010 single_use_p parameter.
1011 (vect_recog_over_widening_pattern): Rewrite to use new
1012 stmt_vec_info infomration. Handle one operation at a time.
1013 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1014 (vect_truncatable_operation_p, vect_set_operation_type)
1015 (vect_set_min_input_precision): New functions.
1016 (vect_determine_min_output_precision_1): Likewise.
1017 (vect_determine_min_output_precision): Likewise.
1018 (vect_determine_precisions_from_range): Likewise.
1019 (vect_determine_precisions_from_users): Likewise.
1020 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1021 (vect_vect_recog_func_ptrs): Put over_widening first.
1023 (vect_pattern_recog): Call vect_determine_precisions.
1025 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1027 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1028 statements that have been replaced by further pattern statements.
1029 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1031 2018-07-03 Richard Biener <rguenther@suse.de>
1033 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1034 always set *dt. Dump vectype in vectype overload.
1035 * dumpfile.h (dump_gimple_expr): New function.
1036 (dump_gimple_expr_loc): Likewise.
1037 * dumpfile.c (dump_gimple_expr): New function.
1038 (dump_gimple_expr_loc): Likewise.
1040 2018-07-02 Jeff Law <law@redhat.com>
1042 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1043 the H8/300, H8/300H and H8/S variants into a single pattern.
1044 (movhi_h8300, movqi_h8300hs): Similarly.
1045 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1046 (QHI mode iterator): New.
1048 * config/h8300/h8300.md: Remove trailing whitespace.
1050 2018-07-02 Jim Wilson <jimw@sifive.com>
1052 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1053 instead of emit_insn for interrupt returns.
1054 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1055 (riscv_sret, riscv_uret): Likewise.
1057 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1059 * pretty-print.c (selftest::test_pp_format): Move save and restore
1060 of quotes to class auto_fix_quotes, and add an instance.
1061 * selftest.c: Include "intl.h".
1062 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1063 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1064 * selftest.h (selftest::auto_fix_quotes): New class.
1066 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1068 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1069 (aarch64_sve_prepare_conditional_op): Remove.
1070 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1071 Allow aarch64_simd_reg_or_zero as select operand; remove
1072 the aarch64_sve_prepare_conditional_op call.
1073 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1074 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1075 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1076 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1077 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1078 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1079 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1080 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1081 and a splitters to match all of the *_any patterns.
1082 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1084 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1085 (SVE_COND_FP_BINARY_REV): Remove.
1086 (sve_int_op_rev, sve_fp_op_rev): New.
1087 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1088 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1089 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1090 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1091 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1092 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1093 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1094 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1095 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1097 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1098 Remove match_dup 1 from the inner unspec.
1099 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1101 * config/aarch64/aarch64.md (movprfx): New attr.
1102 (length): Default movprfx to 8.
1103 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1104 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1105 (*<su>mul<SVE_I>3_highpart): Likewise.
1106 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1107 (*v<ASHIFT><SVE_I>3): Likewise.
1108 (*<su><MAXMIN><SVE_I>3): Likewise.
1109 (*<su><MAXMIN><SVE_F>3): Likewise.
1110 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1111 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1112 (*div<SVE_F>4): Likewise.
1114 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1116 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1119 2018-07-02 Richard Biener <rguenther@suse.de>
1121 PR tree-optimization/86363
1122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1123 memset argument refers to a non-variable address.
1125 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1127 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1128 {PLUS,MINUS}_EXPR code to...
1129 (adjust_symbolic_bound): ...here,
1130 (combine_bound): ...here,
1131 (set_value_range_with_overflow): ...and here.
1133 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1135 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1137 (extract_range_from_abs_expr): ...here.
1139 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1141 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1142 -fno-omit-frame-pointer when not optimizing.
1144 2018-07-02 Martin Liska <mliska@suse.cz>
1147 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1148 (propagate_nothrow): Likewise.
1150 2018-07-02 Martin Liska <mliska@suse.cz>
1153 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1155 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1157 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1158 function in r262149, changing "loc" param from source_location to
1159 const dump_location_t &.
1160 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1161 declaration, as above.
1163 2018-07-01 Paul Koning <ni1d@arrl.net>
1165 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1166 -munit-asm, -mgnu-asm, -mdec-asm.
1167 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1168 (pdp11_output_labelref): New.
1169 (pdp11_output_def): New.
1170 (pdp11_output_addr_vec_elt): New.
1171 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1172 %# and %@ format codes.
1173 (pdp11_option_override): New.
1174 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1175 (pdp11_output_ident): New.
1176 (pdp11_asm_named_section): New.
1177 (pdp11_asm_init_sections): New.
1178 (pdp11_file_start): New.
1179 (pdp11_file_end): New.
1180 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1181 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1182 %o, like %c but octal.
1183 (pdp11_option_override): New.
1184 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1186 (DATA_SECTION_ASM_OP): Ditto.
1187 (READONLY_DATA_SECTION_ASM_OP): New.
1188 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1189 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1190 (ASM_OUTPUT_LABELREF): Ditto.
1191 (ASM_OUTPUT_DEF): Ditto.
1192 (ASM_OUTPUT_EXTERNAL): New.
1193 (ASM_OUTPUT_SOURCE_FILENAME): New.
1194 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1195 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1196 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1197 %# and %@ format codes.
1198 * config/pdp11/pdp11.opt (mgnu-asm): New.
1199 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1200 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1201 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1203 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1205 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1206 dereferencing path[] beyond its length.
1208 (debug_all_paths): New.
1209 (rewire_first_differing_edge): New.
1210 (adjust_paths_after_duplication): New.
1211 (duplicate_thread_path): Call adjust_paths_after_duplication.
1213 (thread_through_all_blocks): Add new argument to
1214 duplicate_thread_path.
1216 2018-06-30 Jim Wilson <jimw@sifive.com>
1218 * config/riscv/predicates.md (p2m1_shift_operand): New.
1219 (high_mask_shift_operand): New.
1220 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1221 pattern using p2m1_shift_operand.
1222 (lshsi3_zero_extend_3+2): New combiner pattern using
1223 high_mask_shift_operand.
1225 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1227 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1229 (vect_recog_rotate_pattern): ...here.
1230 (vect_convert_input): Try to insert casts of invariants in the
1232 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1233 preheader to be empty.
1235 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1237 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1238 vector type. If given, install it in the new statement's
1240 (vect_element_precision): New function.
1241 (vect_unpromoted_value): New struct.
1242 (vect_unpromoted_value::vect_unpromoted_value): New function.
1243 (vect_unpromoted_value::set_op): Likewise.
1244 (vect_look_through_possible_promotion): Likewise.
1245 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1246 (vect_widened_op_tree, vect_convert_input): Likewise.
1247 (vect_convert_inputs, vect_convert_output): Likewise.
1248 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1249 to handle the optional cast of the multiplication result and
1250 vect_widened_op_tree to detect the widened multiplication itself.
1251 Do not require the input and output of promotion casts to have
1252 the same sign, but base the signedness of the operation on the
1253 input rather than the result. If the pattern includes two
1254 promotions, check that those promotions have the same sign.
1255 Do not restrict the MULT_EXPR handling to a double-width result;
1256 handle quadruple-width results and wider. Use vect_convert_inputs
1257 to convert the inputs to the common type.
1258 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1259 to handle the optional cast of the ABS result. Also allow a sign
1260 change or a sign extension between the ABS and MINUS.
1261 Use vect_widened_op_tree to detect the widened subtraction and use
1262 vect_convert_inputs to convert the inputs to the common type.
1263 (vect_handle_widen_op_by_const): Delete.
1264 (vect_recog_widen_op_pattern): New function.
1265 (vect_recog_widen_mult_pattern): Use it.
1266 (vect_recog_widen_shift_pattern): Likewise.
1267 (vect_recog_widen_sum_pattern): Use
1268 vect_look_through_possible_promotion to handle the promoted
1271 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1273 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1274 the containing gimple_seq *.
1275 * gimple-iterator.h (gsi_for_stmt): Declare it.
1276 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1277 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1278 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1279 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1280 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1282 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1284 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1285 statement being replaced is part of an existing pattern
1286 definition sequence, inserting the new pattern statements before
1288 (vect_pattern_recog_1): Don't return a bool. If the statement
1289 is already part of a pattern, instead apply pattern matching
1290 to the pattern definition statements. Don't clear the
1291 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1292 (vect_pattern_recog): Don't break after the first match;
1293 continue processing the pattern definition statements instead.
1294 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1296 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1298 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1299 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1300 (vect_recog_widen_sum_pattern): Use it.
1302 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1304 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1305 phi is not a pattern statement and has not been replaced by
1306 a pattern statement.
1307 * tree-vect-patterns.c (type_conversion_p): Don't check
1308 STMT_VINFO_IN_PATTERN_P.
1309 (vect_recog_vector_vector_shift_pattern): Likewise.
1310 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1311 the pattern statement rather than the original statement; check
1312 directly for a WIDEN_MULT_EXPR here.
1313 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1314 vect_is_simple_use to return the pattern statement rather
1315 than the original statement; use is_pattern_stmt_p to check
1316 for such a pattern statement.
1317 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1318 to return the pattern statement rather than the original statement;
1319 don't do the same transformation here.
1320 (vect_is_simple_use): If the defining statement has been replaced
1321 by a pattern statement, return the pattern statement instead.
1322 Remove the corresponding (local) transformation from the vectype
1325 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1327 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1328 end and default to null.
1329 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1330 (vectorizable_reduction): Update calls accordingly, dropping the
1331 gimple ** argument if the passed-back statement isn't needed.
1332 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1333 (vect_recog_rotate_pattern): Likewise.
1334 (vect_recog_mask_conversion_pattern): Likewise.
1335 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1336 (vect_mask_constant_operand_p): Likewise.
1337 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1338 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1339 (get_group_load_store_type, get_load_store_type): Likewise.
1340 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1341 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1342 (vectorizable_conversion, vectorizable_assignment): Likewise.
1343 (vectorizable_shift, vectorizable_operation): Likewise.
1344 (vectorizable_store, vect_is_simple_cond): Likewise.
1345 (vectorizable_condition, vectorizable_comparison): Likewise.
1346 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1347 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1348 and move it to the end. Cope with null def_stmt_outs.
1350 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1352 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1354 2018-06-29 Jeff Law <law@redhat.com>
1356 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1357 displacements for TARGET_V850E2V3 and newer.
1358 (TARGET_LRA_P): Remove. Defaults to LRA now.
1359 * config/v850/v850.md (sign23byte_load): Remove.
1360 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1361 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1363 2018-06-29 Martin Liska <mliska@suse.cz>
1366 * coverage.c (coverage_init): Mangle full path name.
1367 * doc/invoke.texi: Document the change.
1368 * gcov-io.c (mangle_path): New.
1369 * gcov-io.h (mangle_path): Likewise.
1370 * gcov.c (mangle_name): Use mangle_path for path mangling.
1372 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1374 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1375 if starting source register is not even.
1377 2018-06-29 Martin Liska <mliska@suse.cz>
1379 PR tree-optimization/86263
1380 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1381 Make edge redirection.
1383 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1385 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1386 (dump_scope_depth): New variable.
1387 (get_dump_scope_depth): New function.
1388 (dump_begin_scope): New function.
1389 (dump_end_scope): New function.
1390 * dumpfile.h (get_dump_scope_depth): New declaration.
1391 (dump_begin_scope): New declaration.
1392 (dump_end_scope): New declaration.
1393 (class auto_dump_scope): New class.
1394 (AUTO_DUMP_SCOPE): New macro.
1395 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1398 2018-06-29 Richard Biener <rguenther@suse.de>
1400 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1401 compute_all_dependences succeeds.
1402 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1403 exceed --param loop-max-datarefs-for-datadeps.
1405 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1407 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1409 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1412 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1413 alternative 0 in preferred_for_speed attribute.
1415 2018-06-28 Paul Koning <ni1d@arrl.net>
1417 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1418 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1419 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1420 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1421 for shift insn patterns.
1423 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1425 * cgraph.c (cgraph_node::get_body): Replace assignments to
1426 "dump_file" with calls to set_dump_file.
1427 * dumpfile.c (alt_dump_file): Make static, and group with...
1428 (alt_flags): ...this definition.
1429 (dumps_are_enabled): New variable.
1430 (refresh_dumps_are_enabled): New function.
1431 (set_dump_file): New function.
1432 (set_alt_dump_file): New function.
1433 (gcc::dump_manager::dump_start): Replace assignments to
1434 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1436 (gcc::dump_manager::dump_finish): Likewise.
1437 * dumpfile.h (alt_dump_file): Delete decl.
1438 (dumps_are_enabled): New variable decl.
1439 (set_dump_file): New function decl.
1440 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1442 * tree-nested.c (lower_nested_functions): Replace assignments to
1443 "dump_file" with calls to set_dump_file.
1445 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1447 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1448 goto_locus of each outgoing edge of each basic block.
1450 2018-06-28 Richard Biener <rguenther@suse.de>
1452 * dwarf2out.c (decl_scope_table): Remove.
1453 (push_decl_scope): Likewise.
1454 (pop_decl_scope): Likewise.
1455 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1456 (gen_struct_or_union_type_die): Likewise.
1457 (gen_tagged_type_die): Likewise.
1458 (dwarf2out_init): Do not initialize decl_scope_table.
1459 (dwarf2out_c_finalize): Do not free it.
1461 2018-06-28 Richard Biener <rguenther@suse.de>
1463 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1464 deciding whether to not re-use a DIE.
1466 2018-06-28 Richard Biener <rguenther@suse.de>
1468 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1469 DW_AT_abstract_origin attribute.
1471 2018-06-28 Martin Liska <mliska@suse.cz>
1473 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1474 Use newly introduced constants.
1475 * tree-switch-conversion.h (struct jump_table_cluster):
1476 Define max_ratio_for_size and max_ratio_for_speed.
1478 2018-06-28 Martin Liska <mliska@suse.cz>
1480 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1481 Add new checking assert to catch invalid state.
1482 (jump_table_cluster::can_be_handled): Handle single case
1484 (jump_table_cluster::is_beneficial): Bail out for such case.
1485 (bit_test_cluster::find_bit_tests):
1486 Add new checking assert to catch invalid state.
1487 (bit_test_cluster::can_be_handled): Handle single case
1489 (bit_test_cluster::is_beneficial): Bail out for such case.
1490 (switch_decision_tree::analyze_switch_statement):
1493 2018-06-28 Martin Liska <mliska@suse.cz>
1495 * common.opt: Introduce -completion option.
1496 * gcc.c (driver_handle_option): Handle it.
1497 (driver::main): Print completions if completion
1499 * opt-suggestions.c (option_proposer::get_completions):
1501 (option_proposer::suggest_completion): Likewise.
1502 (option_proposer::find_param_completions): Likewise.
1503 (verify_autocompletions): Likewise.
1504 (test_completion_valid_options): Likewise.
1505 (test_completion_valid_params): Likewise.
1506 (in_completion_p): Likewise.
1507 (empty_completion_p): Likewise.
1508 (test_completion_partial_match): Likewise.
1509 (test_completion_garbage): Likewise.
1510 (opt_proposer_c_tests): Likewise.
1511 * opt-suggestions.h: Declare new functions.
1512 * opts.c (common_handle_option): Handle OPT__completion_.
1513 * selftest-run-tests.c (selftest::run_tests): Add
1514 opt_proposer_c_tests.
1515 * selftest.c (assert_str_startswith): New.
1516 * selftest.h (assert_str_startswith): Likewise.
1517 (opt_proposer_c_tests): New.
1518 (ASSERT_STR_STARTSWITH): Likewise.
1520 2018-06-28 Martin Liska <mliska@suse.cz>
1522 * Makefile.in: Add opt-suggestions.o.
1523 * gcc-main.c: Include opt-suggestions.h.
1524 * gcc.c (driver::driver): Likewise.
1525 (driver::~driver): Remove m_option_suggestions.
1526 (driver::build_option_suggestions): Moved to option_proposer.
1527 (driver::suggest_option): Likewise.
1528 (driver::handle_unrecognized_options): Use option_proposer.
1529 * gcc.h (class driver): Add new memver m_option_proposer.
1530 * opt-suggestions.c: New file.
1531 * opt-suggestions.h: New file.
1533 2018-06-28 Martin Liska <mliska@suse.cz>
1535 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1536 (auto_string_vec::~auto_string_vec): Likewise.
1538 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1540 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1541 prevent_decl_creation_for_types fields up and add reset_location field.
1542 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1543 statement if id->reset_location is true.
1544 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1545 id->reset_location is true.
1546 (copy_phis_for_bb): Force input_location on the arguments if
1547 id->reset_location is true.
1548 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1549 is set on the function to be inlined.
1551 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1553 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1555 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1557 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1558 registers for Pmode.
1559 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1560 hard registers for the clobbered pseudo.
1562 2018-06-27 Paul Koning <ni1d@arrl.net>
1564 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1565 mutually exclusive options.
1566 * config/pdp11/constraints.md (h): New constraint.
1567 (O): Update definition to match shift code generation.
1568 (D): New constraint.
1569 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1571 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1573 (output_jump): Change arguments.
1574 (pdp11_fixed_cc_regs): New function.
1575 (pdp11_cc_mode): Ditto.
1576 (pdp11_expand_shift): Ditto.
1577 (pdp11_assemble_shift): Ditto.
1578 (pdp11_small_shift): Ditto.
1579 (pdp11_branch_cost): Remove.
1580 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1582 (pdp11_register_move_cost): Update for CC registers.
1583 (pdp11_rtx_costs): Add case for LSHIFTRT.
1584 (pdp11_output_jump): Add CCNZ mode conditional branches.
1585 (notice_update_cc_on_set): Remove.
1586 (pdp11_cc_mode): New function.
1587 (simple_memory_operand): Correct pre/post decrement case.
1588 (no_side_effect_operand): New function.
1589 (pdp11_regno_reg_class): Add CC_REGS class.
1590 (pdp11_fixed_cc_regs): New function.
1591 (pdp11_small_shift): New function.
1592 (pdp11_expand_shift): New function to expand shift insns.
1593 (pdp11_assemble_shift): New function to output shifts.
1594 (pdp11_branch_cost): Remove.
1595 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1596 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1597 (WCHAR_TYPE): Ditto.
1598 (PTRDIFF_TYPE): Ditto.
1599 (ADJUST_INSN_LENGTH): New macro.
1600 (FIXED_REGISTERS): Add CC registers.
1601 (CALL_USED_REGISTERS): Ditto.
1603 (REG_CLASS_NAMES): Ditto.
1604 (REG_CLASS_CONTENTS): Ditto.
1605 (SELECT_CC_MODE): Use new function.
1606 (TARGET_FLAGS_REGNUM): New macro.
1607 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1608 (cc0_reg_rtx): Remove.
1609 (CC_STATUS_MDEP): Remove.
1610 (CC_STATUS_MDEFP_INIT): Remove.
1611 (CC_IN_FPU): Remove.
1612 (NOTICE_UPDATE_CC): Remove.
1613 (REGISTER_NAMES): Add CC registers.
1614 (BRANCH_COST): Change to constant 1.
1615 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1617 * config/pdp11/pdp11.opt (mbcopy): Remove.
1618 (mbcopy-builtin): Remove.
1619 (mbranch-cheap): Remove.
1620 (mbranch-expensive): Remove.
1621 * config/pdp11/predicates.md (expand_shift_operand): Update to
1622 match shift code generation.
1623 (ccnz_operator): New predicate.
1624 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1625 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1626 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1628 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1629 description of O constraint.
1631 2018-06-27 Jeff Law <law@redhat.com>
1632 Austin Law <austinklaw@gmail.com>
1634 * config/v850/v850.md (addsi3_set_flags): New pattern.
1635 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1636 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1637 (zero_extendhisi2_v850_set_flags): Likewise.
1638 (zero_extendqisi2_v850_set_flags): Likewise.
1639 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1640 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1641 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1643 * config/v850/v850-protos.h (notice_update_cc): Remove.
1644 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1645 (v850_print_operand): Handle 'D' and "d".
1646 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1647 Add handling of arithmetic/logical operations compared against zero.
1648 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1649 Do not look at v850_compare_op, instead get mode from last argument.
1650 (v850_gen_compare): Remove
1651 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1652 after reload for prologue insns.
1653 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1655 (construct_save_jarl): Likewise.
1656 (TARGET_FLAGS_REGNUM): Define.
1657 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1658 (NOTICE_UPDATE_CC): Remove.
1659 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1660 than cc0. Conditionalize on reload_completed.
1661 (cmpsi_insn, setfcc_insn): Likewise.
1662 (tst1 splitter): Turn into define_and_split which sets the flags
1664 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1665 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1666 (cstoresf4, cstoredf4): Clobber the flags.
1667 (cmpsi, cmpsf, cmpdf): Remove expanders.
1668 (setf_insn): Remove pattern.
1669 (addsi3): Turn into define_and_split which clobbers the flags after
1670 reload and a suitable pattern (addsi3_clobber_flags) for use after
1672 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1673 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1674 (ashrsi3, ashrsi3_v850e2): Likewise.
1675 (bins): Clobber the flags.
1676 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1677 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1678 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1679 (call_value_internal_short, call_value_internal_long): Likewise.
1680 (callt_save_interrupt, callt_return_interrupt): Likewise.
1681 (save_interrupt, return_interrupt): Likewise.
1682 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1683 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1684 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1685 (All FP comparisons): Only allow after reload has completed.
1687 (divh, divhu): Tweak output template.
1688 (branch_z_normal, branch_z_invert): Remove
1689 (branch_nz_normal, branch_nz_invert): Likewise.
1690 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1692 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1693 * config/v850/v850.c (notice_update_cc): Remove.
1694 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1695 (CC_NO_CARRY): Likewise.
1696 (NOTICE_UPDATE_CC): Define to nothing.
1697 * config/v850/v850.md: Remove block comment on cc0 handling
1698 Remove "cc" attribute from all patterns. Remove cc_status handling
1699 from all patterns. Minor formatting fixes.
1701 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1703 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1704 (cortex-a76.cortex-a55): Likewise.
1705 * config/aarch64/aarch64-tune.md: Regenerate.
1706 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1707 cortex-a76.cortex-a55.
1709 2018-06-27 Jeff Law <law@redhat.com>
1711 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1712 (MULTILIB_DIRNAMES): Similarly.
1714 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1716 * gimple.h (gimple_return_retbnd): Delete.
1717 (gimple_return_set_retbnd): Likewise.
1718 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1719 gimple_return_set_retbnd.
1720 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1721 gimple_return_retbnd and adjust.
1722 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1723 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1724 Explicitly return NULL in a couple more cases. Move assertion
1725 on debug statements and remove unreachable code.
1726 (reset_debug_binding): Do not test id->retbnd.
1727 (expand_call_inline): Do not set it.
1729 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1731 * configure.ac: Add --disable-gcov option.
1732 * configure: Regenerate.
1733 * Makefile.in: Honour @enable_gcov@.
1734 * doc/install.texi: Document --disable-gcov.
1736 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1738 * config/arm/arm-cpus.in (cortex-a76): New entry.
1739 (cortex-a76.cortex-a55): Likewise.
1740 * config/arm/arm-tables.opt: Regenerate.
1741 * config/arm/arm-tune.md: Likewise.
1742 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1743 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1744 cortex-a76.cortex-a55.
1746 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1749 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1751 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1753 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1755 (EPILOGUE_USES): Likewise.
1757 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1759 * tree-inline.c (remap_location): New function extracted from...
1760 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1761 (copy_phis_for_bb): ...here. Call remap_location.
1762 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1764 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1766 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1767 unaligned vsx for 16B memset.
1769 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1772 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1773 ieee128_float_type_node to long_double_type_node unless
1774 TARGET_LONG_DOUBLE_128 is set.
1776 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1778 * cfgloop.c (get_loop_location): Convert return type from
1779 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1780 by implicit construction from rtx_insn *, and using
1781 dump_user_location_t::from_function_decl for the fallback case.
1782 * cfgloop.h (get_loop_location): Convert return type from
1783 location_t to dump_user_location_t.
1784 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1785 dump_printf_loc to pass in a dump_location_t rather than a
1786 location_t, via the gimple stmt.
1787 * coverage.c (get_coverage_counts): Update calls to
1788 dump_printf_loc to pass in dump_location_t rather than a
1790 * doc/optinfo.texi (Dump types): Convert example of
1791 dump_printf_loc from taking "locus" to taking "insn". Update
1792 description of the "_loc" calls to cover dump_location_t.
1793 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1795 (dump_user_location_t::dump_user_location_t): New constructors,
1796 from gimple *stmt and rtx_insn *.
1797 (dump_user_location_t::from_function_decl): New function.
1798 (dump_loc): Make static.
1799 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1800 const dump_location_t &.
1801 (dump_generic_expr_loc): Delete.
1802 (dump_printf_loc): Convert param "loc" from location_t to
1803 const dump_location_t &.
1804 (selftest::test_impl_location): New function.
1805 (selftest::dumpfile_c_tests): New function.
1806 * dumpfile.h: Include "profile-count.h".
1807 (class dump_user_location_t): New class.
1808 (struct dump_impl_location_t): New struct.
1809 (class dump_location_t): New class.
1810 (dump_printf_loc): Convert 2nd param from source_location to
1811 const dump_location_t &.
1812 (dump_generic_expr_loc): Delete.
1813 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1814 const dump_location_t &.
1815 * gimple-fold.c (fold_gimple_assign): Update call to
1816 dump_printf_loc to pass in a dump_location_t rather than a
1817 location_t, via the gimple stmt.
1818 (gimple_fold_call): Likewise.
1819 * gimple-loop-interchange.cc
1820 (loop_cand::analyze_iloop_reduction_var): Update for change to
1821 check_reduction_path.
1822 (tree_loop_interchange::interchange): Update for change to
1824 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1825 change in return-type of find_loop_location.
1826 (graphite_regenerate_ast_isl): Likewise.
1827 * graphite-optimize-isl.c (optimize_isl): Likewise.
1828 * graphite.c (graphite_transform_loops): Likewise.
1829 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1830 pass in a dump_location_t rather than a location_t, via the
1832 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1833 * ipa.c (walk_polymorphic_call_targets): Likewise.
1834 * loop-unroll.c (report_unroll): Convert "locus" param from
1835 location_t to dump_location_t.
1836 (decide_unrolling): Update for change to get_loop_location's
1838 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1839 location_t to dump_user_location_t.
1840 (grid_find_single_omp_among_assignments_1): Updates calls to
1841 dump_printf_loc to pass in a dump_location_t rather than a
1842 location_t, via the gimple stmt.
1843 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1844 location_t to dump_location_t. Updates calls to dump_printf_loc
1845 to pass in a dump_location_t rather than a location_t, via the
1847 (grid_inner_loop_gridifiable_p): Likewise.
1848 (grid_dist_follows_simple_pattern): Likewise.
1849 (grid_gfor_follows_tiling_pattern): Likewise.
1850 (grid_target_follows_gridifiable_pattern): Likewise.
1851 (grid_attempt_target_gridification): Convert initialization
1852 of local "grid" from memset to zero-initialization; FIXME: does
1853 this require C++11? Update call to dump_printf_loc to pass in a
1854 optinfo_location rather than a location_t, via the gimple stmt.
1855 * profile.c (read_profile_edge_counts): Updates call to
1856 dump_printf_loc to pass in a dump_location_t rather than a
1858 (compute_branch_probabilities): Likewise.
1859 * selftest-run-tests.c (selftest::run_tests): Call
1861 * selftest.h (dumpfile_c_tests): New decl.
1862 * tree-loop-distribution.c (pass_loop_distribution::execute):
1863 Update for change in return type of find_loop_location.
1864 * tree-parloops.c (parallelize_loops): Likewise.
1865 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1866 "locus" from location_t to dump_user_location_t.
1867 (canonicalize_loop_induction_variables): Likewise.
1868 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1869 for change in return type of find_loop_location.
1870 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1871 to dump_printf_loc to pass in a dump_location_t rather than a
1872 location_t, via the stmt.
1873 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1875 * tree-vect-loop-manip.c (find_loop_location): Convert return
1876 type from source_location to dump_user_location_t.
1877 (vect_do_peeling): Update for above change.
1878 (vect_loop_versioning): Update for change in type of
1880 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1881 from location_t to dump_user_location_t.
1882 (vect_estimate_min_profitable_iters): Update for change in type
1884 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1885 location_t to dump_location_t.
1886 (vect_slp_bb): Update for change in type of vect_location.
1887 * tree-vectorizer.c (vect_location): Convert from source_location
1888 to dump_user_location_t.
1889 (try_vectorize_loop_1): Update for change in vect_location's type.
1890 (vectorize_loops): Likewise.
1891 (increase_alignment): Likewise.
1892 * tree-vectorizer.h (vect_location): Convert from source_location
1893 to dump_user_location_t.
1894 (find_loop_location): Convert return type from source_location to
1895 dump_user_location_t.
1896 (check_reduction_path): Convert 1st param from location_t to
1897 dump_user_location_t.
1898 * value-prof.c (check_counter): Update call to dump_printf_loc to
1899 pass in a dump_user_location_t rather than a location_t; update
1900 call to error_at for change in type of "locus".
1901 (check_ic_target): Update call to dump_printf_loc to
1902 pass in a dump_user_location_t rather than a location_t, via the
1905 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1907 * config/s390/s390.h (enum processor_flags): Do not use
1908 default tune parameter when -march was specified.
1910 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1913 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1914 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1917 2018-06-26 Richard Biener <rguenther@suse.de>
1919 PR tree-optimization/86287
1921 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1922 (vect_analyze_loop): Initialize n_stmts.
1924 2018-06-26 Richard Biener <rguenther@suse.de>
1927 * fold-const.c (fold_convertible_p): Pointer extension
1930 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1933 * dwarf2out.c (loc_list_has_views): Adjust comments.
1934 (dw_loc_list): Split single cross-partition range with
1937 2018-06-25 Jeff Law <law@redhat.com>
1939 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1940 on -mbig-switch by default.
1942 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1944 (const_float_0_operand): Remove unused predicate.
1945 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1946 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1947 (recipsf2): New expander. Original pattern now called
1949 (recipdf2, recipdf2_insn): Similarly.
1950 (rsqrtsf2, rsqrtsf2_insn): Similarly
1951 (rsqrtdf2, rsqrtdf2_insn): Similarly
1953 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1955 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1956 Simplify logic for FreeBSD (twice).
1958 2018-06-25 Martin Sebor <msebor@redhat.com>
1960 PR tree-optimization/86204
1961 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1962 a strnlen result if it's less than the length of the string.
1964 2018-06-25 Martin Sebor <msebor@redhat.com>
1966 PR tree-optimization/85700
1967 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1968 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1969 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1971 2018-06-25 Martin Sebor <msebor@redhat.com>
1973 * doc/extend.texi (Zero-length arrays): Update and clarify.
1975 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1977 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1978 added IEEE/IBM long double multilib support on PowerPC little
1979 endian Linux systems.
1980 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1981 (MULTILIB_DEFAULTS): Likewise.
1982 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1984 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1985 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1986 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1988 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1991 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1992 (REORDER_45): Likewise.
1994 2018-06-25 Jeff Law <law@redhat.com>
1996 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1997 dividend to 32 bits. Adjust length.
1998 (udivmodhi4): Cleanup output template. Fix length.
2000 2018-06-25 Carl Love <cel@us.ibm.com>
2002 * config/rs6000/vsx.md: Change word selector to prefered location.
2004 2018-06-25 Richard Biener <rguenther@suse.de>
2006 PR tree-optimization/86304
2007 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2008 epilogue-if-converted loops as well.
2010 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2012 * lto-section-out.c (lto_begin_section): Do not print section
2013 name for noaddr and unnumbered dumps.
2015 2018-06-25 Richard Biener <rguenther@suse.de>
2017 * tree-vectorizer.h (struct vec_info_shared): New structure
2018 with parts split out from struct vec_info and loop_nest from
2019 struct _loop_vec_info.
2020 (struct vec_info): Adjust accordingly.
2021 (struct _loop_vec_info): Likewise.
2022 (LOOP_VINFO_LOOP_NEST): Adjust.
2023 (LOOP_VINFO_DATAREFS): Likewise.
2024 (LOOP_VINFO_DDRS): Likewise.
2025 (struct _bb_vec_info): Likewise.
2026 (BB_VINFO_DATAREFS): Likewise.
2027 (BB_VINFO_DDRS): Likewise.
2028 (struct _stmt_vec_info): Add dr_aux member.
2029 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2030 (DR_MISALIGNMENT_UNINITIALIZED): New.
2031 (set_dr_misalignment): Adjust.
2032 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2033 (vect_analyze_loop): Adjust prototype.
2034 (vect_analyze_loop_form): Likewise.
2035 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2036 Compute dependences lazily.
2037 (vect_record_base_alignments): Use shared datarefs/ddrs.
2038 (vect_verify_datarefs_alignment): Likewise.
2039 (vect_analyze_data_refs_alignment): Likewise.
2040 (vect_analyze_data_ref_accesses): Likewise.
2041 (vect_analyze_data_refs): Likewise.
2042 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2043 constructor parameter for shared part.
2044 (vect_analyze_loop_form): Pass in shared part and adjust.
2045 (vect_analyze_loop_2): Pass in storage for the number of
2046 stmts. Move loop nest finding to the caller. Compute
2048 (vect_analyze_loop): Pass in shared part.
2049 (vect_transform_loop): Verify shared datarefs are unchanged.
2050 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2051 constructor parameter for shared part.
2052 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2053 (vect_slp_bb): Verify shared datarefs are unchanged before
2055 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2057 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2058 DR_MISALIGNMENT_UNINITIALIZED.
2059 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2060 parameter for shared part.
2061 (vec_info::~vec_info): Adjust.
2062 (vec_info_shared::vec_info_shared): New.
2063 (vec_info_shared::~vec_info_shared): Likewise.
2064 (vec_info_shared::save_datarefs): Likewise.
2065 (vec_info_shared::check_datarefs): Likewise.
2066 (try_vectorize_loop_1): Construct shared part live for analyses
2067 of a single loop for multiple vector sizes.
2068 * tree-parloops.c (gather_scalar_reductions): Adjust.
2070 2018-06-25 Richard Biener <rguenther@suse.de>
2072 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2073 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2074 (vect_analyze_data_refs): Remove similar code from here and
2075 simplify accordingly.
2077 2018-06-25 Richard Biener <rguenther@suse.de>
2079 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2080 for reverse storage order accesses rather than asserting
2081 they cannot happen here.
2083 2018-06-25 Tom de Vries <tdevries@suse.de>
2086 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2087 Use data16 instead of .byte for insn prefix.
2089 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2092 * parser.c (make_char_string_pack): Pass this literal chars
2093 through cpp_interpret_string.
2094 (cp_parser_userdef_numeric_literal): Check the result of
2095 make_char_string_pack.
2097 2018-06-24 Maya Rashish <coypu@sdf.org>
2099 * ginclude/stddef.h: Simplify conditions around avoiding
2100 re-definition of __size_t.
2102 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2104 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2105 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2107 2018-06-22 Maya Rashish <coypu@sdf.org>
2109 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2110 made to OpenBSD and NetBSD through the years.
2112 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2114 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2115 behavior of vec_pack (vector double, vector double) to match
2116 behavior of vec_float2 (vector double, vector double).
2118 2018-06-22 Olivier Hainque <hainque@adacore.com>
2120 * gimplify.c (gimplify_function_tree): Prevent creation
2121 of a trampoline for the address of the current function
2122 passed to entry/exit instrumentation hooks.
2124 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2127 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2130 2018-06-22 Martin Liska <mliska@suse.cz>
2132 PR tree-optimization/86263
2133 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2134 Bail out if is_enabled is false.
2135 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2137 (jump_table_cluster::is_enabled): New function.
2139 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2141 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2142 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2143 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2144 (lto_input_ts_binfo_tree_pointers): Likewise.
2145 * tree-streamer-out.c (streamer_write_tree_bitfields,
2146 write_ts_binfo_tree_pointers): Likewise.
2147 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2149 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2151 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2153 2018-06-22 Martin Liska <mliska@suse.cz>
2155 * symbol-summary.h (get): Make it pure and inline move
2156 functionality from ::get function.
2157 (get): Remove and inline into ::get and ::get_create.
2158 (get_create): Move code from ::get function.
2160 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2163 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2164 -x assembler-with-cpp.
2166 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2168 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2169 _FILE_OFFSET_BITS=64 for C++.
2171 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2173 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2174 conversion insn that shows up when pr85657-3.c is compiled using
2175 IEEE 128-bit long double.
2176 (neg<mode>2_internal): Use the correct mode to check whether the
2177 mode is IBM extended.
2178 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2179 multiply and divide external functions from being created more
2182 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2184 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2186 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2187 the edge can be forwarded.
2188 (cfg_layout_merge_blocks): Likewise.
2190 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2192 * except.c (finish_eh_generation): Commit edge insertions only after
2193 the EH edges have been redirected from post-landing to landing pads.
2195 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2197 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2198 create_tmp_var_for to create the FRAME decl.
2199 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2201 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2203 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2204 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2205 debug statement when resetting its value.
2206 (expand_call_inline): Copy the locus of the call onto the assignment
2207 of the return value, if any. Use local variable in more cases.
2209 2018-06-21 Martin Liska <mliska@suse.cz>
2211 * ipa-pure-const.c (propagate_nothrow): Use
2212 funct_state_summaries->get.
2213 (dump_malloc_lattice): Likewise.
2214 (propagate_malloc): Likewise.
2216 2018-06-21 Richard Biener <rguenther@suse.de>
2218 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2219 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2220 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2222 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2223 BLOCK_ABSTRACT_ORIGIN unconditionally.
2225 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2227 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2229 * ipa-reference.c (ipa_reference_c_finalize): Delete
2230 ipa_ref_opt_sum_summaries and set it to NULL.
2232 2018-06-21 Tom de Vries <tdevries@suse.de>
2234 PR tree-optimization/85859
2235 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2236 test with comment from bb_no_side_effects_p.
2238 2018-06-21 Richard Biener <rguenther@suse.de>
2240 PR tree-optimization/86232
2241 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2242 max for constant niter.
2244 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2246 * config/aarch64/aarch64-simd.md
2247 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2249 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2251 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2252 Make opernads of the unspec commutative.
2254 2018-06-21 Richard Biener <rguenther@suse.de>
2256 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2257 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2258 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2259 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2261 (vect_record_base_alignments): Do not record base alignment
2262 for gathers or scatters.
2263 (vect_compute_data_ref_alignment): Drop return value that is always
2264 true. Bail out early for gathers or scatters.
2265 (vect_enhance_data_refs_alignment): Bail out early for gathers
2267 (vect_find_same_alignment_drs): Likewise.
2268 (vect_analyze_data_refs_alignment): Remove dead code.
2269 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2270 (vect_analyze_data_refs): For possible gathers or scatters do
2271 not create an alternate DR, just check their possible validity
2272 and mark them. Adjust DECL_NONALIASED handling to not rely
2274 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2275 update inits of gathers or scatters.
2276 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2277 Also copy gather/scatter flag to pattern vinfo.
2279 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2281 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2282 behavior of vec_packsu (vector unsigned long long, vector unsigned
2283 long long) to match behavior of vec_packs with same signature.
2285 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2286 Thomas Schwinge <thomas@codesourcery.com>
2287 Cesar Philippidis <cesar@codesourcery.com>
2289 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2290 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2291 (gimplify_adjust_omp_clauses): Likewise.
2292 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2293 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2294 (gimplify_omp_target_update): Update handling of acc update and
2296 * omp-low.c (install_var_field): Remove unused parameter
2297 base_pointers_restrict.
2298 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2299 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2301 (omp_target_base_pointers_restrict_p): Delete.
2302 (scan_omp_target): Update call to scan_sharing_clauses.
2303 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2305 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2306 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2307 (convert_local_omp_clauses): Likewise.
2308 * tree-pretty-print.c (dump_omp_clause): Likewise.
2309 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2311 (omp_clause_code_name): Likewise.
2313 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2316 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2319 PR tree-optimization/86231
2320 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2321 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2323 2018-06-20 Tom de Vries <tdevries@suse.de>
2325 PR tree-optimization/86097
2326 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2327 iv type if signedness of iv type is not the same as that of *nit.
2329 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2331 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2332 EDGE_EH edges, verify they are all EDGE_EH.
2334 2018-06-20 Maya Rashish <coypu@sdf.org>
2336 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2338 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2340 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2341 * config/aarch64/aarch64.c (xgene1_tunings): Add
2342 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2343 (aarch64_mode_valid_for_sched_fusion_p):
2344 Allow 16-byte modes.
2345 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2346 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2348 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2350 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2351 * config/aarch64/iterators.md (VQ2): New mode iterator.
2353 2018-06-20 Martin Liska <mliska@suse.cz>
2355 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2356 Change default ratio from 10 to 8.
2358 2018-06-20 Martin Liska <mliska@suse.cz>
2360 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2362 (bit_test_cluster::find_bit_tests): Likewise.
2363 (switch_decision_tree::analyze_switch_statement): Find clusters.
2364 * tree-switch-conversion.h (struct jump_table_cluster): Document
2367 2018-06-20 Martin Liska <mliska@suse.cz>
2369 * tree-switch-conversion.c (switch_conversion::collect):
2370 Record m_uniq property.
2371 (switch_conversion::expand): Bail out for special conditions.
2372 (group_cluster::~group_cluster): New.
2373 (group_cluster::group_cluster): Likewise.
2374 (group_cluster::dump): Likewise.
2375 (jump_table_cluster::emit): New.
2376 (switch_decision_tree::fix_phi_operands_for_edges): New.
2377 (struct case_node): Remove struct.
2378 (jump_table_cluster::can_be_handled): New.
2379 (case_values_threshold): Moved to header.
2380 (reset_out_edges_aux): Likewise.
2381 (jump_table_cluster::is_beneficial): New.
2382 (bit_test_cluster::can_be_handled): Likewise.
2383 (add_case_node): Remove.
2384 (bit_test_cluster::is_beneficial): New.
2385 (case_bit_test::cmp): New.
2386 (bit_test_cluster::emit): New.
2387 (expand_switch_as_decision_tree_p): Remove.
2388 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2389 (fix_phi_operands_for_edge): Likewise.
2390 (switch_decision_tree::analyze_switch_statement): New.
2391 (compute_cases_per_edge): Move ...
2392 (switch_decision_tree::compute_cases_per_edge): ... here.
2393 (try_switch_expansion): Likewise.
2394 (switch_decision_tree::try_switch_expansion): Likewise.
2395 (record_phi_operand_mapping): Likewise.
2396 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2397 (emit_case_decision_tree): Likewise.
2398 (switch_decision_tree::emit): Likewise.
2399 (balance_case_nodes): Likewise.
2400 (switch_decision_tree::balance_case_nodes): Likewise.
2401 (dump_case_nodes): Likewise.
2402 (switch_decision_tree::dump_case_nodes): Likewise.
2403 (emit_jump): Likewise.
2404 (switch_decision_tree::emit_jump): Likewise.
2405 (emit_cmp_and_jump_insns): Likewise.
2406 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2407 (emit_case_nodes): Likewise.
2408 (switch_decision_tree::emit_case_nodes): Likewise.
2409 (conditional_probability): Remove.
2410 * tree-switch-conversion.h (enum cluster_type): New.
2412 (struct cluster): Likewise.
2413 (cluster::cluster): Likewise.
2414 (struct simple_cluster): Likewise.
2415 (simple_cluster::simple_cluster): Likewise.
2416 (struct group_cluster): Likewise.
2417 (struct jump_table_cluster): Likewise.
2418 (struct bit_test_cluster): Likewise.
2419 (struct min_cluster_item): Likewise.
2420 (struct case_tree_node): Likewise.
2421 (case_tree_node::case_tree_node): Likewise.
2422 (jump_table_cluster::case_values_threshold): Likewise.
2423 (struct case_bit_test): Likewise.
2424 (struct switch_decision_tree): Likewise.
2425 (struct switch_conversion): Likewise.
2426 (switch_decision_tree::reset_out_edges_aux): Likewise.
2428 2018-06-20 Martin Liska <mliska@suse.cz>
2430 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2431 (hoist_edge_and_branch_if_true): Likewise.
2432 (expand_switch_using_bit_tests_p): Likewise.
2433 (struct case_bit_test): Likewise.
2434 (case_bit_test_cmp): Likewise.
2435 (emit_case_bit_tests): Likewise.
2436 (switch_conversion::switch_conversion): New class.
2437 (struct switch_conv_info): Remove old struct.
2438 (collect_switch_conv_info): More to ...
2439 (switch_conversion::collect): ... this.
2440 (check_range): Likewise.
2441 (switch_conversion::check_range): Likewise.
2442 (check_all_empty_except_final): Likewise.
2443 (switch_conversion::check_all_empty_except_final): Likewise.
2444 (check_final_bb): Likewise.
2445 (switch_conversion::check_final_bb): Likewise.
2446 (create_temp_arrays): Likewise.
2447 (switch_conversion::create_temp_arrays): Likewise.
2448 (free_temp_arrays): Likewise.
2449 (gather_default_values): Likewise.
2450 (switch_conversion::gather_default_values): Likewise.
2451 (build_constructors): Likewise.
2452 (switch_conversion::build_constructors): Likewise.
2453 (constructor_contains_same_values_p): Likewise.
2454 (switch_conversion::contains_same_values_p): Likewise.
2455 (array_value_type): Likewise.
2456 (switch_conversion::array_value_type): Likewise.
2457 (build_one_array): Likewise.
2458 (switch_conversion::build_one_array): Likewise.
2459 (build_arrays): Likewise.
2460 (switch_conversion::build_arrays): Likewise.
2461 (gen_def_assigns): Likewise.
2462 (switch_conversion::gen_def_assigns): Likewise.
2463 (prune_bbs): Likewise.
2464 (switch_conversion::prune_bbs): Likewise.
2465 (fix_phi_nodes): Likewise.
2466 (switch_conversion::fix_phi_nodes): Likewise.
2467 (gen_inbound_check): Likewise.
2468 (switch_conversion::gen_inbound_check): Likewise.
2469 (process_switch): Use the newly created class.
2470 (switch_conversion::expand): New.
2471 (switch_conversion::~switch_conversion): New.
2472 * tree-switch-conversion.h: New file.
2474 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2476 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2477 tree-vect-patterns.c.
2478 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2479 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2480 (vect_recog_sad_pattern): Likewise.
2481 (vect_recog_widen_sum_pattern): Likewise.
2482 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2483 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2484 (vect_recog_rotate_pattern): Likewise.
2485 (vect_recog_mult_pattern): Likewise.
2486 (vect_recog_vector_vector_shift_pattern): Likewise.
2487 (vect_recog_divmod_pattern): Likewise.
2488 (vect_recog_mixed_size_cond_pattern): Likewise.
2489 (vect_recog_bool_pattern): Likewise.
2490 (vect_recog_mask_conversion_pattern): Likewise.
2491 (vect_try_gather_scatter_pattern): Likewise.
2492 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2493 (vect_recog_over_widening_pattern): Likewise.
2494 (vect_recog_gather_scatter_pattern): Likewise.
2495 (vect_recog_func_ptr): Move from tree-vectorizer.h
2496 (vect_vect_recog_func_ptrs): Move further down the file.
2497 (vect_recog_func): Likewise. Remove the third argument.
2498 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2499 (vect_pattern_recog_1): Expect the pattern function to do any
2500 necessary target tests. Also expect it to provide a vector type.
2501 Remove the type_in handling.
2503 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2505 * tree-vect-patterns.c (vect_pattern_detected): New function.
2506 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2507 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2508 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2509 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2510 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2511 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2512 (vect_recog_mask_conversion_pattern)
2513 (vect_try_gather_scatter_pattern): Likewise.
2515 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2517 * tree-vect-patterns.c (vect_get_internal_def): New function.
2518 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2519 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2520 (search_type_for_mask_1): Use it.
2522 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2524 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2525 redundant WIDEN_SUM_EXPR handling.
2526 (vect_recog_sad_pattern): Likewise.
2528 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2530 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2531 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2532 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2533 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2534 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2536 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2538 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2539 use the stmt_vec_info of the original bb statement for the
2540 new zero assignment, even if the call is part of a pattern.
2542 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2544 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2545 that the sequence is attached to the original statement rather
2546 than the pattern statement.
2547 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2548 PATTERN_DEF_SEQ from the original statement rather than
2549 the main pattern statement.
2550 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2551 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2552 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2554 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2556 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2557 definition statements before the early exit for statements that aren't
2559 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2561 (vect_transform_loop): ...here. Process pattern definition
2562 statements without first checking whether the main pattern
2563 statement is live or relevant.
2565 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2567 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2568 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2570 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2572 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2573 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2574 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2575 (expand_block_compare): Change select_block_compare_mode call.
2576 (expand_strncmp_align_check): Use new functions, fix comment.
2577 (emit_final_str_compare_gpr): New function.
2578 (expand_strn_compare): Refactor and clean up code.
2579 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2581 2018-06-19 Tony Reix <tony.reix@atos.com>
2582 Damien Bergamini <damien.bergamini@atos.com>
2583 David Edelsohn <dje.gcc@gmail.com>
2585 * collect2.c (static_obj): New variable.
2586 (static_libs): New variable.
2587 (is_in_list): Uncomment declaration.
2588 (main): Track AIX libraries linked statically.
2589 (is_in_list): Uncomment definition.
2590 (scan_prog_file): Don't add AIX shared libraries initializer
2591 to constructor list if linking statically.
2593 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2595 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2597 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2599 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2601 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2604 2018-06-19 Martin Liska <mliska@suse.cz>
2606 * config/i386/i386.c (ix86_can_inline_p): Do not use
2607 ipa_fn_summaries::get_create.
2608 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2610 (devirtualization_time_bonus): Likewise.
2611 (ipcp_propagate_stage): Likewise.
2612 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2613 (edge_set_predicate): Likewise.
2614 (evaluate_conditions_for_known_args): Likewise.
2615 (evaluate_properties_for_edge): Likewise.
2616 (ipa_call_summary::reset): Tranform to ...
2617 (ipa_call_summary::~ipa_call_summary): ... this.
2618 (ipa_fn_summary::reset): Transform to ...
2619 (ipa_fn_summary::~ipa_fn_summary): ... this.
2620 (ipa_fn_summary_t::remove): Rename to ...
2621 (ipa_fn_summary_t::remove_callees): ... this.
2622 (ipa_fn_summary_t::duplicate): Use placement new
2623 instead of memory copy.
2624 (ipa_call_summary_t::duplicate): Likewise.
2625 (ipa_call_summary_t::remove): Remove.
2626 (dump_ipa_call_summary): Change get_create to get.
2627 (ipa_dump_fn_summary): Dump only when summary exists.
2628 (analyze_function_body): Use symbol_summary::get instead
2630 (compute_fn_summary): Likewise.
2631 (estimate_edge_devirt_benefit): Likewise.
2632 (estimate_edge_size_and_time): Likewise.
2633 (inline_update_callee_summaries): Likewise.
2634 (remap_edge_change_prob): Likewise.
2635 (remap_edge_summaries): Likewise.
2636 (ipa_merge_fn_summary_after_inlining): Likewise.
2637 (write_ipa_call_summary): Likewise.
2638 (ipa_fn_summary_write): Likewise.
2639 (ipa_free_fn_summary): Likewise.
2640 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2641 (struct ipa_call_summary): Likewise.
2642 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2644 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2645 (estimate_size_after_inlining): Likewise.
2646 (estimate_growth): Likewise.
2647 (growth_likely_positive): Likewise.
2648 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2649 (inline_call): Likewise.
2650 * ipa-inline.c (caller_growth_limits): Likewise.
2651 (can_inline_edge_p): Likewise.
2652 (can_inline_edge_by_limits_p): Likewise.
2653 (compute_uninlined_call_time): Likewise.
2654 (compute_inlined_call_time): Likewise.
2655 (want_inline_small_function_p): Likewise.
2656 (edge_badness): Likewise.
2657 (update_caller_keys): Likewise.
2658 (update_callee_keys): Likewise.
2659 (inline_small_functions): Likewise.
2660 (inline_to_all_callers_1): Likewise.
2661 (dump_overall_stats): Likewise.
2662 (early_inline_small_functions): Likewise.
2663 (early_inliner): Likewise.
2664 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2665 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2666 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2667 * ipa-split.c (execute_split_functions): Likewise.
2668 * symbol-summary.h: Likewise.
2669 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2671 2018-06-19 Richard Biener <rguenther@suse.de>
2673 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2674 (vectorize_loops): ... here. Fix dbgcnt handling.
2675 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2677 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2680 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2681 ieee128 argument takes up only one (vector) register, not two (floating
2684 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2686 * gimplify.c (gimplify_init_constructor): Really never clear for an
2687 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2689 2018-06-19 Richard Biener <rguenther@suse.de>
2691 PR tree-optimization/86179
2692 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2693 after failed recognition.
2695 2018-06-18 Martin Sebor <msebor@redhat.com>
2698 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2699 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2700 Handle integer subtraction.
2701 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2702 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2704 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2706 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2707 param from rtx to rtx_insn *.
2708 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2710 (frv_ifcvt_modify_insn): Likwise.
2711 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2712 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2713 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2714 as_a <rtx_insn *> cast to local "unprotected_region" once
2715 it's been established that it's not NULL or pc_rtx.
2716 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2717 param "sethi" from rtx to rtx_insn *.
2718 (nds32_group_float_insns): Likewise for param "insn".
2719 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2721 (vax_output_int_subtract): Likewise.
2722 * config/vax/vax.c (vax_output_int_add): Likewise for param
2724 (vax_output_int_subtract): Likewise.
2725 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2726 (emit_pattern_after): Likewise for param "after".
2727 (emit_insn_after): Likewise.
2728 (emit_jump_insn_after): Likewise.
2729 (emit_call_insn_after): Likewise.
2730 (emit_debug_insn_after): Likewise.
2731 (emit_pattern_before): Likewise for param "before".
2732 (emit_insn_before): Likewise.
2733 (emit_jump_insn_before): Likewise.
2734 * final.c (get_insn_template): Likewise for param "insn", removing
2736 * output.h (get_insn_template): Likewise for 2nd param.
2737 * rtl.h (emit_insn_before): Likewise.
2738 (emit_jump_insn_before): Likewise.
2739 (emit_debug_insn_before_noloc): Likewise.
2740 (emit_insn_after): Likewise.
2741 (emit_jump_insn_after): Likewise.
2742 (emit_call_insn_after): Likewise.
2743 (emit_debug_insn_after): Likewise.
2744 (set_insn_deleted): Likewise for param.
2746 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2749 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2750 floating point modes, so that IFmode is numerically greater than
2751 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2752 to declare the ordering. This prevents IFmode from being
2753 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2754 machine. Include rs6000-modes.h to share the fractional values
2755 between genmodes* and the rest of the compiler.
2759 * config/rs6000/rs6000-modes.h: New file.
2760 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2761 meaning of rs6000_long_double_size so that 126..128 selects an
2762 appropriate 128-bit floating point type.
2763 (rs6000_option_override_internal): Likewise.
2764 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2765 (TARGET_LONG_DOUBLE_128): Change the meaning of
2766 rs6000_long_double_size so that 126..128 selects an appropriate
2767 128-bit floating point type.
2768 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2769 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2770 source and destination to match the standard usage.
2771 (truncifkf2): Likewise.
2772 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2773 ISA 2.07 to use an explicit clobber, instead of passing in a
2775 (copysign<mode>3_soft): Likewise.
2777 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2779 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2780 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2781 (vect_slp_analyze_instance_dependence): Likewise.
2782 (vect_enhance_data_refs_alignment): Likewise.
2783 (vect_analyze_data_refs_alignment): Likewise.
2784 (vect_slp_analyze_and_verify_instance_alignment
2785 (vect_analyze_data_ref_accesses): Likewise.
2786 (vect_prune_runtime_alias_test_list): Likewise.
2787 (vect_analyze_data_refs): Likewise.
2788 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2789 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2790 (vect_analyze_scalar_cycles_1): Likewise.
2791 (vect_get_loop_niters): Likewise.
2792 (vect_analyze_loop_form_1): Likewise.
2793 (vect_update_vf_for_slp): Likewise.
2794 (vect_analyze_loop_operations): Likewise.
2795 (vect_analyze_loop): Likewise.
2796 (vectorizable_induction): Likewise.
2797 (vect_transform_loop): Likewise.
2798 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2799 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2800 (vect_make_slp_decision): Likewise.
2801 (vect_detect_hybrid_slp): Likewise.
2802 (vect_slp_analyze_operations): Likewise.
2803 (vect_slp_bb): Likewise.
2804 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2805 (vectorizable_bswap): Likewise.
2806 (vectorizable_call): Likewise.
2807 (vectorizable_simd_clone_call): Likewise.
2808 (vectorizable_conversion): Likewise.
2809 (vectorizable_assignment): Likewise.
2810 (vectorizable_shift): Likewise.
2811 (vectorizable_operation): Likewise.
2812 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2814 2018-06-18 Martin Sebor <msebor@redhat.com>
2816 PR tree-optimization/81384
2817 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2818 * builtins.c (expand_builtin_strnlen): New function.
2819 (expand_builtin): Call it.
2820 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2821 * builtins.def (BUILT_IN_STRNLEN): New.
2822 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2823 Warn for bounds in excess of maximum object size.
2824 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2825 single-value ranges. Handle strnlen.
2826 (handle_builtin_strlen): Handle strnlen.
2827 (strlen_check_and_optimize_stmt): Same.
2828 * doc/extend.texi (Other Builtins): Document strnlen.
2830 2018-06-18 Maya Rashish <coypu@sdf.org>
2832 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2833 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2834 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2836 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2838 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2840 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2842 * tree.c (escaped_string::escape): Replace cast to char * by
2843 const_cast<char *> (unescaped).
2845 2018-06-18 Nick Clifton <nickc@redhat.com>
2848 * tree.c (escaped_string): New class. Converts an unescaped
2849 string into its escaped equivalent.
2850 (warn_deprecated_use): Use the new class to convert the
2851 deprecation message, if present.
2852 (test_escaped_strings): New self test.
2853 (test_c_tests): Add test_escaped_strings.
2854 * doc/extend.texi (deprecated): Add a note that the
2855 deprecation message is affected by the -fmessage-length
2856 option, and that control characters will be escaped.
2857 (#pragma GCC error): Document this pragma.
2858 (#pragma GCC warning): Likewise.
2859 * doc/invoke.texi (-fmessage-length): Document this option's
2860 effect on the #warning and #error preprocessor directives and
2861 the deprecated attribute.
2863 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2865 * tree.c (decl_value_expr_lookup): Revert latest change.
2866 (decl_value_expr_insert): Likewise.
2868 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2870 * gimplify.c (nonlocal_vlas): Delete.
2871 (nonlocal_vla_vars): Likewise.
2872 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2874 (gimplify_body): Do not create and destroy nonlocal_vlas.
2875 * tree-nested.c: Include diagnostic.h.
2876 (use_pointer_in_frame): Tweak.
2877 (lookup_field_for_decl): Add assertion and declare the transformation.
2878 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2879 internal error when the reference is in a wrong context. Do not
2880 create a debug decl by default.
2881 (note_nonlocal_block_vlas): Delete.
2882 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2883 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2884 create a debug decl by default.
2885 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2886 call is in a wrong context.
2887 (fixup_vla_decls): New function.
2888 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2889 debug variables were created.
2890 * tree.c (decl_value_expr_lookup): Add checking assertion.
2891 (decl_value_expr_insert): Likewise.
2893 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2896 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2897 * tree-scalar-evolution.c (interpret_expr): Likewise.
2898 (expression_expensive_p): Likewise.
2899 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2900 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2901 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2902 (ssa_defined_by_minus_one_stmt_p): New.
2904 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2907 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2908 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2909 * dojump.c (do_jump): Likewise.
2910 * expr.c (expand_expr_real_2): Check operand type's sign.
2911 * fold-const.c (const_unop): Handle ABSU_EXPR.
2912 (fold_abs_const): Likewise.
2913 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2914 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2915 (strip_sign_op_1): Likesise.
2916 * match.pd: Add new pattern to generate ABSU_EXPR.
2917 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2918 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2919 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2920 * tree-inline.c (estimate_operator_cost): Likewise.
2921 * tree-pretty-print.c (dump_generic_node): Likewise.
2922 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2923 * tree.def (ABSU_EXPR): New.
2925 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2928 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2929 documented as preserved for backward compatibility only.
2930 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2932 PR rtl-optimization/86108
2933 * bb-reorder.c (create_forwarder_block): Renamed to ...
2934 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2935 jump from new landing pad to the second part.
2936 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2939 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2942 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2943 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2944 Only call store_expr for halves if the mode is the same.
2947 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2948 Fix up comment formatting.
2950 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2952 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2953 (typed_splay_tree::closure,
2954 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2955 (typed_splay_tree::typed_splay_tree,
2956 typed_splay_tree::operator =): Declared private.
2957 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2958 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2959 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2960 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2961 typed_splay_tree::splay_tree_splay,
2962 typed_splay_tree::splay_tree_foreach_helper,
2963 typed_splay_tree::splay_tree_insert,
2964 typed_splay_tree::splay_tree_remove,
2965 typed_splay_tree::splay_tree_lookup,
2966 typed_splay_tree::splay_tree_predecessor,
2967 typed_splay_tree::splay_tree_successor,
2968 typed_splay_tree::splay_tree_min,
2969 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2970 (typed_splay_tree::root, typed_splay_tree::comp,
2971 typed_splay_tree::delete_key,
2972 typed_splay_tree::delete_value): New data members.
2973 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2974 typed_splay_tree::remove.
2976 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2978 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2979 -mginv and -mno-ginv to the assembler.
2980 * config/mips/mips.opt (-mcrc): New option.
2982 * doc/invoke.text (-mcrc): Document.
2985 2018-06-15 Nick Clifton <nickc@redhat.com>
2988 * tree.c (escaped_string): New class. Converts an unescaped
2989 string into its escaped equivalent.
2990 (warn_deprecated_use): Use the new class to convert the
2991 deprecation message, if present.
2992 (test_escaped_strings): New self test.
2993 (test_c_tests): Add test_escaped_strings.
2994 * doc/extend.texi (deprecated): Add a note that the
2995 deprecation message is affected by the -fmessage-length
2996 option, and that control characters will be escaped.
2997 (#pragma GCC error): Document this pragma.
2998 (#pragma GCC warning): Likewise.
2999 * doc/invoke.texi (-fmessage-length): Document this option's
3000 effect on the #warning and #error preprocessor directives and
3001 the deprecated attribute.
3003 2018-06-15 Richard Biener <rguenther@suse.de>
3005 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3006 here, also noting vector size used.
3007 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3008 size used in MSG_OPTIMIZED_LOCATIONS dump.
3009 (pass_slp_vectorize::execute): Adjust.
3011 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3014 * config/arc/arc.c (arc_return_address_register): Fix
3017 2018-06-15 Richard Biener <rguenther@suse.de>
3020 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3021 leave useless conversion stripping to force_gimple_operand_gsi.
3022 (gimplify_build2): Likewise.
3023 (gimplify_build1): Likewise.
3025 2018-06-15 Richard Biener <rguenther@suse.de>
3028 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3029 before adjusting their block.
3031 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3033 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3034 multilibs for *-*-rtems*.
3035 * config/riscv/t-rtems: New file.
3037 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3040 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3041 unsigned_type_for returns NULL.
3044 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3045 subregs of multi-word pseudos unless the float mode has word size.
3047 2018-06-14 Richard Biener <rguenther@suse.de>
3050 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3051 duplicate and harmful type_hash_canon.
3052 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3054 2018-06-14 Richard Biener <rguenther@suse.de>
3057 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3060 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3062 * config/rtems.h (STDINT_LONG32): Define.
3064 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3065 Prachi Godbole <prachi.godbole@imgtec.com>
3067 * config/mips/mips-cpus.def: Define P6600.
3068 * config/mips/mips-tables.opt: Regenerate.
3069 * config/mips/mips.c (mips_ucbranch_type): New enum.
3070 (mips_rtx_cost_data): Add support for P6600.
3071 (mips_issue_rate): Likewise.
3072 (mips_multipass_dfa_lookahead): Likewise.
3073 (mips_avoid_hazard): Likewise.
3074 (mips_reorg_process_insns): Likewise.
3075 (mips_classify_branch_p6600): New function.
3076 * config/mips/mips.h (TUNE_P6600): New define.
3077 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3078 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3079 * config/mips/mips.md: Include p6600.md.
3080 (processor): Add p6600.
3081 * config/mips/p6600.md: New file.
3082 * doc/invoke.texi: Add p6600 to supported architectures.
3084 2018-06-13 Martin Sebor <msebor@redhat.com>
3086 PR tree-optimization/86114
3087 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3089 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3091 2018-06-13 Richard Biener <rguenther@suse.de>
3093 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3094 Properly set vector type of the intermediate stmt.
3095 * tree-vect-stmts.c (vectorizable_operation): The destination
3096 var always has vectype_out type.
3098 2018-06-13 Jeff Law <law@redhat.com>
3100 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3101 integer 0 for argument to print_rtl_with_bb.
3102 (rl78_reorg): Likewise.
3104 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3106 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3107 from rtx to rtx_insn *.
3108 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3110 (add_sched_insns_for_speculation): Likewise for local "target",
3111 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3112 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3113 from rtx_insn ** to rtx_code_label **.
3114 (reorg_emit_nops): Likewise.
3115 (c6x_reorg): Likewise for local "call_labels".
3116 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3118 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3119 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3120 the loops over LABEL_REFS.
3121 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3123 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3124 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3125 (split_branches): Strengthen local "olabel" from rtx to
3126 rtx_insn *, adding a safe_as_a cast.
3127 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3129 (add_insn_after): Likewise for first two params.
3130 (add_insn_before): Likewise.
3131 (remove_insn): Likewise for param.
3132 (emit_pattern_before_noloc): Likewise for second and third params.
3133 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3134 (emit_call_insn_before_noloc): Likewise.
3135 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3137 (emit_barrier_before): Likewise.
3138 (emit_label_before): Strengthen "label" param from "rtx" to
3139 "rtx_code_label *". Strengthen "before" param from "rtx" to
3141 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3143 (emit_pattern_after_noloc): Likewise.
3144 (emit_insn_after_noloc): Likewise.
3145 (emit_jump_insn_after_noloc): Likewise.
3146 (emit_call_insn_after_noloc): Likewise.
3147 (emit_debug_insn_after_noloc): Likewise.
3148 (emit_barrier_after): Likewise.
3149 (emit_label_after): Likewise for both params.
3150 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3151 "loc" param from "int" to "location_t".
3152 (emit_insn_after_setloc): Likewise.
3153 (emit_jump_insn_after_setloc): Likewise.
3154 (emit_call_insn_after_setloc): Likewise.
3155 (emit_debug_insn_after_setloc): Likewise.
3156 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3157 "loc" param from "int" to "location_t".
3158 (emit_pattern_before): Convert NULL_RTX to NULL.
3159 (emit_insn_before_setloc): Convert "loc" param from "int" to
3161 (emit_jump_insn_before_setloc): Likewise.
3162 (emit_call_insn_before_setloc): Likewise.
3163 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3164 rtx_insn *. Convert "loc" param from "int" to "location_t".
3165 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3166 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3167 Convert 3rd param from "int" to "location_t".
3168 (emit_barrier_before, emit_barrier_after, next_real_insn):
3169 Strengthen param from rtx to rtx_insn *.
3170 (emit_label_before): Strengthen 1st param from "rtx" to
3171 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3173 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3174 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3175 Strengthen 2nd param from "rtx" to "rtx_insn *".
3176 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3177 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3178 Likewise. Convert 3rd param from "int" to "location_t".
3179 (emit_label_after): Strengthen 1st param from "rtx" to
3181 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3183 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3184 from "rtx" to "rtx_insn *".
3186 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3188 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3189 bodies streamed in with -Q.
3190 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3191 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3192 * gimple-streamer-out.c: Include gimple-pretty-print.h
3193 (output_bb): Dump stmts streamed.
3194 * lto-section-out.c: Include print-tree.h
3195 (lto_begin_section): Dump sections created.
3196 (lto_output_decl_index): Dump decl encoded.
3197 * lto-streamer-out.c: Include print-tree.h
3198 (create_output_block): Dump output block created.
3199 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3200 (output_function): Dump function output.
3201 (output_constructor): Dump constructor streamed.
3202 (write_global_stream): Output indexes encoded.
3203 (produce_asm_for_decls): Dump streams encoded.
3204 * lto-streamer.c (streamer_dump_file): New global var.
3205 * lto-streamer.h (streamer_dump_file): Declare.
3206 * passes.c (ipa_write_summaries): Initialize streamer dump.
3207 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3210 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3213 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3214 offsets for register save directives. Emit a second batch of save
3215 directives, if need be, when the function accesses prior frames.
3217 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3219 * config/arc/fpu.md (fmasf4): Force operand to register.
3220 (fnmasf4): Likewise.
3222 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3224 * config/arc/arc-protos.h (arc_pad_return): Remove.
3225 * config/arc/arc.c (machine_function): Remove force_short_suffix
3227 (arc_print_operand): Adjust printing of '&'.
3228 (arc_verify_short): Remove conditional printing of short suffix.
3229 (arc_final_prescan_insn): Remove reference to size_reason.
3230 (pad_return): New function.
3231 (arc_reorg): Call pad_return.
3232 (arc_pad_return): Remove.
3233 (arc_init_machine_status): Remove reference to force_short_suffix.
3234 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3235 (attr length): When attribute iscompact is true force to 2
3236 regardless; in the case of maybe check if we want to force the
3237 instruction to have 4 bytes length.
3238 (nopv): Change it to generate 4 byte long nop as well.
3239 (blockage): New pattern.
3240 (simple_return): Remove call to arc_pad_return.
3241 (p_return_i): Likewise.
3243 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3245 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3247 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3249 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3252 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3254 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3255 for ARC700 and ARCv2.
3257 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3260 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3261 operands[2] instead of operands[1].
3264 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3266 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3267 case, check whether the outer register overlaps an unallocatable
3268 register, not just whether it fits the required class.
3270 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3272 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3274 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3275 (memrefs_conflict_p): Likewise.
3276 (init_alias_analysis): Likewise.
3277 * cfgexpand.c (expand_debug_expr): Likewise.
3278 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3279 * cse.c (fold_rtx): Likewise.
3280 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3281 * expr.c (emit_block_move_hints): Likewise.
3282 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3283 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3284 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3285 (emit_group_store): Likewise.
3286 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3287 to read the PRE/POST_MODIFY increment.
3288 * calls.c (store_one_arg): Use strip_offset.
3289 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3291 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3293 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3294 (simplify_binary_operation_1): Extend CONST_INT handling to
3296 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3297 than a HOST_WIDE_INT.
3298 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3300 (adjust_mems, add_stores): Update accodingly.
3301 (vt_canonicalize_addr): Track polynomial offsets.
3302 (emit_note_insn_var_location): Likewise.
3303 (vt_add_function_parameter): Likewise.
3304 (vt_initialize): Likewise.
3306 2018-06-12 Jeff Law <law@redhat.com>
3308 * config.gcc (alpha*-*-freebsd*): Remove.
3309 * config/alpha/freebsd.h: Remove.
3311 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3314 * spellcheck-tree.c (levenshtein_distance): Rename to...
3315 (get_edit_distance): ...this, and update for underlying renaming.
3316 * spellcheck-tree.h (levenshtein_distance): Rename to...
3317 (get_edit_distance): ...this.
3318 * spellcheck.c (levenshtein_distance): Rename to...
3319 (get_edit_distance): ...this. Convert from Levenshtein distance
3320 to Damerau-Levenshtein distance by supporting transpositions of
3321 adjacent characters. Rename "v1" to "v_next" and "v0" to
3323 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3324 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3325 update for underlying renaming.
3326 (selftest::levenshtein_distance_unit_test): Rename to...
3327 (selftest::test_get_edit_distance_unit): ...this, and update for
3328 underlying renaming.
3329 (selftest::test_find_closest_string): Add example from PR 69968
3330 where transposition helps
3331 (selftest::test_metric_conditions): Update for renaming.
3332 (selftest::test_metric_conditions): Likewise.
3333 (selftest::spellcheck_c_tests): Likewise.
3334 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3335 (get_edit_distance): ...this.
3336 (best_match::consider): Update for renaming.
3338 2018-06-12 Martin Sebor <msebor@redhat.com>
3340 PR tree-optimization/85259
3341 * builtins.c (compute_objsize): Handle constant offsets.
3342 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3343 true iff a warning has been issued.
3344 * gimple.h (gimple_nonartificial_location): New function.
3345 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3346 gimple_nonartificial_location and handle -Wno-system-headers.
3347 (handle_builtin_stxncpy): Same.
3349 2018-06-12 Martin Sebor <msebor@redhat.com>
3352 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3354 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3356 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3357 BUILTIN_VEC_XST entries for pointer to double and long long.
3359 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3362 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3364 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3367 2018-06-12 Martin Liska <mliska@suse.cz>
3369 * doc/options.texi: Document IntegerRange.
3371 2018-06-12 Martin Liska <mliska@suse.cz>
3373 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3374 * opt-functions.awk: Handle Deprecated flag.
3375 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3377 (read_cmdline_option): Report warning for a deprecated option.
3378 * opts.h (struct cl_option): Add new field cl_deprecated.
3379 (CL_ERR_DEPRECATED): New.
3381 2018-06-12 Martin Liska <mliska@suse.cz>
3383 * doc/options.texi: Document Deprecated option flag.
3385 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3387 * config/arc/arc-arch.h (arc_extras): New enum.
3388 (arc_cpu_t):Add field extra.
3389 (arc_cpu_types): Consider the extras.
3390 * config/arc/arc-cpus.def: Add extras info.
3391 * config/arc/arc-opts.h (processor_type): Consider extra field.
3392 * config/arc/arc.c (arc_override_options): Handle extra field.
3394 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3396 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3397 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3399 * config/arc/arc.c (arc_init): Update pic warning.
3400 (irq_range): Update irq range parsing warnings.
3401 (arc_override_options): Update various warning messages.
3402 (arc_handle_aux_attribute): Likewise.
3404 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3406 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3408 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3410 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3411 compiler messages directives.
3413 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3415 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3416 * config/mips/mips-tables.opt: Regenerate.
3417 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3419 * doc/invoke.texi: Document -march=i6500.
3421 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3423 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3424 (i6400_gpmul): Add cpu_unit.
3425 (i6400_gpdiv): Likewise.
3426 (i6400_msa_add_d): Update reservations.
3427 (i6400_msa_int_add) Likewise.
3428 (i6400_msa_short_logic3) Likewise.
3429 (i6400_msa_short_logic2) Likewise.
3430 (i6400_msa_short_logic) Likewise.
3431 (i6400_msa_move) Likewise.
3432 (i6400_msa_cmp) Likewise.
3433 (i6400_msa_short_float2) Likewise.
3434 (i6400_msa_div_d) Likewise.
3435 (i6400_msa_long_logic1) Likewise.
3436 (i6400_msa_long_logic2) Likewise.
3437 (i6400_msa_mult) Likewise.
3438 (i6400_msa_long_float2) Likewise.
3439 (i6400_msa_long_float4) Likewise.
3440 (i6400_msa_long_float5) Likewise.
3441 (i6400_msa_long_float8) Likewise.
3442 (i6400_fpu_fadd): Include frint type.
3443 (i6400_fpu_store): New define_insn_reservation.
3444 (i6400_fpu_load): Likewise.
3445 (i6400_fpu_move): Likewise.
3446 (i6400_fpu_fcmp): Likewise.
3447 (i6400_fpu_fmadd): Likewise.
3448 (i6400_int_mult): Include imul3nc type and update reservation.
3449 (i6400_int_div): Include idiv3 type and update reservation.
3450 (i6400_int_load): Update to check type not move_type.
3451 (i6400_int_store): Likewise.
3452 (i6400_int_prefetch): Set zero latency.
3454 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3456 * gcc.c: Document new %@{...} sequence.
3457 (LINK_COMMAND_SPEC): Use it for the -L switches.
3458 (cpp_unique_options): Use it for the -I switches.
3459 (at_file_argbuf): New global variable.
3460 (in_at_file): Likewise.
3461 (alloc_args): Create at_file_argbuf.
3462 (clear_args): Truncate at_file_argbuf.
3463 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3464 (open_at_file): New function.
3465 (close_at_file): Likewise.
3466 (create_at_file): Delete.
3467 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3470 (validate_switches_from_spec): Deal with %@{...} sequence.
3471 (validate_switches): Likewise.
3472 (driver::finalize): Call clear_args.
3474 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3476 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3478 2018-06-11 Martin Sebor <msebor@redhat.com>
3480 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3481 (Wno-class-memaccess): Add @opindex.
3482 (Wno-templates, Wno-multiple-inheritance): Same.
3483 (Wno-virtual-inheritance, Wno-namespaces): Same.
3484 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3485 (Wno-format-overflow, Wno-format-truncation): Same.
3486 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3487 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3488 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3489 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3490 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3491 (Wno-misspelled-isr): Same.
3493 2018-06-11 Martin Sebor <msebor@redhat.com>
3495 * PR tree-optimization/86083
3496 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3498 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3500 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3502 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3505 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3506 on the correct operand.
3507 (*movdi_internal64): Ditto.
3509 2018-06-11 Martin Liska <mliska@suse.cz>
3511 PR tree-optimization/86089
3512 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3514 2018-06-11 Julia Koval <julia.koval@intel.com>
3516 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3517 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3518 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3520 2018-06-11 Olivier Hainque <hainque@adacore.com>
3522 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3523 for Ada with strict dwarf2.
3525 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3528 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3531 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3533 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3535 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3537 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3540 2018-06-08 Martin Liska <mliska@suse.cz>
3542 * tree-cfg.h (debug_function): Fix argument type to match
3545 2018-06-08 Martin Liska <mliska@suse.cz>
3547 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3548 Remove usage of MPX-related (and removed) fields.
3549 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3551 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3553 * cfg.c (debug): Use TDF_NONE rather than 0.
3554 * cfghooks.c (debug): Likewise.
3555 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3556 (struct dump_option_value_info): Convert to...
3557 (struct kv_pair): ...this template type.
3558 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3560 (optinfo_verbosity_options): Likewise.
3561 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3563 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3564 than int for "optgroup_flags" param.
3565 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3567 (dump_dec): Likewise.
3568 (dump_finish): Use TDF_NONE rather than 0.
3569 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3570 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3571 than 0. Update for change to option_ptr.
3572 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3573 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3574 0. Update for changes to optinfo_verbosity_options and
3576 (opt_info_switch_p): Convert optgroup_flags from int to
3578 (dump_basic_block): Use dump_flags_t rather than int
3579 for "dump_kind" param.
3580 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3581 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3582 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3583 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3584 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3585 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3586 TDF_NONE): Convert from macros to...
3587 (enum dump_flag): ...this new enum.
3588 (dump_flags_t): Update to use enum.
3589 (operator|, operator&, operator~, operator|=, operator&=):
3590 Implement for dump_flags_t.
3591 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3592 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3593 Convert from macros to...
3594 (enum optgroup_flag): ...this new enum.
3595 (optgroup_flags_t): New typedef.
3596 (operator|, operator|=): Implement for optgroup_flags_t.
3597 (struct dump_file_info): Convert field "alt_flags" to
3598 dump_flags_t. Convert field "optgroup_flags" to
3600 (dump_basic_block): Use dump_flags_t rather than int for param.
3601 (dump_generic_expr_loc): Likewise.
3602 (dump_dec): Likewise.
3603 (dump_register): Convert param "optgroup_flags" to
3605 (opt_info_enable_passes): Likewise.
3606 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3608 * gimple-pretty-print.c (debug): Likewise.
3609 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3610 (merged_store_group::apply_stores): Likewise.
3611 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3612 * gimple.c (verify_gimple_pp): Likewise.
3613 * graphite-poly.c (print_pbb_body): Likewise.
3614 * passes.c (pass_manager::register_one_dump_file): Convert
3615 local "optgroup_flags" to optgroup_flags_t.
3616 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3618 (debug_body): Likewise.
3619 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3620 to optgroup_flags_t.
3621 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3623 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3624 (convert_mult_to_fma): Likewise.
3625 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3626 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3627 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3628 "dump_kind" to dump_flags_t.
3630 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3632 * config/rs6000/rs6000.c (min, max): Delete.
3634 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3636 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3637 -mabi=spe and -mabi=no-spe.
3639 2018-06-08 Martin Liska <mliska@suse.cz>
3641 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3642 where we expect an existing summary.
3644 2018-06-08 Martin Liska <mliska@suse.cz>
3646 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3647 * ipa-inline.h (estimate_edge_growth): Likewise.
3649 2018-06-08 Martin Liska <mliska@suse.cz>
3651 * cgraph.c (function_version_hasher::hash): Use
3652 cgraph_node::get_uid ().
3653 (function_version_hasher::equal):
3654 * cgraph.h (cgraph_node::get_uid): New method.
3655 * ipa-inline.c (update_caller_keys): Use
3656 cgraph_node::get_uid ().
3657 (update_callee_keys): Likewise.
3658 * ipa-utils.c (searchc): Likewise.
3659 (ipa_reduced_postorder): Likewise.
3660 * lto-cgraph.c (input_node): Likewise.
3661 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3662 * symbol-summary.h (symtab_insertion): Likewise.
3663 (symtab_removal): Likewise.
3664 (symtab_duplication): Likewise.
3665 * tree-pretty-print.c (dump_function_header): Likewise.
3666 * tree-sra.c (convert_callers_for_node): Likewise.
3668 2018-06-08 Martin Liska <mliska@suse.cz>
3670 * cgraph.c (symbol_table::create_edge): Always assign a new
3672 (symbol_table::free_edge): Do not recycle numbers.
3673 * cgraph.h (cgraph_edge::get): New method.
3674 * symbol-summary.h (symtab_removal): Use it.
3675 (symtab_duplication): Likewise.
3676 (call_summary::hashable_uid): Remove.
3678 2018-06-08 Martin Liska <mliska@suse.cz>
3680 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3681 (initialize_growth_caches): Remove.
3682 (free_growth_caches): Likewise.
3683 (do_estimate_edge_time): Use edge_growth_cache.
3684 (do_estimate_edge_size): Likewise.
3685 (do_estimate_edge_hints): Likewise.
3686 * ipa-inline.c (reset_edge_caches): Likewise.
3687 (recursive_inlining): Likewise.
3688 (inline_small_functions): Likewise.
3689 * ipa-inline.h (initialize_growth_caches): Remove.
3690 (estimate_edge_size): Likewise.
3691 (estimate_edge_time): Likewise.
3692 (estimate_edge_hints): Likewise.
3693 (reset_edge_growth_cache): Likewise.
3694 * symbol-summary.h (call_summary::remove): New method.
3696 2018-06-08 Martin Liska <mliska@suse.cz>
3698 * ipa-cp.c (class edge_clone_summary): New summary.
3699 (grow_edge_clone_vectors): Remove.
3700 (ipcp_edge_duplication_hook): Remove.
3701 (class edge_clone_summary_t): New call_summary class.
3702 (ipcp_edge_removal_hook): Remove.
3703 (edge_clone_summary_t::duplicate): New function.
3704 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3705 (create_specialized_node): Likewise.
3706 (ipcp_driver): Initialize edge_clone_summaries and do not
3709 2018-06-08 Martin Liska <mliska@suse.cz>
3711 * symbol-summary.h (get): New function.
3712 (call_summary::m_initialize_when_cloning): New class member.
3714 2018-06-08 Martin Liska <mliska@suse.cz>
3716 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3717 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3718 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3719 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3720 (remove_cgraph_node_from_order): Use the removed_nodes set.
3721 (do_per_function_toporder): Likwise.
3722 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3723 instead of summary_uid.
3724 (symtab_removal): Likewise.
3725 (symtab_duplication): Likewise.
3727 2018-06-08 Martin Liska <mliska@suse.cz>
3729 * ipa-cp.c (ipcp_store_bits_results): Use
3730 ipcp_transformation_sum.
3731 (ipcp_store_vr_results): Likewise.
3732 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3734 (ipcp_transformation_initialize): ... this.
3735 (ipa_set_node_agg_value_chain):
3736 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3737 (write_ipcp_transformation_info): Likewise.
3738 (read_ipcp_transformation_info): Likewise.
3739 (ipcp_update_bits): Likewise.
3740 (ipcp_update_vr): Likewise.
3741 (ipcp_transform_function): Likewise.
3742 * ipa-prop.h: Rename ipcp_transformation_summary to
3743 ipcp_transformation.
3744 (class ipcp_transformation_t): New function summary.
3745 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3746 (ipa_get_agg_replacements_for_node): Likewise.
3748 2018-06-08 Martin Liska <mliska@suse.cz>
3750 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3752 (class funct_state_summary_t): New function_summary class.
3753 (has_function_state): Remove.
3754 (get_function_state): Likewise.
3755 (set_function_state): Likewise.
3756 (add_new_function): Likewise.
3757 (funct_state_summary_t::insert): New function.
3758 (duplicate_node_data): Remove.
3759 (remove_node_data): Remove.
3760 (funct_state_summary_t::duplicate): New function.
3761 (register_hooks): Create new funct_state_summaries.
3762 (pure_const_generate_summary): Use it.
3763 (pure_const_write_summary): Likewise.
3764 (pure_const_read_summary): Likewise.
3765 (propagate_pure_const): Likewise.
3766 (propagate_nothrow): Likewise.
3767 (dump_malloc_lattice): Likewise.
3768 (propagate_malloc): Likewise.
3769 (execute): Do not register hooks, just remove summary
3771 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3774 2018-06-08 Martin Liska <mliska@suse.cz>
3776 * ipa-reference.c (remove_node_data): Remove.
3777 (duplicate_node_data): Likewise.
3778 (class ipa_ref_var_info_summary_t): New class.
3779 (class ipa_ref_opt_summary_t): Likewise.
3780 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3781 (get_reference_optimization_summary): Use
3782 ipa_ref_opt_sum_summaries.
3783 (set_reference_vars_info): Remove.
3784 (set_reference_optimization_summary): Likewise.
3785 (ipa_init): Create summaries.
3786 (init_function_info): Use function summary.
3787 (ipa_ref_opt_summary_t::duplicate): New function.
3788 (ipa_ref_opt_summary_t::remove): New function.
3789 (get_read_write_all_from_node): Fix GNU coding style.
3790 (propagate): Use function summary.
3791 (write_node_summary_p): Fix GNU coding style.
3792 (stream_out_bitmap): Likewise.
3793 (ipa_reference_read_optimization_summary): Use function summary.
3794 (ipa_reference_c_finalize): Do not release hooks.
3796 2018-06-08 Martin Liska <mliska@suse.cz>
3798 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3799 (analyze_function_body): Extract multiple calls of get_create.
3800 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3801 * ipa-inline.c (recursive_inlining): Use ::get method.
3802 * ipa-inline.h (estimate_edge_growth): Likewise.
3804 2018-06-08 Martin Liska <mliska@suse.cz>
3806 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3808 (hsa_function_summary::hsa_function_summary): Use the new enum
3810 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3811 * hsa-gen.c (hsa_get_host_function): Likewise.
3812 (get_brig_function_name): Likewise.
3813 * ipa-hsa.c (process_hsa_functions): Likewise.
3814 (ipa_hsa_write_summary): Likewise.
3815 * symbol-summary.h (symtab_duplication): Use ::get function/
3816 (get): New function.
3818 2018-06-08 Martin Liska <mliska@suse.cz>
3820 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3822 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3823 (hsa_register_kernel): Likewise.
3824 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3825 * hsa-gen.c (hsa_get_host_function): Likewise.
3826 (get_brig_function_name): Likewise.
3827 (generate_hsa): Likewise.
3828 (pass_gen_hsail::execute): Likewise.
3829 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3830 (devirtualization_time_bonus): Likewise.
3831 (ipcp_propagate_stage): Likewise.
3832 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3833 (edge_set_predicate): Likewise.
3834 (evaluate_conditions_for_known_args): Likewise.
3835 (evaluate_properties_for_edge): Likewise.
3836 (ipa_fn_summary::reset): Likewise.
3837 (ipa_fn_summary_t::duplicate): Likewise.
3838 (dump_ipa_call_summary): Likewise.
3839 (ipa_dump_fn_summary): Likewise.
3840 (analyze_function_body): Likewise.
3841 (compute_fn_summary): Likewise.
3842 (estimate_edge_devirt_benefit): Likewise.
3843 (estimate_edge_size_and_time): Likewise.
3844 (estimate_calls_size_and_time): Likewise.
3845 (estimate_node_size_and_time): Likewise.
3846 (inline_update_callee_summaries): Likewise.
3847 (remap_edge_change_prob): Likewise.
3848 (remap_edge_summaries): Likewise.
3849 (ipa_merge_fn_summary_after_inlining): Likewise.
3850 (ipa_update_overall_fn_summary): Likewise.
3851 (read_ipa_call_summary): Likewise.
3852 (inline_read_section): Likewise.
3853 (write_ipa_call_summary): Likewise.
3854 (ipa_fn_summary_write): Likewise.
3855 (ipa_free_fn_summary): Likewise.
3856 * ipa-hsa.c (process_hsa_functions): Likewise.
3857 (ipa_hsa_write_summary): Likewise.
3858 (ipa_hsa_read_section): Likewise.
3859 * ipa-icf.c (sem_function::merge): Likewise.
3860 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3861 (do_estimate_edge_time): Likewise.
3862 (estimate_size_after_inlining): Likewise.
3863 (estimate_growth): Likewise.
3864 (growth_likely_positive): Likewise.
3865 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3866 (inline_call): Likewise.
3867 * ipa-inline.c (caller_growth_limits): Likewise.
3868 (can_inline_edge_p): Likewise.
3869 (can_inline_edge_by_limits_p): Likewise.
3870 (compute_uninlined_call_time): Likewise.
3871 (compute_inlined_call_time): Likewise.
3872 (want_inline_small_function_p): Likewise.
3873 (edge_badness): Likewise.
3874 (update_caller_keys): Likewise.
3875 (update_callee_keys): Likewise.
3876 (recursive_inlining): Likewise.
3877 (inline_small_functions): Likewise.
3878 (inline_to_all_callers_1): Likewise.
3879 (dump_overall_stats): Likewise.
3880 (early_inline_small_functions): Likewise.
3881 (early_inliner): Likewise.
3882 * ipa-inline.h (estimate_edge_growth): Likewise.
3883 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3884 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3885 * ipa-prop.h (IPA_NODE_REF): Likewise.
3886 (IPA_EDGE_REF): Likewise.
3887 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3888 (propagate_malloc): Likewise.
3889 * ipa-split.c (execute_split_functions): Likewise.
3890 * symbol-summary.h: Rename get to get_create.
3892 (get_create): Likewise.
3893 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3895 2018-06-08 Martin Liska <mliska@suse.cz>
3897 * symbol-summary.h (release): Move definition out of class
3899 (symtab_removal): Likewise.
3900 (symtab_duplication): Likewise.
3902 2018-06-08 Martin Liska <mliska@suse.cz>
3904 * symbol-summary.h (function_summary): Move constructor
3905 implementation out of class declaration.
3906 (release): Likewise.
3907 (symtab_insertion): Likewise.
3908 (symtab_removal): Likewise.
3909 (symtab_duplication): Likewise.
3912 2018-06-08 Martin Liska <mliska@suse.cz>
3914 * Makefile.in: Remove support for MPX (macros, related functions,
3915 fields in cgraph_node, ...).
3916 * builtin-types.def (BT_BND): Likewise.
3917 (BT_FN_BND_CONST_PTR): Likewise.
3918 (BT_FN_CONST_PTR_BND): Likewise.
3919 (BT_FN_VOID_PTR_BND): Likewise.
3920 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3921 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3922 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3923 (expand_builtin_mempcpy_with_bounds): Likewise.
3924 (expand_builtin_memset_with_bounds): Likewise.
3925 (expand_builtin_memset_args): Likewise.
3926 (std_expand_builtin_va_start): Likewise.
3927 (expand_builtin): Likewise.
3928 (expand_builtin_with_bounds): Likewise.
3929 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3930 (DEF_LIB_BUILTIN_CHKP): Likewise.
3931 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3932 (DEF_CHKP_BUILTIN): Likewise.
3933 (BUILT_IN_MEMCPY): Likewise.
3934 (BUILT_IN_MEMMOVE): Likewise.
3935 (BUILT_IN_MEMPCPY): Likewise.
3936 (BUILT_IN_MEMSET): Likewise.
3937 (BUILT_IN_STPCPY): Likewise.
3938 (BUILT_IN_STRCAT): Likewise.
3939 (BUILT_IN_STRCHR): Likewise.
3940 (BUILT_IN_STRCPY): Likewise.
3941 (BUILT_IN_STRLEN): Likewise.
3942 (BUILT_IN_MEMCPY_CHK): Likewise.
3943 (BUILT_IN_MEMMOVE_CHK): Likewise.
3944 (BUILT_IN_MEMPCPY_CHK): Likewise.
3945 (BUILT_IN_MEMSET_CHK): Likewise.
3946 (BUILT_IN_STPCPY_CHK): Likewise.
3947 (BUILT_IN_STRCAT_CHK): Likewise.
3948 (BUILT_IN_STRCPY_CHK): Likewise.
3949 * calls.c (store_bounds): Likewise.
3950 (emit_call_1): Likewise.
3951 (special_function_p): Likewise.
3952 (maybe_warn_nonstring_arg): Likewise.
3953 (initialize_argument_information): Likewise.
3954 (finalize_must_preallocate): Likewise.
3955 (compute_argument_addresses): Likewise.
3956 (expand_call): Likewise.
3957 * cfgexpand.c (expand_call_stmt): Likewise.
3958 (expand_return): Likewise.
3959 (expand_gimple_stmt_1): Likewise.
3960 (pass_expand::execute): Likewise.
3961 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3962 (cgraph_node::remove): Likewise.
3963 (cgraph_node::dump): Likewise.
3964 (cgraph_node::verify_node): Likewise.
3965 * cgraph.h (chkp_function_instrumented_p): Likewise.
3966 (symtab_node::get_alias_target): Likewise.
3967 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3968 (cgraph_local_p): Likewise.
3969 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3970 (cgraph_edge::rebuild_references): Likewise.
3971 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3972 (walk_polymorphic_call_targets): Likewise.
3973 (cgraph_node::expand_thunk): Likewise.
3974 (symbol_table::output_weakrefs): Likewise.
3975 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3976 (ix86_handle_option): Likewise.
3977 * config/i386/constraints.md: Likewise.
3978 * config/i386/i386-builtin-types.def (BND): Likewise.
3982 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3983 (BDESC_FIRST): Likewise.
3985 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3986 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3987 * config/i386/i386.c (enum reg_class): Likewise.
3988 (ix86_target_string): Likewise.
3989 (ix86_option_override_internal): Likewise.
3990 (ix86_conditional_register_usage): Likewise.
3991 (ix86_valid_target_attribute_inner_p): Likewise.
3992 (ix86_set_indirect_branch_type): Likewise.
3993 (ix86_set_current_function): Likewise.
3994 (ix86_function_arg_regno_p): Likewise.
3995 (init_cumulative_args): Likewise.
3996 (ix86_function_arg_advance): Likewise.
3997 (ix86_function_arg): Likewise.
3998 (ix86_pass_by_reference): Likewise.
3999 (ix86_function_value_regno_p): Likewise.
4000 (ix86_function_value_1): Likewise.
4001 (ix86_function_value_bounds): Likewise.
4002 (ix86_return_in_memory): Likewise.
4003 (ix86_setup_incoming_vararg_bounds): Likewise.
4004 (ix86_va_start): Likewise.
4005 (indirect_thunk_need_prefix): Likewise.
4006 (print_reg): Likewise.
4007 (ix86_print_operand): Likewise.
4008 (ix86_expand_call): Likewise.
4009 (ix86_output_function_return): Likewise.
4010 (reg_encoded_number): Likewise.
4011 (BDESC_VERIFYS): Likewise.
4012 (ix86_init_mpx_builtins): Likewise.
4013 (ix86_init_builtins): Likewise.
4014 (ix86_emit_cmove): Likewise.
4015 (ix86_emit_move_max): Likewise.
4016 (ix86_expand_builtin): Likewise.
4017 (ix86_builtin_mpx_function): Likewise.
4018 (ix86_get_arg_address_for_bt): Likewise.
4019 (ix86_load_bounds): Likewise.
4020 (ix86_store_bounds): Likewise.
4021 (ix86_load_returned_bounds): Likewise.
4022 (ix86_store_returned_bounds): Likewise.
4023 (ix86_class_likely_spilled_p): Likewise.
4024 (ix86_hard_regno_mode_ok): Likewise.
4025 (x86_order_regs_for_local_alloc): Likewise.
4026 (ix86_mitigate_rop): Likewise.
4027 (ix86_bnd_prefixed_insn_p): Likewise.
4028 (ix86_mpx_bound_mode): Likewise.
4029 (ix86_make_bounds_constant): Likewise.
4030 (ix86_initialize_bounds): Likewise.
4031 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4032 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4033 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4034 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4035 (TARGET_CHKP_BOUND_MODE): Likewise.
4036 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4037 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4038 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4039 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4040 * config/i386/i386.h (TARGET_MPX): Likewise.
4041 (TARGET_MPX_P): Likewise.
4042 (VALID_BND_REG_MODE): Likewise.
4043 (FIRST_BND_REG): Likewise.
4044 (LAST_BND_REG): Likewise.
4045 (enum reg_class): Likewise.
4046 (BND_REG_P): Likewise.
4047 (BND_REGNO_P): Likewise.
4048 (BNDmode): Likewise.
4049 (ADJUST_INSN_LENGTH): Likewise.
4050 * config/i386/i386.md: Likewise.
4051 * config/i386/i386.opt: Likewise.
4052 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4053 (defined): Likewise.
4054 (LINK_MPX): Likewise.
4055 (MPX_SPEC): Likewise.
4056 (LIBMPX_SPEC): Likewise.
4057 (LIBMPXWRAPPERS_SPEC): Likewise.
4058 (CHKP_SPEC): Likewise.
4059 * config/i386/predicates.md: Likewise.
4060 * dbxout.c (dbxout_type): Likewise.
4061 * doc/extend.texi: Likewise.
4062 * doc/invoke.texi: Likewise.
4063 * doc/md.texi: Likewise.
4064 * doc/tm.texi: Likewise.
4065 * doc/tm.texi.in: Likewise.
4066 * dwarf2out.c (is_base_type): Likewise.
4067 (gen_formal_types_die): Likewise.
4068 (gen_subprogram_die): Likewise.
4069 (gen_type_die_with_usage): Likewise.
4070 (gen_decl_die): Likewise.
4071 (dwarf2out_late_global_decl): Likewise.
4072 * expr.c (expand_assignment): Likewise.
4073 (emit_storent_insn): Likewise.
4074 (store_expr_with_bounds): Likewise.
4075 (store_expr): Likewise.
4076 (expand_expr_real_1): Likewise.
4077 * expr.h (store_expr_with_bounds): Likewise.
4078 * function.c (use_register_for_decl): Likewise.
4079 (struct bounds_parm_data): Likewise.
4080 (assign_parms_augmented_arg_list): Likewise.
4081 (assign_parm_find_entry_rtl): Likewise.
4082 (assign_parm_is_stack_parm): Likewise.
4083 (assign_parm_load_bounds): Likewise.
4084 (assign_bounds): Likewise.
4085 (assign_parms): Likewise.
4086 (expand_function_start): Likewise.
4087 * gcc.c (CHKP_SPEC): Likewise.
4088 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4089 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4090 (wrestrict_dom_walker::check_call): Likewise.
4091 * gimple.c (gimple_build_call_from_tree): Likewise.
4092 * gimple.h (enum gf_mask): Likewise.
4093 (gimple_call_with_bounds_p): Likewise.
4094 (gimple_call_set_with_bounds): Likewise.
4095 * gimplify.c (gimplify_init_constructor): Likewise.
4096 * ipa-cp.c (initialize_node_lattices): Likewise.
4097 (propagate_constants_across_call): Likewise.
4098 (find_more_scalar_values_for_callers_subset): Likewise.
4099 * ipa-hsa.c (process_hsa_functions): Likewise.
4100 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4101 * ipa-icf.c (sem_function::merge): Likewise.
4102 * ipa-inline.c (early_inliner): Likewise.
4103 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4104 (warn_function_cold): Likewise.
4105 (propagate_pure_const): Likewise.
4106 * ipa-ref.h (enum GTY): Likewise.
4107 * ipa-split.c (find_retbnd): Likewise.
4108 (consider_split): Likewise.
4109 (split_function): Likewise.
4110 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4111 * ipa.c (walk_polymorphic_call_targets): Likewise.
4112 (symbol_table::remove_unreachable_nodes): Likewise.
4113 (process_references): Likewise.
4114 (cgraph_build_static_cdtor_1): Likewise.
4115 * lto-cgraph.c (lto_output_node): Likewise.
4116 (output_refs): Likewise.
4117 (compute_ltrans_boundary): Likewise.
4118 (input_overwrite_node): Likewise.
4119 (input_node): Likewise.
4120 (input_cgraph_1): Likewise.
4121 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4122 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4123 (class pass_chkp_instrumentation_passes): Likewise.
4124 (make_pass_chkp_instrumentation_passes): Likewise.
4125 * passes.def: Likewise.
4126 * rtl.h (struct GTY): Likewise.
4127 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4128 * stor-layout.c (layout_type): Likewise.
4129 * symtab.c: Likewise.
4130 * target.def: Likewise.
4131 * targhooks.c (default_chkp_bound_type): Likewise.
4132 (default_chkp_bound_mode): Likewise.
4133 (default_builtin_chkp_function): Likewise.
4134 (default_chkp_function_value_bounds): Likewise.
4135 (default_chkp_make_bounds_constant): Likewise.
4136 (default_chkp_initialize_bounds): Likewise.
4137 * targhooks.h (default_chkp_bound_type): Likewise.
4138 (default_chkp_bound_mode): Likewise.
4139 (default_builtin_chkp_function): Likewise.
4140 (default_chkp_function_value_bounds): Likewise.
4141 (default_chkp_make_bounds_constant): Likewise.
4142 (default_chkp_initialize_bounds): Likewise.
4143 * toplev.c (compile_file): Likewise.
4144 (process_options): Likewise.
4145 * tree-core.h (DEF_BUILTIN): Likewise.
4146 (DEF_BUILTIN_CHKP): Likewise.
4147 * tree-inline.c (declare_return_variable): Likewise.
4148 (remap_gimple_stmt): Likewise.
4149 (copy_bb): Likewise.
4150 (initialize_inlined_parameters): Likewise.
4151 (expand_call_inline): Likewise.
4152 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4153 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4154 (make_pass_ipa_chkp_produce_thunks): Likewise.
4155 (make_pass_chkp): Likewise.
4156 (make_pass_chkp_opt): Likewise.
4157 (make_pass_chkp_instrumentation_passes): Likewise.
4158 * tree-pretty-print.c (dump_generic_node): Likewise.
4159 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4160 * tree-ssa-dce.c (propagate_necessity): Likewise.
4161 (eliminate_unnecessary_stmts): Likewise.
4162 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4163 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4164 * tree-ssa-sccvn.h: Likewise.
4165 * tree-ssa-strlen.c (get_string_length): Likewise.
4166 (valid_builtin_call): Likewise.
4167 (adjust_last_stmt): Likewise.
4168 (handle_builtin_strchr): Likewise.
4169 (handle_builtin_strcpy): Likewise.
4170 (handle_builtin_stxncpy): Likewise.
4171 (handle_builtin_memcpy): Likewise.
4172 (handle_builtin_strcat): Likewise.
4173 (strlen_check_and_optimize_stmt): Likewise.
4174 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4175 * tree-streamer-in.c: Likewise.
4176 * tree-streamer.c (record_common_node): Likewise.
4177 * tree.c (tree_code_size): Likewise.
4178 (wide_int_to_tree_1): Likewise.
4179 (type_contains_placeholder_1): Likewise.
4180 (build_common_tree_nodes): Likewise.
4181 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4182 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4183 (POINTER_BOUNDS_P): Likewise.
4184 (BOUNDED_TYPE_P): Likewise.
4185 (BOUNDED_P): Likewise.
4186 (CALL_WITH_BOUNDS_P): Likewise.
4187 (pointer_bounds_type_node): Likewise.
4188 * value-prof.c (gimple_ic): Likewise.
4189 * var-tracking.c (vt_add_function_parameters): Likewise.
4190 * varasm.c (make_decl_rtl): Likewise.
4191 (assemble_start_function): Likewise.
4192 (output_constant): Likewise.
4193 (maybe_assemble_visibility): Likewise.
4194 * varpool.c (ctor_for_folding): Likewise.
4195 * chkp-builtins.def: Remove.
4196 * ipa-chkp.c: Remove.
4197 * ipa-chkp.h: Remove.
4198 * rtl-chkp.c: Remove.
4199 * rtl-chkp.h: Remove.
4200 * tree-chkp-opt.c: Remove.
4201 * tree-chkp.c: Remove.
4202 * tree-chkp.h: Remove.
4204 2018-06-07 Carl Love <cel@us.ibm.com>
4206 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4207 vextract_fp_from_shortl): Add BE support.
4209 2018-06-07 Paul Koning <ni1d@arrl.net>
4211 * compare-elim.c (try_merge_compare): Don't merge compare if
4212 address contains a side effect.
4213 (try_eliminate_compare): Likewise.
4215 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4217 * config.gcc: Support "tremont".
4218 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4219 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4221 * config/i386/i386.c (m_TREMONT): Define.
4222 (processor_target_table): Add "tremont".
4223 (PTA_TREMONT): Define.
4224 (ix86_lea_outperforms): Add TARGET_TREMONT.
4225 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4226 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4227 and M_INTEL_GOLDMONT_PLUS.
4228 (fold_builtin_cpu): Add "tremont".
4229 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4230 (ix86_option_override_internal): Add "tremont".
4231 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4232 (processor_type): Add PROCESSOR_TREMONT.
4233 * config/i386/x86-tune.def: Add m_TREMONT.
4234 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4236 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4238 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4239 symbol defined for msp430i* devices to be lower case.
4241 gcc/testsuite/gcc.target/msp430/
4242 * msp430i-device-symbol.c: New test.
4243 * msp430f-device-symbol.c: New test.
4244 * msp430.h: New test header file.
4246 2018-06-07 Richard Biener <rguenther@suse.de>
4248 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4249 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4250 Properly wrap signed arithmetic if overflow wraps.
4252 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4254 PR tree-optimization/69615
4255 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4256 of a cast from a same precision integral SSA_NAME in a bb dominated
4257 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4258 cast to utype if rhs2 has already a compatible type.
4260 2018-06-07 Richard Biener <rguenther@suse.de>
4262 PR tree-optimization/85935
4263 * graphite-scop-detection.c (find_params_in_bb): Analyze
4264 condition operands with respect to the correct loop. Assert
4265 the analysis doesn't fail.
4267 2018-06-04 Carl Love <cel@us.ibm.com>
4269 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4270 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4271 as it is slightly cheaper.
4272 (first_match_or_eos_index_<mode>):
4273 Calculate index using natural element order.
4274 (first_match_index_<mode>):
4275 Calculate index using natural element order.
4276 (first_match_or_eos_index_<mode>):
4277 Calculate index using natural order.
4278 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4279 for BE and LE modes.
4280 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4281 P9V_BUILTIN_VCLZLSBB_V16QI.
4282 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4285 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4287 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4288 indentation and line wrap for many prototypes. Add missing
4289 @smallexample directives around block of prototypes for vec_xl and
4292 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4294 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4295 track if we pass or return IEEE 128-bit floating point.
4296 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4297 C++ mangling that is compatible with GCC 8.1.
4298 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4299 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4300 floating point types.
4301 (rs6000_function_arg_advance_1): Likewise.
4302 (rs6000_mangle_type): Optionally generate mangled names that match
4303 what GCC 8.1 generated for IEEE 128-bit floating point types.
4304 (rs6000_globalize_decl_name): If we have an external function that
4305 passes or returns IEEE 128-bit types, generate a weak reference
4306 from the mangled name used in GCC 8.1 to the current mangled
4308 (rs6000_init_builtins): Make __ibm128 use the long double type if
4309 long double is IBM extended double. Make __float128 use the long
4310 double type if long double is IEEE 128-bit.
4313 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4314 macro for __ibm128 built-in functions.
4315 (PACK_IF): Add __ibm128 pack/unpack functions.
4316 (UNPACK_IF): Likewise.
4317 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4318 enable long double built-in functions if long double is IEEE
4319 128-bit floating point.
4320 (rs6000_invalid_builtin): Update long double built-in function
4322 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4323 functions, adjust the built-in function to use the long double
4324 built-in function if __ibm128 and long double are the same type.
4325 * doc/extend.texi (PowerPC builtins): Update documention for
4326 __builtin_{,un}pack_longdouble. Add documentation for
4327 __builtin_{,un}pack_ibm128.
4329 2018-06-06 Jim Wilson <jimw@sifive.com>
4331 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4332 (struct machine_function): New field interrupt_mode.
4333 (riscv_handle_type_attribute): New function. Add forward declaration.
4334 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4335 (riscv_expand_epilogue): Check interrupt_mode field.
4336 (riscv_set_current_function): Check interrupt attribute args and
4337 set interrupt_mode field.
4338 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4339 (riscv_sret, riscv_uret): New.
4340 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4341 new arguments to interrupt attribute.
4343 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4346 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4347 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4349 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4351 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4352 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4353 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4354 several redundant entries.
4356 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4358 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4359 type from "rtx" to "rtx_insn *".
4360 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4361 for local "call_insn", removing cast.
4362 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4364 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4366 PR tree-optimization/86066
4367 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4368 for BIT_INSERT_EXPR stores.
4370 2018-06-06 Richard Biener <rguenther@suse.de>
4372 PR tree-optimization/86062
4373 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4374 component refs ontop
4375 of to be offsetted base.
4377 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4379 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4380 to be static and remove check on interrupt attribute name.
4382 gcc/testsuite/gcc.target/msp430/
4383 * function-attributes-4.c: New test.
4384 * static-interrupts.c: New test.
4386 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4388 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4389 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4391 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4394 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4396 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4397 Remove second argument, change how error is called.
4398 (aarch64_layout_arg): Remove second argument from
4399 aarch64_err_no_fpadvsimd call.
4400 (aarch64_init_cumulative_args): Ditto.
4401 (aarch64_gimplify_va_arg_expr): Ditto.
4402 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4404 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4406 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4407 (*simple_return_indirect_internal<mode>): Rename from
4408 simple_return_indirect_internal. Use W mode iterator.
4409 (rstorssp): New expander.
4410 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4411 (clrssbsy): New expander.
4412 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4414 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4416 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4418 (cmse_check_pointed_object): Likewise.
4420 2018-06-05 Martin Liska <mliska@suse.cz>
4422 PR gcov-profile/47618
4423 * doc/invoke.texi: Document how -fprofile-dir format
4426 2018-06-05 Richard Biener <rguenther@suse.de>
4428 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4429 removal pretend DOM info isn't available so we do not update
4430 it and only remove edges, not dominated blocks. Actually free
4431 DOM info in case we removed something. Remove unreachable blocks.
4432 (mfb_keep_latches): Work with either DOM info or marked backedges.
4433 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4434 first. Mark backedges if DOM info isn't available.
4435 (Re-)compute DOM info after cleanup_control_flow_pre.
4437 2018-06-05 Richard Biener <rguenther@suse.de>
4439 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4440 (locus_discrim_hasher::hash): Adjust.
4441 (locus_discrim_hasher::equal): Likewise.
4442 (next_discriminator_for_locus): Work on line directly.
4443 (same_line_p): Pass in expanded locus1 as well.
4444 (assign_discriminators): Avoid redundant location expansions.
4446 2018-06-05 Richard Biener <rguenther@suse.de>
4448 PR tree-optimization/86046
4449 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4450 if required after clearing TREE_ADDRESSABLE.
4452 2018-06-05 Richard Biener <rguenther@suse.de>
4454 PR tree-optimization/86047
4455 * tree-ssa-loop.c (for_each_index): Glob handling of all
4456 decls and constants and really handle all of them.
4458 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4461 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4462 qualifier_void_pointer and qualifier_const_void_pointer.
4463 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4464 (arm_init_builtins): Handle the above.
4465 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4466 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4467 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4470 2018-06-05 Martin Liska <mliska@suse.cz>
4472 * auto-profile.c (read_autofdo_file): Do not use
4473 gcov_ctr_summary struct.
4474 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4475 * coverage.c (struct counts_entry): Likewise.
4476 (read_counts_file): Read just single summary entry.
4477 (get_coverage_counts): Use gcov_summary struct.
4478 * coverage.h (get_coverage_counts): Likewise.
4479 * gcov-dump.c (dump_working_sets): Likewise.
4480 (tag_summary): Dump just single summary.
4481 * gcov-io.c (gcov_write_summary): Write just histogram
4483 (gcov_read_summary): Read just single summary.
4484 (compute_working_sets): Use gcov_summary struct.
4485 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4486 of GCOV_COUNTERS_SUMMABLE.
4487 (GCOV_COUNTERS_SUMMABLE): Remove.
4488 (GCOV_FIRST_VALUE_COUNTER): Replace with
4489 GCOV_COUNTER_V_INTERVAL.
4490 (struct gcov_ctr_summary): Remove.
4491 (struct gcov_summary): Directly use fields of former
4493 (compute_working_sets): Use gcov_summary struct.
4494 * gcov.c (read_count_file): Do not use ctrs fields.
4495 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4497 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4499 * profile.c: Likewise.
4500 * profile.h: Likewise.
4502 2018-06-05 Martin Liska <mliska@suse.cz>
4504 PR gcov-profile/84846
4505 * gcov.c (output_lines): Print working directory only
4506 in intermediate format.
4508 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4510 * config/s390/s390-builtin-types.def: Add void function type.
4511 * config/s390/s390-builtins.def: Use the function type for the
4514 2018-06-04 Jim Wilson <jimw@sifive.com>
4516 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4518 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4519 and maybe_eh_return. Change regno to unsigned int. Use new args to
4520 handle EH_RETURN_DATA_REGNO registers properly.
4521 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4522 (riscv_expand_epilogue): Update comment. Change argument name and
4523 type. Update code to use new name and type. Pass new args to
4524 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4526 * config/riscv/riscv.md (NORMAL_RETURN): New.
4527 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4528 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4529 (eh_return): Call gen_eh_return_internal and emit barrier.
4530 (eh_return_internal): Call riscv_expand_epilogue.
4532 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4534 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4535 bit_insertion field and declare can_be_merged_into method.
4536 (merged_store_group::can_be_merged_into): New method.
4537 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4538 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4539 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4541 2018-06-04 Richard Biener <rguenther@suse.de>
4543 PR tree-optimization/85955
4544 * builtins.c (fold_builtin_sincos): Convert pointers to
4545 destination to appropriate type before dereferencing.
4547 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4549 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4551 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4553 * expr.c (expand_expr_real_1): Force the operand into memory if
4554 its TYPE_MODE is BLKmode and if there is no integer mode for
4555 the number of bits being extracted.
4557 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4561 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4562 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4564 2018-06-04 Richard Biener <rguenther@suse.de>
4566 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4567 (cleanup_tree_cfg_noloop): ... single caller. Do
4568 start_recording_case_labels later.
4570 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4572 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4573 to _IMMINTRIN_H_INCLUDED.
4574 * config/i386/pconfigintrin.h: Ditto.
4575 * config/i386/waitpkgintrin.h: Ditto.
4576 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4577 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4578 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4579 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4580 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4581 waitpkgintrin.h and cldemoteintrin.h.
4583 2018-06-04 Richard Biener <rguenther@suse.de>
4585 PR tree-optimization/86038
4586 * tracer.c (find_best_successor): Check probability for
4587 being initialized, bail out if not.
4589 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4592 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4593 of bits to ignore when comparing architectures.
4595 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4597 PR tree-optimization/69615
4598 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4599 maximum or minimum of the type, try to merge it also as if
4600 range1 is + [-, x - 1] or + [x + 1, -].
4603 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4605 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4607 PR tree-optimization/86034
4608 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4609 the unsigned bitfield type in a bit insertion sequence if it does not
4610 have a larger precision than the bitfield size.
4611 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4613 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4615 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4617 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4619 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4620 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4621 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4622 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4624 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4626 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4627 Disable -fdelete-null-pointer-checks for ELF toolchain.
4629 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4630 Kito Cheng <kito.cheng@gmail.com>
4632 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4633 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4634 (nds32*-*-*): Add glibc and uclibc conditions.
4635 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4636 (TARGET_EXCEPT_UNWIND_INFO): Define.
4637 * config/nds32/elf.h: New file.
4638 * config/nds32/linux.h: New file.
4639 * config/nds32/nds32-elf.opt: New file.
4640 * config/nds32/nds32-linux.opt: New file.
4641 * config/nds32/nds32-fp-as-gp.c
4642 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4643 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4645 (nds32_asm_file_end): Ditto.
4646 (nds32_print_operand): Ditto.
4647 (nds32_insert_attributes): Ditto.
4648 (nds32_init_libfuncs): New function.
4649 (TARGET_HAVE_TLS): Define.
4650 (TARGET_INIT_LIBFUNCS): Define.
4651 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4653 (TARGET_ELF): Apply different mcmodel setting.
4654 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4655 been migrated into elf.h and linux.h files.
4656 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4657 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4658 (mcmodel): The content has been migrated into nds32-elf.opt and
4659 nds32-linux.opt files.
4660 * config/nds32/t-elf: New file.
4661 * config/nds32/t-linux: New file.
4663 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4664 Shiva Chen <shiva0217@gmail.com>
4666 * config/nds32/constants.md (unspec_volatile_element): Add
4667 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4668 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4670 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4671 (make_pass_nds32_fp_as_gp): Declare.
4672 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4674 (nds32_asm_function_end_prologue): Remove unused asm output.
4675 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4676 (nds32_asm_file_start): Output necessary fp_as_gp information.
4677 (nds32_option_override): Adjust register usage.
4678 (nds32_expand_prologue): Consider fp_as_gp situation.
4679 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4680 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4683 (simple_return): Ditto.
4684 (omit_fp_begin): Output special directive for fp_as_gp.
4685 (omit_fp_end): Output special directive for fp_as_gp.
4686 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4687 mforbid-fp-as-gp): New options.
4689 2018-06-01 Mark Wielaard <mark@klomp.org>
4691 * dwarf2out.c (dwarf2out_finish): Remove generation of
4692 DW_AT_loclists_base.
4694 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4696 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4697 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4698 (struct merged_store_group): Add bit_insertion field.
4699 (dump_char_array): Use standard hexadecimal format.
4700 (merged_store_group::merged_store_group): Set bit_insertion to false.
4701 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4702 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4703 also print the mask in the dump file.
4704 (pass_store_merging::gate): Minor tweak.
4705 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4706 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4707 stores with INTEGER_CST stores.
4708 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4709 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4710 and use it throughout. Generate bit insertion sequences if need be.
4711 (pass_store_merging::process_store): Remove redundant condition.
4712 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4714 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4716 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4717 the 128-bit floating point types. Fix function comment.
4719 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4721 * config/aarch64/aarch64-simd.md
4722 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4724 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4727 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4729 PR tree-optimization/85989
4730 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4732 (backprop::intersect_uses): Check it when deciding whether this
4733 is a backedge reference.
4734 (backprop::process_block): Add each phi to m_visited_phis
4735 after visiting it, then clear it at the end.
4737 2018-06-01 Richard Biener <rguenther@suse.de>
4739 * tree-vectorizer.h (vect_dr_stmt): New function.
4740 (vect_get_load_cost): Adjust.
4741 (vect_get_store_cost): Likewise.
4742 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4743 Use vect_dr_stmt instead of DR_SMTT.
4744 (vect_record_base_alignments): Likewise.
4745 (vect_calculate_target_alignment): Likewise.
4746 (vect_compute_data_ref_alignment): Likewise and make static.
4747 (vect_update_misalignment_for_peel): Likewise.
4748 (vect_verify_datarefs_alignment): Likewise.
4749 (vector_alignment_reachable_p): Likewise.
4750 (vect_get_data_access_cost): Likewise. Pass down
4751 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4752 (vect_get_peeling_costs_all_drs): Likewise.
4753 (vect_peeling_hash_get_lowest_cost): Likewise.
4754 (vect_enhance_data_refs_alignment): Likewise.
4755 (vect_find_same_alignment_drs): Likewise.
4756 (vect_analyze_data_refs_alignment): Likewise.
4757 (vect_analyze_group_access_1): Likewise.
4758 (vect_analyze_group_access): Likewise.
4759 (vect_analyze_data_ref_access): Likewise.
4760 (vect_analyze_data_ref_accesses): Likewise.
4761 (vect_vfa_segment_size): Likewise.
4762 (vect_small_gap_p): Likewise.
4763 (vectorizable_with_step_bound_p): Likewise.
4764 (vect_prune_runtime_alias_test_list): Likewise.
4765 (vect_analyze_data_refs): Likewise.
4766 (vect_supportable_dr_alignment): Likewise.
4767 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4768 (vect_gen_prolog_loop_niters): Likewise.
4769 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4770 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4772 (vect_recog_mask_conversion_pattern): Likewise.
4773 (vect_try_gather_scatter_pattern): Likewise.
4774 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4775 to vect_get_store_cost.
4776 (vect_get_store_cost): Get stmt_info instead of DR.
4777 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4778 (vect_get_load_cost): Get stmt_info instead of DR.
4780 2018-06-01 Richard Biener <rguenther@suse.de>
4783 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4784 vars wrapped in MEM_REFs.
4786 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4788 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4789 Fix subreg tests so that we only return a choice between
4790 GENERAL_REGS and FP_REGS if the original classes included both.
4792 2018-06-01 Richard Biener <rguenther@suse.de>
4795 * tree-ssa-structalias.c (get_function_part_constraint):
4796 Handle NULL fi->decl.
4797 (find_func_aliases_for_call): Properly handle indirect
4798 fi from direct call.
4799 (find_func_clobbers): Likewise.
4800 (ipa_pta_execute): Likewise.
4801 (create_variable_info_for): For functions that are ifunc_resolver
4802 resolve to a varinfo that contains the result of the resolver call.
4803 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4806 2018-05-31 Michael Collison <michael.collison@arm.com>
4808 * config/aarch64/aarch64.md:
4809 (*fix_to_zero_extenddfdi2): New pattern.
4810 * gcc.target/aarch64/fix_extend1.c: New testcase.
4812 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4816 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4817 and BUILT_IN_STRNCMP_EQ.
4818 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4819 BUILT_IN_STRNCMP_EQ.
4820 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4821 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4822 (gimple_fold_builtin): Likewise.
4823 * tree-ssa-strlen.c (compute_string_length): New function.
4824 (determine_min_obsize): New function.
4825 (handle_builtin_string_cmp): New function to handle calls to
4826 string compare functions.
4827 (strlen_optimize_stmt): Add handling to builtin string compare
4829 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4830 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4831 * tree.c (build_common_builtin_nodes): Add new defines of
4832 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4834 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4837 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4838 functions with naked attribute.
4840 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4842 * config/i386/sse.md (avx_vec_concat<mode>):
4843 Substitute concat_tg_mode mode attribute with xtg_mode.
4844 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4845 (concat_tg_mode): Remove mode attribute.
4847 2018-05-31 Martin Sebor <msebor@redhat.com>
4850 * calls.c (alloc_max_size): Correct a logic error/typo.
4851 Treat excessive arguments as infinite. Warn for invalid arguments.
4852 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4854 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4857 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4858 and movx for Haswell.
4860 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4861 Cesar Philippidis <cesar@codesourcery.com>
4864 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4865 when emitting error on private/firstprivate reductions.
4866 * omp-low.c (lower_omp_target): Avoid reference-type processing
4867 on pointers for firstprivate clause.
4869 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4871 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4874 * config/aarch64/aarch64-simd.md
4875 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4876 (aarch64_ld1_x3_<mode>): Likewise
4877 (aarch64_st1x2<VALLDIF:mode>): Likewise
4878 (aarch64_st1_x2_<mode>): Likewise
4879 (aarch64_st1x3<VALLDIF:mode>): Likewise
4880 (aarch64_st1_x3_<mode>): Likewise
4881 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4882 (vld1_s8_x3): Likewise.
4883 (vld1_u16_x3): Likewise.
4884 (vld1_s16_x3): Likewise.
4885 (vld1_u32_x3): Likewise.
4886 (vld1_s32_x3): Likewise.
4887 (vld1_u64_x3): Likewise.
4888 (vld1_s64_x3): Likewise.
4889 (vld1_f16_x3): Likewise.
4890 (vld1_f32_x3): Likewise.
4891 (vld1_f64_x3): Likewise.
4892 (vld1_p8_x3): Likewise.
4893 (vld1_p16_x3): Likewise.
4894 (vld1_p64_x3): Likewise.
4895 (vld1q_u8_x3): Likewise.
4896 (vld1q_s8_x3): Likewise.
4897 (vld1q_u16_x3): Likewise.
4898 (vld1q_s16_x3): Likewise.
4899 (vld1q_u32_x3): Likewise.
4900 (vld1q_s32_x3): Likewise.
4901 (vld1q_u64_x3): Likewise.
4902 (vld1q_s64_x3): Likewise.
4903 (vld1q_f16_x3): Likewise.
4904 (vld1q_f32_x3): Likewise.
4905 (vld1q_f64_x3): Likewise.
4906 (vld1q_p8_x3): Likewise.
4907 (vld1q_p16_x3): Likewise.
4908 (vld1q_p64_x3): Likewise.
4909 (vst1_s64_x2): Likewise.
4910 (vst1_u64_x2): Likewise.
4911 (vst1_f64_x2): Likewise.
4912 (vst1_s8_x2): Likewise.
4913 (vst1_p8_x2): Likewise.
4914 (vst1_s16_x2): Likewise.
4915 (vst1_p16_x2): Likewise.
4916 (vst1_s32_x2): Likewise.
4917 (vst1_u8_x2): Likewise.
4918 (vst1_u16_x2): Likewise.
4919 (vst1_u32_x2): Likewise.
4920 (vst1_f16_x2): Likewise.
4921 (vst1_f32_x2): Likewise.
4922 (vst1_p64_x2): Likewise.
4923 (vst1q_s8_x2): Likewise.
4924 (vst1q_p8_x2): Likewise.
4925 (vst1q_s16_x2): Likewise.
4926 (vst1q_p16_x2): Likewise.
4927 (vst1q_s32_x2): Likewise.
4928 (vst1q_s64_x2): Likewise.
4929 (vst1q_u8_x2): Likewise.
4930 (vst1q_u16_x2): Likewise.
4931 (vst1q_u32_x2): Likewise.
4932 (vst1q_u64_x2): Likewise.
4933 (vst1q_f16_x2): Likewise.
4934 (vst1q_f32_x2): Likewise.
4935 (vst1q_f64_x2): Likewise.
4936 (vst1q_p64_x2): Likewise.
4937 (vst1_s64_x3): Likewise.
4938 (vst1_u64_x3): Likewise.
4939 (vst1_f64_x3): Likewise.
4940 (vst1_s8_x3): Likewise.
4941 (vst1_p8_x3): Likewise.
4942 (vst1_s16_x3): Likewise.
4943 (vst1_p16_x3): Likewise.
4944 (vst1_s32_x3): Likewise.
4945 (vst1_u8_x3): Likewise.
4946 (vst1_u16_x3): Likewise.
4947 (vst1_u32_x3): Likewise.
4948 (vst1_f16_x3): Likewise.
4949 (vst1_f32_x3): Likewise.
4950 (vst1_p64_x3): Likewise.
4951 (vst1q_s8_x3): Likewise.
4952 (vst1q_p8_x3): Likewise.
4953 (vst1q_s16_x3): Likewise.
4954 (vst1q_p16_x3): Likewise.
4955 (vst1q_s32_x3): Likewise.
4956 (vst1q_s64_x3): Likewise.
4957 (vst1q_u8_x3): Likewise.
4958 (vst1q_u16_x3): Likewise.
4959 (vst1q_u32_x3): Likewise.
4960 (vst1q_u64_x3): Likewise.
4961 (vst1q_f16_x3): Likewise.
4962 (vst1q_f32_x3): Likewise.
4963 (vst1q_f64_x3): Likewise.
4964 (vst1q_p64_x3): Likewise.
4966 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4968 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4969 user_label_prefix to name.
4971 * tree-core.h: Update comment about the format of NAME string
4972 passed to handler in attribute_spec.
4974 * config/msp430/msp430.md: Remove erroneous subreg expression from
4975 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4976 zero_extend{q,h}isi2.
4978 2018-05-30 Borislav Petkov <bp@suse.de>
4980 * doc/extend.texi: Document some architecture specific
4981 constraints and sort entries.
4983 2018-05-30 Martin Sebor <msebor@redhat.com>
4986 * builtins.c (expand_builtin_stpcpy_1): New function.
4987 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4988 only if the former succeeds.
4990 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4992 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4995 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4997 * doc/invoke.texi (-flinker-output): Document
4999 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5001 * passes.c (ipa_write_summaries): Only modify statements if body
5003 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5004 incrementally linking.
5005 (ipa_passes): Likewise.
5006 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5007 pass down resolution info.
5008 * common.opt (flag_incremental_link): Update info.
5009 * gcc.c (plugin specs): Turn flinker-output=* to
5010 -plugin-opt=-linker-output-known
5011 * toplev.c (compile_file): Also cut compilation when doing incremental
5013 * flag-types. (enum lto_partition_model): Add
5014 LTO_LINKER_OUTPUT_NOLTOREL.
5015 (invoke.texi): Add -flinker-output docs.
5016 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5017 link same way as WPA; do not stream in dead initializers.
5019 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5020 darf2out_register_external_decl): Support incremental link.
5022 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5024 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5026 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5028 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5029 it down to simple_object_copy_lto_debug_sections.
5030 (run_gcc): Determine incremental LTO link time and configure
5031 lto1 into non-wpa mode, disable renaming of debug sections.
5033 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5035 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5036 descriptions of various incorrectly documented functions.
5038 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5041 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5042 address check not strict.
5044 2018-05-30 Richard Biener <rguenther@suse.de>
5046 PR tree-optimization/85964
5047 * tracer.c (better_p): Drop initialized count check, we only
5048 call the function with initialized counts now.
5049 (find_best_successor): Do find a best edge if one
5050 has uninitialized count.
5051 (find_best_predecessor): Likewise. Do BB frequency check only
5052 if count is initialized.
5054 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5056 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5057 (aarch64_ldrstr_offset_compare): New.
5058 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5059 load/store orderings.
5060 (aarch64_gen_adjusted_ldpstp): Likewise.
5062 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5064 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5065 Check for subset of GENERAL_REGS and FP_REGS.
5066 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5069 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5071 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5072 and wi::to_poly_offset. Add the current offset and then check
5073 whether the sum fits, rather than using an unchecked addition of
5074 a checked term. Check for a shwi rather than a uhwi.
5075 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5076 (store_constructor): Use poly_int_tree_p.
5077 (expand_expr_real_1): Likewise.
5078 * function.c (assign_temp): Likewise.
5079 * fold-const.c (const_binop): Use poly_int_tree_p and
5081 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5083 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5084 to_poly_offset for MEM offsets.
5085 * ipa-icf.c (sem_variable::equals): Likewise.
5086 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5087 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5088 wi::to_poly_offset for BIT_FIELD_REF offsets.
5089 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5091 * var-tracking.c (emit_note_insn_var_location): Use
5092 tree_to_poly_uint64.
5094 2018-05-29 Jim Wilson <jimw@sifive.com>
5096 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5098 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5101 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5102 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5104 (sse4_1_round<mode>2): Use nonimmediate_operand
5105 for operand 1 predicate.
5107 2018-05-29 Martin Sebor <msebor@redhat.com>
5108 Richard Biener <rguenther@suse.de>
5111 * calls.c (get_size_range): Call determine_value_range instead
5112 of get_value_range..
5113 * tree-vrp.h (determine_value_range): Declared new function.
5114 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5116 2018-05-29 Richard Biener <rguenther@suse.de>
5118 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5119 sure to use non-pattern stmts for get_earlier_stmt arguments.
5120 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5121 called on pattern stmts.
5122 (get_later_stmt): Likewise.
5124 2018-05-29 Martin Liska <mliska@suse.cz>
5126 PR gcov-profile/85759
5127 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5130 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5132 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5134 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5138 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5139 VEC_PACK_FLOAT_EXPR): New tree codes.
5140 * tree-pretty-print.c (op_code_prio): Handle
5141 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5142 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5143 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5144 * tree-inline.c (estimate_operator_cost): Likewise.
5145 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5146 * fold-const.c (const_binop): Likewise.
5147 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5148 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5149 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5150 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5151 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5152 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5153 * expr.c (expand_expr_real_2): Likewise.
5154 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5155 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5156 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5158 * optabs.c (expand_widen_pattern_expr): For
5159 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5160 sign from result type rather than operand's type.
5161 (expand_binop_directly): For vec_packu_float_optab and
5162 vec_packs_float_optab allow result type to be different from operand's
5164 * optabs-tree.c (optab_for_tree_code): Handle
5165 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5166 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5167 * tree-vect-generic.c (expand_vector_operations_1): Handle
5168 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5169 VEC_PACK_FLOAT_EXPR.
5170 * tree-vect-stmts.c (supportable_widening_operation): Handle
5172 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5173 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5174 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5175 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5176 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5178 (vec_pack<floatprefix>_float_<mode>): New expander.
5179 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5181 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5182 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5183 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5184 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5185 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5187 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5188 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5189 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5190 VEC_PACK_FLOAT_EXPR): Document.
5192 2018-05-29 Richard Biener <rguenther@suse.de>
5194 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5196 (stmt_vec_info_vec): Make pointer.
5197 (init_stmt_vec_info_vec): Remove.
5198 (free_stmt_vec_info_vec): Likewise.
5199 (set_stmt_vec_info_vec): New function.
5200 (free_stmt_vec_infos): Likewise.
5201 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5202 (set_vinfo_for_stmt): Likewise.
5203 (get_earlier_stmt): Likewise.
5204 (get_later_stmt): Likewise.
5205 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5206 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5207 (vec_info::~vec_info): Free stmt_vec_infos.
5208 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5209 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5210 (pass_slp_vectorize::execute): Likewise.
5211 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5212 (free_stmt_vec_info_vec): Likewise.
5213 (set_stmt_vec_info_vec): New function.
5214 (free_stmt_vec_infos): Likewise.
5215 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5216 the global stmt_vec_info_vec.
5217 * tree-parloops.c (gather_scalar_reductions): Use
5218 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5221 2018-05-29 Richard Biener <rguenther@suse.de>
5223 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5225 2018-05-29 Martin Liska <mliska@suse.cz>
5226 David Malcolm <dmalcolm@redhat.com>
5228 * vec.c (test_reverse): New.
5229 (vec_c_tests): Add new test.
5230 * vec.h (vl_ptr>::reverse): New function.
5232 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5234 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5236 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5239 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5241 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5243 2018-05-28 Richard Biener <rguenther@suse.de>
5245 PR tree-optimization/85933
5246 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5247 look at stmts marked as vectorizable.
5249 2018-05-28 Richard Biener <rguenther@suse.de>
5251 PR tree-optimization/85934
5252 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5253 vector boolean check before scalar optimization.
5255 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5257 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5260 2018-05-28 Mark Wielaard <mark@klomp.org>
5262 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5263 if it is an expression containing a minus sign.
5265 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5267 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5269 2018-05-27 Paul Koning <ni1d@arrl.net>
5271 * config/pdp11/pdp11.md (truncsihi2): Remove.
5273 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5274 Chung-Ju Wu <jasonwucj@gmail.com>
5276 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5278 (unaligned_store_dw): Ditto.
5279 * config/nds32/nds32-memory-manipulation.c
5280 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5281 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5282 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5283 (emit_setmem_word_loop): Rename to ...
5284 (emit_setmem_doubleword_loop): ... this.
5285 (nds32_gen_dup_4_byte_to_word_value): New function.
5286 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5287 (nds32_expand_setmem_loop): Refine implementation.
5288 (nds32_expand_setmem_loop_v3m): Ditto.
5289 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5292 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5294 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5296 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5298 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5299 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5300 machine->attr_no_prologue_p.
5301 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5302 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5303 (nds32_expand_epilogue): Consider attr_naked_p.
5304 (nds32_expand_epilogue_v3pop): Likewise.
5305 (nds32_can_use_return_insn): Likewise.
5306 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5307 attr_no_prologue_p fields.
5308 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5310 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5313 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5315 * config/i386/sse.md
5316 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5318 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5320 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5322 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5324 (*<floatsuffix>floatv2div2sf2): Rename to ...
5325 (*float<floatunssuffix>v2div2sf2): ... this.
5326 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5327 (float<floatunssuffix>v2div2sf2_mask): ... this.
5328 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5329 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5330 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5332 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5334 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5336 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5338 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5340 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5342 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5343 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5344 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5345 gen_ufix_truncv8dfv8si2.
5346 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5347 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5348 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5349 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5350 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5351 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5352 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5353 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5355 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5359 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5361 2018-05-25 Jim Wilson <jimw@sifive.com>
5363 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5364 * config/riscv/riscv.c (struct machine_function): Add
5365 interrupt_handler_p and attribute_checked_p fields.
5366 (riscv_attribute_table): Add interrupt.
5367 (riscv_interrupt_type_p): New.
5368 (riscv_save_reg_p): Save extra regs for interrupt handler.
5369 (riscv_use_save_libcall): Return false for interrupt handler.
5370 (riscv_first_stack_step): Add forward declaration.
5371 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5372 for interrupt handler with large frame. Use it for saved reg list.
5373 (riscv_expand_prologue): Move flag_stack_usage_info support to
5374 eliminate duplication.
5375 (riscv_expand_epilogue): Generate mret for interrupt handler.
5376 (riscv_epilogue_uses): New.
5377 (riscv_can_use_return_insn): Return false for interrupt handler.
5378 (riscv_function_ok_for_sibcall): Likewise.
5379 (riscv_set_current_function): Add interrupt handler support.
5380 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5381 * config/riscv/riscv.md (UNSPECV_MRET): New.
5383 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5385 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5387 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5389 PR tree-optimization/85712
5390 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5391 this candidate has already been replaced in-situ by a copy.
5393 2018-05-25 Jason Merrill <jason@redhat.com>
5395 PR c++/80485 - inline function non-zero address.
5396 * symtab.c (nonzero_address): Check DECL_COMDAT.
5398 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5401 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5402 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5403 extension of SImode operation. Use const123_operand predicate.
5404 (*saddsi_1): Remove.
5405 (*saddl_se_1): Ditto.
5407 (*ssubl_se_1): Ditto.
5408 * config/alpha/predicates.md (const123_operand): New predicate.
5409 * config/alpha/constraints.md (P): Use IN_RANGE.
5411 2018-05-25 Richard Biener <rguenther@suse.de>
5413 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5415 (ref_maybe_used_by_stmt_p): Likewise.
5416 (stmt_may_clobber_ref_p): Likewise.
5417 (stmt_may_clobber_ref_p_1): Likewise.
5418 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5420 (ref_maybe_used_by_stmt_p): Likewise.
5421 (stmt_may_clobber_ref_p): Likewise.
5422 (stmt_may_clobber_ref_p_1): Likewise.
5423 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5424 the alias oracle to disambiguate DRs with stmts DR analysis
5426 (vect_analyze_data_refs): Do not give up on not analyzable
5427 DRs for BB vectorization. Remove code truncating the dataref
5430 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5433 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5434 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5435 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5437 2018-05-25 Richard Biener <rguenther@suse.de>
5439 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5440 function, combining stmt data ref gathering and fatal analysis
5442 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5443 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5444 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5445 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5446 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5448 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5450 PR tree-optimization/85720
5451 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5452 SCC if all partitions are builtins.
5453 (version_loop_by_alias_check): New parameter. Generate cancelable
5454 runtime alias check if all partitions are builtins.
5455 (distribute_loop): Update call to above function.
5457 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5459 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5460 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5461 (parm_default_def_partition_arg): Ditto.
5462 (set_parm_default_def_partition): Ditto.
5463 (get_parm_default_def_partitions): Ditto and make it static.
5464 (get_undefined_value_partitions): Ditto and make it static.
5465 (remove_ssa_form): Refactor call to init_var_map here.
5466 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5467 computation for loop region.
5468 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5469 (register_default_def): Delete.
5470 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5471 (parm_default_def_partition_arg): Ditto.
5472 (set_parm_default_def_partition): Ditto.
5473 (get_parm_default_def_partitions): Ditto and make it static.
5474 (get_undefined_value_partitions): Ditto and make it static.
5475 (coalesce_with_default, coalesce_with_default): Update comment.
5476 (create_coalesce_list_for_region): New func factored out from
5477 create_outofssa_var_map.
5478 (populate_coalesce_list_for_outofssa): New func factored out from
5479 create_outofssa_var_map and coalesce_ssa_name.
5480 (create_outofssa_var_map): Delete.
5481 (coalesce_ssa_name): Refactor to support live range computation.
5482 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5483 (get_parm_default_def_partitions): Delete.
5484 (get_undefined_value_partitions): Ditto.
5485 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5486 computation for loop region.
5487 (new_tree_live_info, loe_visit_block): Ditto.
5488 (live_worklist, set_var_live_on_entry): Ditto.
5489 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5490 * tree-ssa-live.h (struct _var_map): New fields.
5491 (init_var_map): Change decl.
5492 (region_contains_p): New.
5494 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5496 * tree-ssa-live.h (live_merge_and_clear): Delete.
5498 2018-05-25 Richard Biener <rguenther@suse.de>
5501 * tree-dump.c (dequeue_and_dump): Remove access to removed
5502 operand 2 of a SWITCH_EXPR.
5504 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5506 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5507 multiplication and division.
5508 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5509 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5510 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5511 (cond_udiv_optab, cond_umod_optab): New optabs.
5512 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5513 (IFN_COND_RDIV): New internal functions.
5514 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5515 TRUNC_MOD_EXPR and RDIV_EXPR.
5516 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5517 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5519 (SVE_INT_BINARY): Include mult.
5520 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5521 (optab, sve_int_op): Handle mult.
5522 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5524 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5525 for SVE_INT_BINARY_SD.
5527 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5529 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5530 (optab, sve_int_op): Handle div and udiv.
5531 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5532 for SVE_INT_BINARY_SD.
5533 (*<optab><mode>3): New insn for the same.
5535 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5537 * tree-vect-patterns.c: Include predict.h.
5538 (vect_recog_divmod_pattern): Restrict check for division support
5539 to when optimizing for size.
5541 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5543 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5544 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5545 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5546 (gimple_match_op::set_op): Likewise.
5547 (gimple_resimplify4): Declare.
5548 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5549 (expr::gen_transform): Likewise.
5550 (decision_tree::gen): Generate a simplification routine for 4 operands.
5551 * gimple-match-head.c (gimple_simplify): Add an overload for
5552 4 operands. In the top-level function, handle up to 4 call
5553 arguments and call gimple_resimplify4.
5554 (gimple_resimplify4): New function.
5555 (build_call_internal): Pass a fourth operand.
5556 (maybe_push_to_seq): Likewise.
5557 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5558 Fold VEC_COND_EXPRs of an operation and a default value into
5559 an IFN_COND_* function if possible.
5560 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5562 (SVE_COND_FP_BINARY): Include them.
5563 (optab, sve_fp_op): Handle them.
5564 (SVE_INT_BINARY_REV): New code iterator.
5565 (SVE_COND_FP_BINARY_REV): New int iterator.
5566 (commutative): New int attribute.
5567 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5569 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5571 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5572 (*cond_<optab><mode>): New patterns for reversed operands.
5574 2018-05-25 Richard Biener <rguenther@suse.de>
5576 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5577 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5578 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5579 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5580 * tree-vect-data-refs.c (everywhere): Adjust users.
5581 * tree-vect-loop.c (everywhere): Likewise.
5582 * tree-vect-slp.c (everywhere): Likewise.
5583 * tree-vect-stmts.c (everywhere): Likewise.
5584 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5586 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5588 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5590 (gcc_cv_as_section_exclude): ... this.
5591 Try Solaris as #exclude syntax.
5592 * configure: Regenerate.
5593 * config.in: Regenerate.
5594 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5596 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5597 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5599 * varasm.c (default_elf_asm_named_section): Don't check if
5600 HAVE_GAS_SECTION_EXCLUDE is defined.
5602 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5604 * doc/md.texi: Update the documentation of the cond_* optabs
5605 to mention the new final operand. Fix GET_MODE_NUNITS call.
5606 Describe the scalar case too.
5607 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5608 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5610 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5611 (get_conditional_internal_fn): Update comment.
5612 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5613 accumulator value as a final argument to conditional functions.
5614 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5615 a define_expand and add an "else" operand. Assert for now that
5616 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5617 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5618 (*cond_<optab><mode>): New patterns.
5619 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5620 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5621 (UNSPEC_COND_EOR): Delete.
5622 (optab): Remove associated mappings.
5623 (SVE_INT_BINARY): New code iterator.
5624 (sve_int_op): Remove int attribute and add "minus" to the code
5626 (SVE_COND_INT_OP): Delete.
5627 (SVE_COND_FP_OP): Rename to...
5628 (SVE_COND_FP_BINARY): ...this.
5630 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5632 * optabs.c (can_reuse_operands_p): New function.
5633 (maybe_legitimize_operands): Try to reuse the results for
5636 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5638 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5639 Add {q} suffix to insn mnemonic.
5641 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5643 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5644 (msp430_warn_func_return): New.
5646 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5648 * fold-const.c (tree_nonzero_bits): New function.
5649 * fold-const.h (tree_nonzero_bits): Likewise.
5650 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5651 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5653 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5657 * varasm.c (assemble_alias): Check ifunc_resolver only on
5660 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5663 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5664 when memory input operand is handled.
5666 2018-05-24 Luis Machado <luis.machado@linaro.org>
5668 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5670 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5672 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5674 * match.pd: Delay FMA folds until after vectorization.
5676 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5679 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5680 address check not strict.
5682 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5684 * gimple-match.h (gimple_match_op): New class.
5685 (mprts_hook): Replace parameters with a gimple_match_op *.
5686 (maybe_build_generic_op): Likewise.
5687 (gimple_simplified_result_is_gimple_val): Replace parameters with
5688 a const gimple_match_op *.
5689 (gimple_simplify): Replace code_helper * and tree * parameters with
5690 a gimple_match_op * parameter.
5691 (gimple_resimplify1): Replace code_helper *, tree and tree *
5692 parameters with a gimple_match_op * parameter.
5693 (gimple_resimplify2): Likewise.
5694 (gimple_resimplify3): Likewise.
5695 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5696 parameters with a gimple_match_op * parameter.
5697 * gimple-match-head.c (gimple_simplify): Change prototypes of
5698 auto-generated functions to take a gimple_match_op * instead of
5699 separate code_helper * and tree * parameters. Make the same
5700 change in the top-level overload and update calls to the
5701 gimple_resimplify routines. Update calls to the auto-generated
5702 functions and to maybe_push_res_to_seq in the publicly-facing
5703 operation-specific gimple_simplify overloads.
5704 (gimple_match_op::MAX_NUM_OPS): Define.
5705 (gimple_resimplify1): Replace rcode and ops with a single res_op
5706 parameter. Update call to gimple_simplify.
5707 (gimple_resimplify2): Likewise.
5708 (gimple_resimplify3): Likewise.
5709 (mprts_hook): Replace parameters with a gimple_match_op *.
5710 (maybe_build_generic_op): Likewise.
5711 (build_call_internal): Replace type, nargs and ops with
5712 a gimple_match_op *.
5713 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5714 with a single gimple_match_op *. Update calls to mprts_hook,
5715 build_call_internal and gimple_simplified_result_is_gimple_val.
5716 Factor out code that is common to the tree_code and combined_fn cases.
5717 * genmatch.c (expr::gen_transform): Replace tem_code and
5718 tem_ops with a gimple_match_op called tem_op. Update calls
5719 to the gimple_resimplify functions and maybe_push_res_to_seq.
5720 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5721 res_ops. Update call to the gimple_resimplify functions.
5722 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5723 (decision_tree::gen): Make the functions take a gimple_match_op *
5724 called res_op instead of separate res_code and res_ops parameters.
5725 Update call accordingly.
5726 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5727 and ops with a single res_op parameter. Update calls to
5728 maybe_build_generic_op and maybe_push_res_to_seq.
5729 (fold_stmt_1): Update calls to gimple_simplify and
5730 replace_stmt_with_simplification.
5731 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5732 and gimple_simplified_result_is_gimple_val.
5733 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5735 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5736 with a gimple_match_op *.
5737 (vn_nary_build_or_lookup): Likewise. Update call to
5738 vn_nary_build_or_lookup_1.
5739 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5740 gimple_match_op *. Update calls to the gimple_resimplify routines
5741 and to gimple_simplified_result_is_gimple_val.
5742 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5743 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5744 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5745 (visit_nary_op): Likewise.
5746 (visit_reference_op_load): Likewise.
5748 2018-05-23 Luis Machado <luis.machado@linaro.org>
5750 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5751 modifier for printing the step amount.
5753 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5756 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5759 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5761 * doc/sourcebuild.texi (Endianness): New subsubsection.
5763 2018-05-23 Luis Machado <luis.machado@linaro.org>
5765 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5766 <prefetch_dynamic_strides>: New const bool field.
5767 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5768 prefetch_dynamic_strides.
5769 (exynosm1_prefetch_tune): Likewise.
5770 (thunderxt88_prefetch_tune): Likewise.
5771 (thunderx_prefetch_tune): Likewise.
5772 (thunderx2t99_prefetch_tune): Likewise.
5773 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5775 (aarch64_override_options_internal): Update to set
5776 PARAM_PREFETCH_DYNAMIC_STRIDES.
5777 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5778 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5779 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5780 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5781 prefetch-dynamic-strides setting.
5783 2018-05-23 Luis Machado <luis.machado@linaro.org>
5785 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5786 <minimum_stride>: New const int field.
5787 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5788 minimum_stride field defaulting to -1.
5789 (exynosm1_prefetch_tune): Likewise.
5790 (thunderxt88_prefetch_tune): Likewise.
5791 (thunderx_prefetch_tune): Likewise.
5792 (thunderx2t99_prefetch_tune): Likewise.
5793 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5794 <default_opt_level>: Set to 3.
5795 (aarch64_override_options_internal): Update to set
5796 PARAM_PREFETCH_MINIMUM_STRIDE.
5797 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5798 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5799 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5800 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5801 stride is constant and is below the minimum stride threshold.
5803 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5805 * config/arm/arm-cpus.in (mode26): Delete.
5806 (armv4): Delete mode26 reference.
5807 * config/arm/arm.c (arm_configure_build_target): Delete use of
5810 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5812 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5814 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5815 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5817 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5818 (floatunsdidf2): Ditto.
5820 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5822 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5823 (fixuns_trunc<mode>si2_avx512f): Ditto.
5824 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5825 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5826 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5828 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5830 PR rtl-optimization/79985
5831 * df-scan.c (df_insn_refs_collect): Remove special case for
5832 global registers and asm statements.
5834 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5836 * extend.texi (Global Register Variables): Rewrite the bullet list.
5837 Note that the register is available for allocation. Note that access
5838 via inline asm must use constraints. Add note about async-signal
5839 handlers. Remove paragraph about automagic register selection.
5841 2018-05-23 Richard Biener <rguenther@suse.de>
5843 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5844 of fixed offset from memset VN.
5846 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5848 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5850 (alloc_cand_and_find_basis): Initialize first_interp field.
5851 (slsr_process_mul): Modify first_interp field.
5852 (slsr_process_add): Likewise.
5853 (slsr_process_cast): Modify first_interp field for each new
5855 (slsr_process_copy): Likewise.
5856 (dump_candidate): Dump first_interp field.
5857 (replace_mult_candidate): Process all interpretations, not just
5859 (replace_rhs_if_not_dup): Likewise.
5860 (replace_one_candidate): Likewise.
5862 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5864 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5866 (aarch64_needs_frame_chain): New function.
5867 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5869 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5872 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5873 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5874 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5875 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5876 as true for strict-align.
5877 (aarch64_can_inline_p): Perform checks even when callee has no
5878 attributes to check for strict alignment.
5879 * doc/extend.texi (AArch64 Function Attributes): Document
5881 * doc/invoke.texi: (AArch64 Options): Likewise.
5883 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5885 PR tree-optimization/85853
5886 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5887 the handling of the root of the node to...
5888 (vect_slp_analyze_node_operations_1): ...this new function,
5889 and run the whole thing with the child nodes' def types
5890 set according to their SLP node's def type.
5892 2018-05-23 Richard Biener <rguenther@suse.de>
5895 * tree-data-ref.c (create_runtime_alias_checks): Defer
5896 and ignore overflow warnings.
5898 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5900 PR tree-optimization/85822
5901 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5904 2018-05-23 Richard Biener <rguenther@suse.de>
5906 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5907 memset constants via native_interpret_expr.
5909 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5912 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5914 (cgraph_node::create_alias): Likewise.
5915 (cgraph_node::get_availability): Check ifunc_resolver instead
5916 of looking up ifunc attribute.
5917 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5918 * varasm.c (do_assemble_alias): Likewise.
5919 (assemble_alias): Likewise.
5920 (default_binds_local_p_3): Likewise.
5921 * cgraph.h (cgraph_node): Add ifunc_resolver.
5922 (cgraph_node::only_called_directly_or_aliased_p): Return false
5924 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5926 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5927 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5928 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5929 instead of looking up ifunc attribute.
5931 2018-05-22 Luis Machado <luis.machado@linaro.org>
5933 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5935 2018-05-22 Martin Sebor <msebor@redhat.com>
5938 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5939 only when expasion succeeds.
5940 (expand_builtin_strcmp): Same.
5941 (expand_builtin_strncmp): Same.
5943 2018-05-22 Martin Sebor <msebor@redhat.com>
5945 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5947 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5948 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5950 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5951 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5953 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5954 Add check for legitimate_address.
5955 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5956 (aarch64_swap_ldrstr_operands): New.
5957 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5960 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5961 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5963 * config/aarch64/aarch64.md: New patterns to generate stp
5965 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5966 single words and double words.
5967 (load_pair_sw, load_pair_dw): Likewise.
5968 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5970 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5972 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5973 for different mode ldpstp and add peephole for merged zero stores.
5975 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5977 (aarch64_gen_store_pair): Rename calls to match new patterns.
5978 (aarch64_gen_load_pair): Rename calls to match new patterns.
5979 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5980 (load_pair<DREG:mode><DREG2:mode>): ... This.
5981 (store_pair<mode>): Rename to...
5982 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5983 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5985 (V_INT_EQUIV): Handle SImode.
5986 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5989 2018-05-22 Martin Sebor <msebor@redhat.com>
5992 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5993 or ajust the presumed bound on an operation to avoid unnecessary
5996 2018-05-22 Martin Sebor <msebor@redhat.com>
5998 PR tree-optimization/85826
5999 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6000 assuming that a DECL necesarily has a constant size.
6002 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6005 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6007 2018-05-22 Richard Biener <rguenther@suse.de>
6009 PR tree-optimization/85834
6010 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6011 non-constant and non-zero memset arguments.
6013 2018-05-22 Martin Liska <mliska@suse.cz>
6016 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6018 2018-05-22 Richard Biener <rguenther@suse.de>
6020 PR tree-optimization/85863
6021 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6022 comparisons when vectype is specified.
6023 (vectorizable_condition): Do not specify vectype for
6024 vect_is_simple_cond when SLP vectorizing.
6026 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6029 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6030 define __ibm128 as long double.
6031 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6032 as a distinct type when IEEE 128-bit support is enabled.
6033 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6034 128-bit types to use the correct functions.
6035 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6036 convert between 128-bit floating point types that have different
6037 modes but the same representation, instead of using gen_lowpart to
6039 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6041 (IFKF_reg): New attributes to give the register constraints for
6043 (extend<mode>tf2_internal): New insns to mark an explicit
6044 conversion between 128-bit floating point types that have a
6045 different mode but share the same representation.
6047 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6049 PR tree-optimization/85814
6050 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6051 a null return from get_strinfo when unsharing the next
6052 strinfo in the chain.
6054 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6057 * varasm.c (weak_finish): Clean up weak_decls.
6059 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6061 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6062 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6063 UNSPEC_UADALP values.
6064 * config/aarch64/iterators.md (ABAL): New int iterator.
6067 (sur): Add mappings for the above.
6068 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6070 (aarch64_<sur>abal<mode>_4): Likewise.
6071 (aarch64_<sur>adalp<mode>_3): Likewise.
6072 (<sur>sadv16qi): New define_expand.
6074 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6076 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6077 (*movdf_internal): Ditto.
6078 (*rcpsf2_sse): Ditto.
6079 (*rsqrtsf2_sse): Ditto.
6080 (*sqrt<mode>2_sse): Ditto.
6082 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6084 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6086 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6087 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6088 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6089 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6091 * config/aarch64/arm_neon.h: Likewise.
6092 * config/aarch64/iterators.md (VQ_I): New.
6094 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6096 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6098 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6101 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6103 * config/nds32/constraints.md (S): New constraint.
6104 * config/nds32/nds32.md (call_internal): Use constraint S.
6105 (call_value_internal): Likewise.
6106 (sibcall_internal): Likewise.
6107 (sibcall_value_internal): Likewise.
6109 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6110 Chung-Ju Wu <jasonwucj@gmail.com>
6112 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6115 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6116 Chung-Ju Wu <jasonwucj@gmail.com>
6118 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6119 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6120 (nds32_rtx_costs_impl): Simplify.
6121 (nds32_address_cost_impl): Simplify.
6122 (nds32_init_rtx_costs): New function.
6123 (nds32_rtx_costs_speed_prefer): Likewise.
6124 (nds32_rtx_costs_size_prefer): Likewise.
6125 (nds32_address_cost_speed_prefer): Likewise.
6126 (nds32_address_cost_speed_fwprop): Likewise.
6127 (nds32_address_cost_size_prefer): Likewise.
6128 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6129 * config/nds32/nds32.c (nds32_option_override): Use
6130 nds32_init_rtx_costs function.
6132 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6134 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6135 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6136 (TARGET_PIPELINE_N8): Likewise.
6137 (TARGET_PIPELINE_N10): Likewise.
6138 (TARGET_PIPELINE_N13): Likewise.
6139 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6141 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6143 * config/nds32/nds32-fpu.md: Update copyright year.
6145 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6147 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6149 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6151 * config/nds32/nds32.c
6152 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6153 * config/nds32/nds32.opt (minline-asm-r15): New option.
6155 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6157 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6159 * config/nds32/nds32.md (abssi2): New pattern.
6161 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6163 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6164 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6165 Merge insn pattern from sse_cvtsi2ss<round_name> and
6166 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6167 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6168 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6169 using SWI48 mode iterator.
6170 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6171 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6172 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6173 pattern from sse_cvttss2si<round_saeonly_name>
6174 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6175 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6176 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6177 using SWI48 mode iterator.
6178 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6179 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6180 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6181 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6182 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6183 using SWI48 mode iterator.
6184 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6185 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6186 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6187 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6188 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6189 SWI48 mode iterator.
6190 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6191 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6192 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6193 pattern from sse_cvttsd2si<round_saeonly_name>
6194 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6196 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6198 * config/nds32/nds32-md-auxiliary.c
6199 (nds32_valid_smw_lwm_base_p): Refine.
6200 (nds32_output_smw_single_word): Refine.
6201 (nds32_output_smw_double_word): New.
6202 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6204 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6206 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6207 (nds32_output_stack_pop): Refine.
6208 (nds32_expand_unaligned_load): Refine.
6209 (nds32_expand_unaligned_store): Refine.
6211 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6212 Chung-Ju Wu <jasonwucj@gmail.com>
6214 * config/nds32/constants.md: Add TP_REGNUM constant.
6215 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6216 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6218 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6219 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6220 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6221 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6222 and PIC code generation.
6223 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6225 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6227 * config/nds32/nds32.md: Support TLS and PIC.
6228 * config/nds32/nds32.c: Support TLS and PIC.
6229 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6230 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6233 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6235 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6236 mode with E_ prefix.
6238 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6239 Chung-Ju Wu <jasonwucj@gmail.com>
6241 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6242 * config/nds32/nds32-md-auxiliary.c
6243 (symbolic_reference_mentioned_p): New.
6244 (nds32_legitimize_ict_address): New.
6245 (nds32_expand_ict_move): New.
6246 (nds32_indirect_call_referenced_p): New.
6247 (nds32_symbol_binds_local_p): Delete.
6248 (nds32_long_call_p): Modify.
6249 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6250 * config/nds32/nds32-protos.h
6251 (symbolic_reference_mentioned_p): Declare.
6252 (nds32_legitimize_ict_address): Declare.
6253 (nds32_expand_ict_move): Declare.
6254 (nds32_indirect_call_referenced_p): Declare.
6255 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6256 (nds32_relax_group): Use nds32_ict_const_p as condition.
6257 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6258 (nds32_asm_file_start): Output ict_model directive in asm code.
6259 (nds32_legitimate_address_p): Consider indirect call.
6260 (nds32_print_operand): Consider indirect call.
6261 (nds32_print_operand_address): Consider indirect call.
6262 (nds32_insert_attributes): Handle "indirect_call" attribute.
6263 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6264 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6265 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6266 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6267 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6268 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6269 (TARGET_ICT_MODEL_SMALL): Define.
6270 (TARGET_ICT_MODEL_LARGE): Define.
6271 * config/nds32/nds32.md (movsi): Consider ict model.
6272 (call, call_value): Consider ict model.
6273 (sibcall, sibcall_value): Consider ict model.
6274 * config/nds32/nds32.opt (mict-model): New option.
6275 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6278 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6279 Monk Chiang <sh.chiang04@gmail.com>
6280 Jim Wilson <jimw@sifive.com>
6282 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6283 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6284 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6285 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6286 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6287 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6288 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6289 compute save_libcall_adjustment properly.
6290 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6291 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6292 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6293 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6294 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6295 (ABI_SPEC): Handle mabi=ilp32e.
6296 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6297 (RVE): Add RVE mask.
6298 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6299 <-march>: Add rv32e as an example.
6301 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6304 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6305 (intra_create_variable_infos): Handle C++ constructors.
6307 2018-05-18 Martin Liska <mliska@suse.cz>
6309 * passes.def: Remove a redundant pass.
6311 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6314 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6316 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6318 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6320 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6322 (armv2, armv2a, armv3, armv3m): Delete architectures.
6323 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6324 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6325 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6327 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6328 (*mulsidi3adddi): Likewise.
6329 (mulsidi3): Likewise.
6330 (*mulsidi3_nov6): Likewise.
6331 (umulsidi3): Likewise.
6332 (umulsidi3_nov6): Likewise.
6333 (umaddsidi4): Likewise.
6334 (*umulsidi3adddi): Likewise.
6335 (smulsi3_highpart): Likewise.
6336 (*smulsi3_highpart_nov6): Likewise.
6337 (umulsi3_highpart): Likewise.
6338 (*umulsi3_highpart_nov6): Likewise.
6339 * config/arm/arm.h (arm_arch3m): Delete.
6340 * config/arm/arm.c (arm_arch3m): Delete.
6341 (arm_option_override_internal): Update armv3-related comment.
6342 (arm_configure_build_target): Delete use of isa_bit_mode32.
6343 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6344 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6345 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6347 (mulusa3): Likewise.
6348 * config/arm/arm-protos.h (arm_arch3m): Delete.
6349 * config/arm/arm-tables.opt: Regenerate.
6350 * config/arm/arm-tune.md: Likewise.
6351 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6352 deleted architectures.
6354 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6356 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6357 (armv5t, armv5te): New features.
6358 (ARMv5, ARMv5e): Delete fgroups.
6359 (ARMv5t, ARMv5te): Adjust for above changes.
6361 (armv5, armv5e): Delete arches.
6362 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6364 (*call_reg_arm): Likewise.
6365 (*call_value_reg_armv5): Likewise.
6366 (*call_value_reg_arm): Likewise.
6367 (*call_symbol): Likewise.
6368 (*call_value_symbol): Likewise.
6369 (*sibcall_insn): Likewise.
6370 (*sibcall_value_insn): Likewise.
6372 (prefetch): Likewise.
6373 (define_split and define_peephole2 dependent on arm_arch5):
6375 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6377 (TARGET_ARM_QBIT): Likewise.
6378 (TARGET_DSP_MULTIPLY): Likewise.
6379 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6380 (arm_arch5, arm_arch5e): Delete.
6381 (arm_arch5t, arm_arch5te): Declare.
6382 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6383 (arm_arch5t): Declare.
6384 (arm_option_reconfigure_globals): Update for the above.
6385 (arm_options_perform_arch_sanity_checks): Update comment, replace
6386 use of arm_arch5 with arm_arch5t.
6387 (use_return_insn): Likewise.
6388 (arm_emit_call_insn): Likewise.
6389 (output_return_instruction): Likewise.
6390 (arm_final_prescan_insn): Likewise.
6391 (arm_coproc_builtin_available): Likewise.
6392 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6393 arm_arch5e with arm_arch5t and arm_arch5te.
6394 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6395 (arm_arch5t, arm_arch5te): Declare.
6396 * config/arm/arm-tables.opt: Regenerate.
6397 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6398 * config/arm/t-multilib: Likewise.
6399 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6400 instead of arm_arch5.
6401 (*call_reg_thumb1): Likewise.
6402 (*call_value_reg_thumb1_v5): Likewise.
6403 (*call_value_reg_thumb1): Likewise.
6404 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6406 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6408 2018-05-18 Martin Liska <mliska@suse.cz>
6410 PR gcov-profile/84846
6411 * doc/gcov.texi: Document -t option of gcov tool.
6413 2018-05-18 Martin Liska <mliska@suse.cz>
6415 PR gcov-profile/84846
6416 * gcov.c (print_usage): Add new -t option.
6417 (process_args): Handle the option.
6418 (generate_results): Use stdout as output when requested by
6421 2018-05-18 Martin Liska <mliska@suse.cz>
6423 PR gcov-profile/84846
6424 * coverage.c (coverage_init): Write PWD to .gcno file.
6425 * doc/gcov.texi: Document how working directory is printed.
6426 * gcov-dump.c (dump_gcov_file): Print PWD.
6427 * gcov.c (output_intermediate_file): Likewise.
6428 (read_graph_file): Read PWD string.
6429 (output_lines): Print PWD.
6431 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6434 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6435 for retval and return false if all args to phi are zero.
6437 2018-05-18 Richard Biener <rguenther@suse.de>
6439 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6441 (evrp_dom_walker::before_dom_children): Call it.
6443 2018-05-18 Richard Biener <rguenther@suse.de>
6445 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6446 results when processing array refs with variable index.
6448 2018-05-18 Toon Moene <toon@moene.org>
6450 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6451 directly after that of -floop-interchange. Indicate that both
6452 options are enabled by default when specifying -O3.
6454 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6456 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6457 iterator. Delete separate integer-mode vec_set<mode> expander.
6458 (aarch64_simd_vec_setv2di): Delete.
6459 (vec_setv2di): Delete.
6460 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6461 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6462 the "w, r" alternative.
6464 2018-05-18 Martin Liska <mliska@suse.cz>
6466 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6467 * tree-pass.h (make_pass_lower_switch_O0): New function.
6468 * tree-switch-conversion.c (node_has_low_bound): Remove.
6469 (node_has_high_bound): Likewise.
6470 (node_is_bounded): Likewise.
6471 (class pass_lower_switch): Make it a template type and create
6473 (pass_lower_switch::execute): Add template argument.
6474 (make_pass_lower_switch): New function.
6475 (make_pass_lower_switch_O0): New function.
6476 (do_jump_if_equal): Remove.
6477 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6478 all the hard work to tree optimization passes.
6480 2018-05-18 Martin Liska <mliska@suse.cz>
6482 * dbgcnt.c (limit_low): Renamed from limit.
6483 (limit_high): New variable.
6484 (dbg_cnt_is_enabled): Check for upper limit.
6485 (dbg_cnt): Adjust dumping.
6486 (dbg_cnt_set_limit_by_index): Add new argument for high
6488 (dbg_cnt_set_limit_by_name): Likewise.
6489 (dbg_cnt_process_single_pair): Parse new format.
6490 (dbg_cnt_process_opt): Use strtok.
6491 (dbg_cnt_list_all_counters): Remove 'value' and add
6493 * doc/invoke.texi: Document changes.
6495 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6497 * doc/sourcebuild.texi (scalar_all_fma): Document.
6498 * tree.def (FMA_EXPR): Delete.
6499 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6500 * internal-fn.c (ternary_direct): New macro.
6501 (expand_ternary_optab_fn): Likewise.
6502 (direct_ternary_optab_supported_p): Likewise.
6503 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6504 * builtins.c (fold_builtin_fma): Delete.
6505 (fold_builtin_3): Don't call it.
6506 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6507 * expr.c (expand_expr_real_2): Likewise.
6508 * fold-const.c (operand_equal_p): Likewise.
6509 (fold_ternary_loc): Likewise.
6510 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6511 * gimple.c (DEFTREECODE): Likewise.
6512 * gimplify.c (gimplify_expr): Likewise.
6513 * optabs-tree.c (optab_for_tree_code): Likewise.
6514 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6515 * tree-eh.c (operation_could_trap_p): Likewise.
6516 (stmt_could_throw_1_p): Likewise.
6517 * tree-inline.c (estimate_operator_cost): Likewise.
6518 * tree-pretty-print.c (dump_generic_node): Likewise.
6519 (op_code_prio): Likewise.
6520 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6521 * tree-ssa-operands.c (get_expr_operands): Likewise.
6522 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6523 * fold-const-call.h (fold_fma): Delete.
6524 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6525 CFN_FNMA and CFN_FNMS.
6527 * genmatch.c (combined_fn): New enum.
6528 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6529 (commutative_op): New function.
6530 (commutate): Use it. Handle more than 2 operands.
6531 (dt_operand::gen_gimple_expr): Use commutative_op.
6532 (parser::parse_expr): Allow :c to be used with non-binary
6533 operators if the commutative operand is known.
6534 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6535 CFN_FMS, CFN_FNMA and CFN_FNMS.
6536 (backprop::process_assign_use): Remove FMA_EXPR handling.
6537 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6538 (gen_hsa_fma): New function.
6539 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6540 IFN_FNMA and IFN_FNMS.
6541 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6542 * gimple-fold.h (follow_all_ssa_edges): Declare.
6543 * gimple-fold.c (follow_all_ssa_edges): New function.
6544 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6545 gimple_build interface and use follow_all_ssa_edges to fold the result.
6546 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6547 instead of checking for optabs directly.
6548 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6549 rather than FMA_EXPRs.
6550 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6551 call to IFN_FMA instead of an FMA_EXPR.
6553 2018-05-17 Jim Wilson <jimw@sifive.com>
6555 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6556 sign extended promoted subreg, and the range does not have the sign bit
6557 set, then do a sign extend.
6559 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6560 test, check for sign extended subreg and/or constant operands, and
6561 do a sign extend in that case.
6563 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6565 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6566 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6568 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6569 Change logics_shift_reg to logics_shift_imm.
6570 (thunderx2t99_fp_loadpair_basic): Delete.
6571 (thunderx2t99_fp_storepair_basic): Delete.
6572 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6573 (thunderx2t99_asimd_polynomial): Delete.
6574 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6575 and neon_fp_mul_d_scalar_q.
6576 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6577 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6578 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6579 (thunderx2t99_asimd_lut): Add missing tbl types.
6580 (thunderx2t99_asimd_ext): Delete.
6581 (thunderx2t99_asimd_load1_1_mult): Delete.
6582 (thunderx2t99_asimd_load1_2_mult): Delete.
6583 (thunderx2t99_asimd_load1_ldp): New.
6584 (thunderx2t99_asimd_load1): New.
6585 (thunderx2t99_asimd_load2): Add missing *load2* types.
6586 (thunderx2t99_asimd_load3): New.
6587 (thunderx2t99_asimd_load4): New.
6588 (thunderx2t99_asimd_store1_1_mult): Delete.
6589 (thunderx2t99_asimd_store1_2_mult): Delete.
6590 (thunderx2t99_asimd_store2_mult): Delete.
6591 (thunderx2t99_asimd_store2_onelane): Delete.
6592 (thunderx2t99_asimd_store_stp): New.
6593 (thunderx2t99_asimd_store1): New.
6594 (thunderx2t99_asimd_store2): New.
6595 (thunderx2t99_asimd_store3): New.
6596 (thunderx2t99_asimd_store4): New.
6598 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6600 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6601 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6603 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6604 Segher Boessenkool <segher@kernel.crashing.org>
6607 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6610 2018-05-17 Richard Biener <rguenther@suse.de>
6612 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6613 for pruning loop and prune defs feeding only already visited PHIs.
6615 2018-05-17 Richard Biener <rguenther@suse.de>
6617 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6619 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6620 Richard Biener <rguenther@suse.de>
6622 PR tree-optimization/85793
6623 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6624 for VMAT_ELEMENTWISE.
6626 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6628 * internal-fn.h (lookup_internal_fn): Declare
6629 * internal-fn.c (lookup_internal_fn): New function.
6630 * gimple.c (gimple_build_call_from_tree): Handle calls to
6632 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6633 internal function names.
6634 * tree-pretty-print.c (dump_generic_node): Likewise.
6635 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6637 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6639 * gimple-fold.h (gimple_build): Make the function forms take
6640 combined_fn rather than built_in_function.
6641 (gimple_simplify): Likewise.
6642 * gimple-match-head.c (gimple_simplify): Likewise.
6643 * gimple-fold.c (gimple_build): Likewise.
6644 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6645 rather than gimple_build_call_internal.
6646 (get_initial_defs_for_reduction): Likewise.
6647 (vect_create_epilog_for_reduction): Likewise.
6648 (vectorizable_live_operation): Likewise.
6650 2018-05-17 Martin Liska <mliska@suse.cz>
6652 * gimple-ssa-sprintf.c (format_directive): Do not use
6653 space in between 'G_' and '('.
6655 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6658 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6659 even if the mask is not all ones.
6662 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6664 (ix86_gimple_fold_builtin): Likewise.
6667 * config/i386/i386.c: Include tree-vector-builder.h.
6668 (ix86_vector_shift_count): New function.
6669 (ix86_fold_builtin): Fold shift builtins by scalar count.
6670 (ix86_gimple_fold_builtin): Likewise.
6672 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6673 _mm512_setzero): New intrinsics.
6675 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6676 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6678 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6679 code generation for cases where splatting a value is not useful.
6680 * simplify-rtx.c (simplify_ternary_operation): Simplify
6681 vec_merge across a vec_duplicate and a paradoxical subreg forming
6682 a vector mode to a vec_concat.
6684 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6686 * config.gcc: Support "goldmont-plus".
6687 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6689 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6690 PROCESSOR_GOLDMONT_PLUS.
6691 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6692 (processor_target_table): Add "goldmont-plus".
6693 (PTA_GOLDMONT_PLUS): Define.
6694 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6695 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6696 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6697 (fold_builtin_cpu): Add "goldmont-plus".
6698 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6699 (ix86_option_override_internal): Add "goldmont-plus".
6700 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6701 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6702 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6703 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6705 2018-05-17 Richard Biener <rguenther@suse.de>
6707 PR tree-optimization/85757
6708 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6709 remove defs that only feed that PHI from further processing.
6711 2018-05-16 Jim Wilson <jimw@sifive.com>
6713 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6715 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6717 2018-05-16 Mark Wielaard <mark@klomp.org>
6719 * dwarf2out.c (count_index_strings): New function.
6720 (output_indirect_strings): Call count_index_strings and generate
6721 header for dwarf_version >= 5.
6723 2018-05-16 Mark Wielaard <mark@klomp.org>
6725 * dwarf2out.c (dwarf_FORM): New function.
6726 (set_indirect_string): Use dwarf_FORM.
6727 (reset_indirect_string): Likewise.
6728 (size_of_die): Likewise.
6729 (value_format): Likewise.
6730 (output_die): Likewise.
6731 (add_skeleton_AT_string): Likewise.
6732 (output_macinfo_op): Likewise.
6733 (index_string): Likewise.
6734 (output_index_string_offset): Likewise.
6735 (output_index_string): Likewise.
6736 (count_index_strings): Likewise.
6738 2018-05-16 Carl Love <cel@us.ibm.com>
6740 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6741 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6743 2018-05-16 Martin Jambor <mjambor@suse.cz>
6745 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6746 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6748 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6750 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6751 (fnma<mode>4): Likewise.
6752 (fms<mode>4): Likewise.
6753 (fnms<mode>4): Likewise.
6754 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6755 (aarch64_fnma<mode>4): Likewise.
6756 (aarch64_fms<mode>4): Likewise.
6757 (aarch64_fnms<mode>4): Likewise.
6758 (aarch64_fnmadd<mode>4): Likewise.
6760 2018-05-16 Jason Merrill <jason@redhat.com>
6762 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6764 2018-05-16 Richard Biener <rguenther@suse.de>
6766 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6767 (dump_stmt_cost): Declare.
6768 (add_stmt_cost): Dump cost we add.
6769 (add_stmt_costs): New function.
6770 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6772 (vect_analyze_stmt): Adjust prototype.
6773 (vectorizable_condition): Likewise.
6774 (vectorizable_live_operation): Likewise.
6775 (vectorizable_reduction): Likewise.
6776 (vectorizable_induction): Likewise.
6777 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6778 cost vector to pass to vectorizable_ and record afterwards.
6779 (vect_model_reduction_cost): Take cost vector argument and adjust.
6780 (vect_model_induction_cost): Likewise.
6781 (vectorizable_reduction): Likewise.
6782 (vectorizable_induction): Likewise.
6783 (vectorizable_live_operation): Likewise.
6784 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6785 SLP_TREE_NUMBER_OF_VEC_STMTS.
6786 (vect_analyze_slp_cost_1): Remove.
6787 (vect_analyze_slp_cost): Likewise.
6788 (vect_slp_analyze_node_operations): Take visited args and
6789 a target cost vector. Avoid processing already visited stmt sets.
6790 (vect_slp_analyze_operations): Use a local cost vector to gather
6791 costs and register those of non-discarded instances.
6792 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6793 (vect_schedule_slp_instance): Remove copying of
6794 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6796 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6797 adding cost. Record cost entry location.
6798 (vect_prologue_cost_for_slp_op): Function to compute cost of
6799 a constant or invariant generated for SLP vect in the prologue,
6800 split out from vect_analyze_slp_cost_1.
6801 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6802 (vect_model_promotion_demotion_cost): Likewise.
6803 (vect_model_store_cost): Likewise, make static.
6804 (vect_model_load_cost): Likewise.
6805 (vectorizable_bswap): Add cost vector arg and adjust.
6806 (vectorizable_call): Likewise.
6807 (vectorizable_simd_clone_call): Likewise.
6808 (vectorizable_conversion): Likewise.
6809 (vectorizable_assignment): Likewise.
6810 (vectorizable_shift): Likewise.
6811 (vectorizable_operation): Likewise.
6812 (vectorizable_store): Likewise.
6813 (vectorizable_load): Likewise.
6814 (vectorizable_condition): Likewise.
6815 (vectorizable_comparison): Likewise.
6816 (can_vectorize_live_stmts): Likewise.
6817 (vect_analyze_stmt): Likewise.
6818 (vect_transform_stmt): Adjust calls to vectorizable_*.
6819 * tree-vectorizer.c: Include gimple-pretty-print.h.
6820 (dump_stmt_cost): New function.
6822 2018-05-16 Richard Biener <rguenther@suse.de>
6824 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6825 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6826 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6827 (check_name): New callback.
6828 (dse_classify_store): Track cycles via a visited bitmap of PHI
6829 defs and simplify handling of in-loop and across loop dead stores
6830 and properly fail for loop-variant refs. Handle byte-tracking with
6831 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6834 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6836 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6837 (vect_get_mask_type_for_stmt): Likewise.
6838 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6840 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6841 to determine the statement's vector type and the vector type that
6842 should be used for calculating nunits. Deal with cases in which
6843 the type has to be deferred.
6844 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6845 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6846 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6847 (vect_determine_vf_for_stmt): New functions, split out from...
6848 (vect_determine_vectorization_factor): ...here.
6849 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6850 (vect_get_mask_type_for_stmt): New functions, split out from
6851 vect_determine_vectorization_factor.
6853 2018-05-16 Richard Biener <rguenther@suse.de>
6855 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6856 verify the [VEC_]COND_EXPR embedded comparison.
6858 2018-05-15 Martin Sebor <msebor@redhat.com>
6860 PR tree-optimization/85753
6861 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6862 RECORD_TYPE in addition to ARRAY_TYPE.
6864 2018-05-15 Martin Sebor <msebor@redhat.com>
6867 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6869 2018-05-15 Richard Biener <rguenther@suse.de>
6871 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6872 add by_clobber_p one. Change algorithm to collect all defs
6873 representing uses we need to walk and try reducing them to
6874 a single one before failing.
6875 (dse_dom_walker::dse_optimize_stmt): Adjust.
6877 2018-05-13 Mark Wielaard <mark@klomp.org>
6879 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6880 (size_of_loc_descr): Likewise.
6881 (output_loc_operands): Likewise.
6882 (output_loc_operands_raw): Likewise.
6883 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6884 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6885 (hash_loc_operands): Likewise.
6886 (compare_loc_operands): Likewise.
6888 2018-05-14 Mark Wielaard <mark@klomp.org>
6890 * dwarf2out.c (count_index_addrs): New function.
6891 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6893 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6895 PR tree-optimization/83648
6896 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6897 return value as malloc candidate.
6899 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6902 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6903 param as true in call to suggest_attribute.
6905 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6907 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6908 -mreadonly-in-sdata.
6910 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6912 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6914 (aarch64_crypto_aesd_fused): Likewise.
6916 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6918 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6919 (movsi_aarch64): Likewise.
6920 (load_pairsi): Likewise.
6921 (load_pairdi): Likewise.
6922 (store_pairsi): Likewise.
6923 (store_pairdi): Likewise.
6924 (load_pairsf): Likewise.
6925 (load_pairdf): Likewise.
6926 (store_pairsf): Likewise.
6927 (store_pairdf): Likewise.
6928 (zero_extend): Likewise.
6929 (trunc): Swap alternatives.
6930 (fcvt_target): Add '?' to prefer w over r.
6932 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6935 * config/i386/i386.md: Disallow non-commutative arithmetics in
6936 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6937 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6938 in the peephole2 before it.
6940 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6942 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6943 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6944 (ix86_handle_option): Handle -mcldemote.
6945 * config.gcc: New header.
6946 * config/i386/cldemoteintrin.h: New file.
6947 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6948 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6950 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6951 OPTION_MASK_ISA_CLDEMOTE.
6952 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6953 (ix86_valid_target_attribute_inner_p): Ditto.
6954 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6955 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6956 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6957 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6958 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6960 * config/i386/i386.opt: Add -mcldemote.
6961 * config/i386/x86intrin.h: New header.
6962 * doc/invoke.texi: Add -mcldemote.
6964 2018-05-14 Richard Biener <rguenther@suse.de>
6966 * doc/match-and-simplify.texi: Adjust :s documentation.
6968 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6970 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6971 intended memcpy size.
6972 (REORDER_45): Likewise.
6974 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6976 * sort.cc: New file.
6977 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6978 * vec.c (qsort_chk): Use gcc_qsort.
6979 * Makefile.in (OBJS-libcommon): Add sort.o.
6980 (build/sort.o): New target. Use it...
6981 (BUILD_RTL): ... here, and...
6982 (build/gencfn-macros): ... here, and...
6983 (build/genmatch): ... here.
6985 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6986 Chung-Ju Wu <jasonwucj@gmail.com>
6988 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6989 * config/nds32/nds32-graywolf.md: New file.
6990 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6991 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6993 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6994 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6995 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6996 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6997 * config/nds32/pipelines.md: Include n15 settings.
6999 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7000 Chung-Ju Wu <jasonwucj@gmail.com>
7002 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7003 * config/nds32/nds32-n13.md: New file.
7004 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7005 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7007 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7008 * config/nds32/nds32.md (pipeline_model): Add n13.
7009 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7010 * config/nds32/pipelines.md: Include n13 settings.
7012 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7013 Chung-Ju Wu <jasonwucj@gmail.com>
7015 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7016 * config/nds32/nds32-n10.md: New file.
7017 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7018 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7020 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7021 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7022 * config/nds32/nds32.md (pipeline_model): Add n10.
7023 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7024 * config/nds32/pipelines.md: Include n10 settings.
7026 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7027 Kito Cheng <kito.cheng@gmail.com>
7028 Chung-Ju Wu <jasonwucj@gmail.com>
7030 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7031 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7032 Add enum values for DSP extension instructions.
7033 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7035 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7036 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7038 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7039 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7040 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7041 * config/nds32/nds32-intrinsic.md: Likewise.
7042 * config/nds32/nds32_intrinsic.h: Likewise.
7043 * config/nds32/nds32-md-auxiliary.c: Likewise.
7044 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7045 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7046 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7047 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7048 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7049 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7050 TYPE_DMAC in switch statement.
7051 * config/nds32/nds32.c: New checking and implementation for DSP
7052 extension instructions.
7053 * config/nds32/nds32.h: Likewise.
7054 * config/nds32/nds32.md: Likewise.
7055 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7056 * config/nds32/predicates.md: Implement new predicates for DSP
7059 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7061 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7062 Reformat alternatives and attributes so it is easier to identify
7063 which constraints/attributes go with which instruction.
7064 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7065 (mov<mode>_softfloat32, FMOVE64): Likewise.
7066 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7067 (mov<mode>_softfloat64, FMOVE64): Likewise.
7069 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7071 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7073 (Basic PowerPC Built-in Functions): The new name of the
7074 subsection previously known as "PowerPC Built-in Functions".
7075 (Basic PowerPC Built-in Functions Available on all Configurations):
7077 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7078 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7079 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7080 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7082 2018-05-11 Martin Jambor <mjambor@suse.cz>
7085 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7088 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7091 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7093 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7095 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7096 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7097 (ix86_handle_option): Handle -mwaitpkg.
7098 * config.gcc: New header.
7099 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7100 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7101 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7103 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7104 OPTION_MASK_ISA_WAITPKG.
7105 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7106 (ix86_option_override_internal): Add PTA_WAITPKG.
7107 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7108 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7109 IX86_BUILTIN_TPAUSE.
7110 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7111 __builtin_ia32_umwait and __builtin_ia32_tpause.
7112 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7113 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7114 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7115 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7116 UNSPECV_TPAUSE): New.
7117 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7118 * config/i386/i386.opt: Add -mwaitpkg.
7119 * config/i386/waitpkgintrin.h: New file.
7120 * config/i386/x86intrin.h: New header.
7121 * doc/invoke.texi: Add -mwaitpkg.
7123 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7126 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7128 (cortex-m0): Use armv6s-m isa.
7129 (cortex-m0plus): Likewise.
7130 (cortex-m1): Likewise.
7131 (cortex-m0.small-multiply): Likewise.
7132 (cortex-m0plus.small-multiply): Likewise.
7133 (cortex-m1.small-multiply): Likewise.
7135 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7136 Jakub Jelinek <jakub@redhat.com>
7138 PR tree-optimization/85692
7139 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7140 source permute as well.
7142 2018-05-11 Martin Liska <mliska@suse.cz>
7145 * doc/extend.texi: Document LLVM style format for no_sanitize
7148 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7150 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7151 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7152 (mode_supports_vsx_dform_quad): Likewise.
7153 (mode_supports_vmx_dform): Move these functions to be next to the
7154 other mode_supports functions.
7155 (mode_supports_dq_form): Likewise.
7156 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7157 mode_supports_dq_form.
7158 (reg_offset_addressing_ok_p): Likewise.
7159 (offsettable_ok_by_alignment): Likewise.
7160 (rs6000_legitimate_offset_address_p): Likewise.
7161 (legitimate_lo_sum_address_p): Likewise.
7162 (rs6000_legitimize_address): Likewise.
7163 (rs6000_legitimize_reload_address): Likewise.
7164 (rs6000_secondary_reload_inner): Likewise.
7165 (rs6000_preferred_reload_class): Likewise.
7166 (rs6000_output_move_128bit): Likewise.
7168 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7170 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7171 Generate SImode target register for null target.
7172 <case IX86_BUILTIN_XGETBV>: Ditto.
7173 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7174 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7176 2018-05-10 Carl Love <cel@us.ibm.com>
7178 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7179 dcbtt and dcbtstt if operands[2] is 0.
7181 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7184 * config/i386/sse.md (usadv64qi): New expander.
7186 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7188 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7189 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7190 -maltivec=be support.
7191 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7192 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7193 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7194 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7195 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7196 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7197 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7198 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7199 altivec_vsumsws): Adjust.
7200 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7201 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7202 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7204 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7205 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7206 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7207 (altivec_lve<VI_char>x): Delete expand.
7208 (*altivec_lve<VI_char>x_internal): Rename to...
7209 (altivec_lve<VI_char>x): ... this.
7210 (altivec_lvxl_<mode>): Delete expand.
7211 (*altivec_lvxl_<mode>_internal): Rename to ...
7212 (altivec_lvxl_<mode>): ... this.
7213 (altivec_stvxl_<mode>): Delete expand.
7214 (*altivec_stvxl_<mode>_internal): Rename to ...
7215 (altivec_stvxl_<mode>): ... this.
7216 (altivec_stve<VI_char>x): Delete expand.
7217 (*altivec_stve<VI_char>x_internal): Rename to ...
7218 (altivec_stve<VI_char>x): ... this.
7219 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7220 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7221 reduc_plus_scal_<mode>): Adjust.
7222 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7224 (rs6000_cpu_cpp_builtins): Adjust.
7225 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7226 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7227 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7228 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7229 -maltivec=be support.
7230 (rs6000_split_vec_extract_var): Adjust.
7231 (rs6000_split_v4si_init): Adjust.
7232 (swap_selector_for_mode): Delete.
7233 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7234 altivec_expand_stvex_be): Delete.
7235 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7236 -maltivec=be support.
7237 (rs6000_gimple_fold_builtin): Ditto.
7238 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7240 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7241 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7242 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7243 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7244 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7245 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7246 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7247 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7248 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7249 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7250 anonymous split): Adjust.
7251 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7252 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7254 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7256 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7257 when --with-gxx-include-dir is also specified.
7258 * configure: Regenerate.
7260 2018-05-09 Jim Wilson <jimw@sifive.com>
7263 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7264 * config/riscv/t-withmultilib: New.
7265 * config/riscv/withmultilib.h: New.
7266 * doc/install.texi: Document RISC-V --with-multilib-list support.
7268 2018-05-09 Richard Biener <rguenther@suse.de>
7270 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7272 (vect_bb_vectorization_profitable_p): Adjust. Compute
7273 actual scalar cost using the cost vector and the add_stmt_cost
7276 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7278 PR rtl-optimization/85645
7279 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7280 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7282 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7284 PR rtl-optimization/85645
7285 * shrink-wrap.c (spread_components): Return a boolean saying if
7286 anything was changed.
7287 (try_shrink_wrapping_separate): Iterate spread_components until
7288 nothing changes anymore.
7290 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7292 PR rtl-optimization/85645
7293 * regrename.c (build_def_use): Also kill the chains that include the
7294 destination of a REG_CFA_REGISTER note.
7296 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7298 PR rtl-optimization/85645
7299 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7300 insn that has a REG_CFA_REGISTER note.
7302 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7304 * cfgexpand.c (expand_clobber): New function.
7305 (expand_gimple_stmt_1): Use it.
7306 * tree-vect-stmts.c (vect_clobber_variable): New function,
7308 (vectorizable_simd_clone_call): ...here.
7309 (vectorizable_store): Emit a clobber either side of an
7310 IFN_STORE_LANES sequence.
7311 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7313 2018-05-09 Tom de Vries <tom@codesourcery.com>
7316 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7317 (define_insn "trap_if_false"): Add exit after trap.
7319 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7321 PR rtl-optimization/85638
7322 * bb-reorder.c: Include common/common-target.h.
7323 (create_forwarder_block): New function extracted from...
7324 (fix_up_crossing_landing_pad): ...here. Rename into...
7325 (dw2_fix_up_crossing_landing_pad): ...this.
7326 (sjlj_fix_up_crossing_landing_pad): New function.
7327 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7328 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7329 from both partitions and exit the loop after one iteration.
7331 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7334 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7336 (Basic PowerPC Built-in Functions): The new name of the
7337 subsection previously known as "PowerPC Built-in Functions".
7338 (Basic PowerPC Built-in Functions Available on all Configurations):
7340 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7342 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7344 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7346 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7349 2018-05-08 Jim Wilson <jimw@sifive.com>
7351 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7352 (LD_EMUL_SUFFIX): New.
7353 (LINK_SPEC): Use it.
7355 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7357 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7359 (Basic PowerPC Built-in Functions): The new name of the
7360 subsection previously known as "PowerPC Built-in Functions".
7361 (Basic PowerPC Built-in Functions Available on all Configurations):
7363 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7365 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7367 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7369 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7372 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7375 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7376 after cmpelim optimization.
7378 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7380 * config.gcc: Support "goldmont".
7381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7382 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7384 * config/i386/i386.c (m_GOLDMONT): Define.
7385 (processor_target_table): Add "goldmont".
7386 (PTA_GOLDMONT): Define.
7387 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7388 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7389 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7390 (fold_builtin_cpu): Add "goldmont".
7391 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7392 (ix86_option_override_internal): Add "goldmont".
7393 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7394 (processor_type): Add PROCESSOR_GOLDMONT.
7395 * config/i386/i386.md: Add CPU "glm".
7396 * config/i386/glm.md: New file.
7397 * config/i386/x86-tune.def: Add m_GOLDMONT.
7398 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7400 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7403 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7405 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7406 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7407 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7410 * config/i386/i386.c (ix86_fold_builtin): Handle
7411 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7414 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7415 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7417 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7420 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7421 (check_arch): Likewise.
7422 (check_fpu): Return the result rather than printing it.
7423 (end arch): Fix operator precedence.
7424 (end cpu): Likewise.
7425 (END): Print the result from check_fpu.
7427 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7428 Alan Hayward <alan.hayward@arm.com>
7429 David Sherwood <david.sherwood@arm.com>
7431 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7432 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7433 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7434 (*fcmuo<mode>_and): New patterns.
7436 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7438 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7439 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7440 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7441 (cmp_op, sve_imm_con): New code attributes.
7442 (SVE_COND_INT_CMP, imm_con): Delete.
7443 (cmp_op): Remove above unspecs from int attribute.
7444 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7446 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7447 comparison-specific unspecs.
7448 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7449 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7450 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7451 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7452 (*vec_fcm<cmp_op><mode>): Rename to...
7453 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7454 (*vec_fcmuo<mode>): Rename to...
7455 (*fcmuo<mode>): ...this and adjust likewise.
7456 (*pred_fcm<cmp_op><mode>): New pattern.
7457 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7458 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7460 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7462 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7463 (aarch64_emit_sve_predicated_cond): New function.
7464 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7465 (aarch64_emit_unspec_cond_or): Replace with...
7466 (aarch64_emit_sve_or_conds): ...this new function. Use
7467 aarch64_emit_sve_ptrue_op for the individual comparisons and
7468 aarch64_emit_binop to OR them together.
7469 (aarch64_emit_inverted_unspec_cond): Replace with...
7470 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7471 aarch64_emit_sve_ptrue_op for the comparison and
7472 aarch64_emit_unop to invert the result.
7473 (aarch64_expand_sve_vec_cmp_float): Update after the above
7474 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7476 2018-05-07 Nathan Sidwell <nathan@acm.org>
7478 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7479 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7480 (Backwards Compatibility): Likewise.
7482 2018-05-07 Luis Machado <luis.machado@linaro.org>
7486 2018-05-07 Luis Machado <luis.machado@linaro.org>
7488 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7489 <prefetch_dynamic_strides>: New const bool field.
7490 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7491 prefetch_dynamic_strides.
7492 (exynosm1_prefetch_tune): Likewise.
7493 (thunderxt88_prefetch_tune): Likewise.
7494 (thunderx_prefetch_tune): Likewise.
7495 (thunderx2t99_prefetch_tune): Likewise.
7496 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7498 (aarch64_override_options_internal): Update to set
7499 PARAM_PREFETCH_DYNAMIC_STRIDES.
7500 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7501 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7502 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7503 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7504 prefetch-dynamic-strides setting.
7506 2018-05-07 Luis Machado <luis.machado@linaro.org>
7508 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7509 <minimum_stride>: New const int field.
7510 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7511 minimum_stride field.
7512 (exynosm1_prefetch_tune): Likewise.
7513 (thunderxt88_prefetch_tune): Likewise.
7514 (thunderx_prefetch_tune): Likewise.
7515 (thunderx2t99_prefetch_tune): Likewise.
7516 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7517 (aarch64_override_options_internal): Update to set
7518 PARAM_PREFETCH_MINIMUM_STRIDE.
7519 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7520 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7521 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7522 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7523 stride is constant and is below the minimum stride threshold.
7525 2018-05-07 Luis Machado <luis.machado@linaro.org>
7527 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7530 2018-05-07 Luis Machado <luis.machado@linaro.org>
7532 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7533 <prefetch_dynamic_strides>: New const bool field.
7534 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7535 prefetch_dynamic_strides.
7536 (exynosm1_prefetch_tune): Likewise.
7537 (thunderxt88_prefetch_tune): Likewise.
7538 (thunderx_prefetch_tune): Likewise.
7539 (thunderx2t99_prefetch_tune): Likewise.
7540 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7542 (aarch64_override_options_internal): Update to set
7543 PARAM_PREFETCH_DYNAMIC_STRIDES.
7544 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7545 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7546 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7547 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7548 prefetch-dynamic-strides setting.
7550 2018-05-07 Luis Machado <luis.machado@linaro.org>
7552 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7553 <minimum_stride>: New const int field.
7554 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7555 minimum_stride field.
7556 (exynosm1_prefetch_tune): Likewise.
7557 (thunderxt88_prefetch_tune): Likewise.
7558 (thunderx_prefetch_tune): Likewise.
7559 (thunderx2t99_prefetch_tune): Likewise.
7560 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7561 (aarch64_override_options_internal): Update to set
7562 PARAM_PREFETCH_MINIMUM_STRIDE.
7563 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7564 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7565 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7566 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7567 stride is constant and is below the minimum stride threshold.
7569 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7572 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7573 the type is addressable. Don't force op into register if it has
7576 2018-05-05 Roland McGrath <mcgrathr@google.com>
7579 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7580 any section for which we don't know a specific type it should have,
7581 regardless of name. Previously this was done only for the exact
7582 names ".init_array", ".fini_array", and ".preinit_array".
7583 (default_elf_asm_named_section): Add comment about
7584 relationship with default_section_type_flags and SECTION_NOTYPE.
7585 (get_section): Don't consider it a type conflict if one side has
7586 SECTION_NOTYPE and the other doesn't, as long as neither has the
7587 SECTION_BSS et al used in the default_section_type_flags logic.
7589 2018-05-05 Tom de Vries <tom@codesourcery.com>
7592 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7593 (workaround_barsyncs): New function.
7594 (nvptx_reorg): Use workaround_barsyncs.
7595 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7596 (define_expand "nvptx_membar_cta"): New define_expand.
7597 (define_insn "*nvptx_membar_cta"): New insn.
7599 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7601 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7602 To improve optimization opportunities.
7603 * builtin-types.def: The new needed builtin types for the above.
7605 2018-05-04 Richard Biener <rguenther@suse.de>
7607 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7608 * gimple-ssa-store-merging.c
7609 (imm_store_chain_info::output_merged_store): Remove redundant create,
7610 release split_store vector contents on failure.
7611 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7612 scalar stmt vector on cache hit.
7614 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7616 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7618 * config.gcc (powerpc-xilinx-eabi*): Remove.
7619 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7621 (fusion_addis_mem_combo_load): Ditto.
7622 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7624 (rs6000_cpu_cpp_builtins): Ditto.
7625 * config/rs6000/rs6000-linux.c
7626 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7627 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7628 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7630 (rs6000_setup_reg_addr_masks): Ditto.
7631 (rs6000_init_hard_regno_mode_ok): Ditto.
7632 (rs6000_option_override_internal): Ditto.
7633 (legitimate_lo_sum_address_p): Ditto.
7634 (rs6000_legitimize_address): Ditto.
7635 (rs6000_legitimize_reload_address): Ditto.
7636 (rs6000_legitimate_address_p): Ditto.
7637 (abi_v4_pass_in_fpr): Ditto.
7638 (setup_incoming_varargs): Ditto.
7639 (rs6000_gimplify_va_arg): Ditto.
7640 (rs6000_split_multireg_move): Ditto.
7641 (rs6000_savres_strategy): Ditto.
7642 (rs6000_emit_prologue_components): Ditto.
7643 (rs6000_emit_epilogue_components): Ditto.
7644 (rs6000_emit_prologue): Ditto.
7645 (rs6000_emit_epilogue): Ditto.
7646 (rs6000_elf_file_end): Ditto.
7647 (rs6000_function_value): Ditto.
7648 (rs6000_libcall_value): Ditto.
7649 * config/rs6000/rs6000.h: Ditto.
7650 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7651 (TARGET_MINMAX): ... this. New.
7652 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7653 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7654 (*movsi_internal1_single): Delete.
7655 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7656 mfpu=, mxilinx-fpu): Delete.
7657 * config/rs6000/singlefp.h: Delete.
7658 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7659 * config/rs6000/t-rs6000: Ditto.
7660 * config/rs6000/t-xilinx: Delete.
7661 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7662 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7663 (VStype_simple): Delete.
7664 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7665 * config/rs6000/xfpu.h: Delete.
7666 * config/rs6000/xfpu.md: Delete.
7667 * config/rs6000/xilinx.h: Delete.
7668 * config/rs6000/xilinx.opt: Delete.
7669 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7670 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7672 2018-05-04 Tom de Vries <tom@codesourcery.com>
7675 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7678 2018-05-04 Richard Biener <rguenther@suse.de>
7681 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7682 (expand_complex_div_wide): Likewise.
7683 (expand_complex_operations_1): Likewise.
7684 (expand_complex_libcall): Preserve EH info of the original stmt.
7685 (tree_lower_complex): Handle removed blocks.
7686 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7687 on complex multiplication and division libcall builtins.
7689 2018-05-04 Richard Biener <rguenther@suse.de>
7692 * fold-const.c (negate_expr_p): Restrict negation of operand
7693 zero of a division to when we know that can happen without
7695 (fold_negate_expr_1): Likewise.
7697 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7700 * real.h (real_nextafter): Declare.
7701 * real.c (real_nextafter): New function.
7702 * fold-const-call.c (fold_const_nextafter): New function.
7703 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7704 CASE_CFN_NEXTTOWARD.
7705 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7706 even when arg1_mode is different from arg0_mode.
7708 2018-05-03 Nathan Sidwell <nathan@acm.org>
7710 * doc/extend.texi (Deprecated Features): Remove
7712 (Backwards Compatibility): Likewise.
7713 * doc/invoke.texi (C++ Language Options): Likewise.
7714 (C++ Dialect Options): Likewise.
7716 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7719 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7720 _mm512_mask_mullox_epi64): New intrinsics.
7722 2018-05-03 Tom de Vries <tom@codesourcery.com>
7725 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7726 dump files): Add offload-tree.
7728 2018-05-03 Richard Biener <rguenther@suse.de>
7730 PR tree-optimization/85615
7731 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7732 to loops not nested in BBs loop father to avoid creating multi-entry
7735 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7737 PR tree-optimization/70291
7738 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7739 arguments. Change return type to tree. Emit libcall as a new
7740 statement rather than replacing existing one when inplace_p is true.
7741 (expand_complex_multiplication_components): New function.
7742 (expand_complex_multiplication): Expand floating-point complex
7743 multiplication using the above.
7744 (expand_complex_division): Rename inner_type parameter to type.
7745 Update expand_complex_libcall call-site.
7746 (expand_complex_operations_1): Update expand_complex_multiplication
7747 and expand_complex_division call-sites.
7749 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7752 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7753 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7754 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7755 the highest significant bit of the shift count mask is clear. In
7756 check whether and[sq]i3 is needed verify that all significant bits
7757 of the shift count other than the highest are set.
7759 2018-05-02 Tom de Vries <tom@codesourcery.com>
7762 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7763 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7764 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7765 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7766 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7767 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7768 __builtin_goacc_parlevel_size.
7770 2018-05-02 Richard Biener <rguenther@suse.de>
7772 PR tree-optimization/85597
7773 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7774 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7777 2018-05-02 Tom de Vries <tom@codesourcery.com>
7780 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7781 dump files): Add ltrans-tree.
7783 2018-05-02 Tom de Vries <tom@codesourcery.com>
7786 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7787 dump files): Add wpa-ipa.
7789 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7791 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7792 powerpc*-*-linux*paired* target.
7793 * config/rs6000/750cl.h: Delete.
7794 * config/rs6000/paired.h: Delete.
7795 * config/rs6000/paired.md: Delete.
7796 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7798 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7799 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7800 comment. Remove paired float support.
7801 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7802 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7804 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7805 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7807 * config/rs6000/rs6000.c: Remove paired float support.
7808 (paired_expand_vector_init, paired_expand_vector_move,
7809 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7810 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7811 paired_expand_builtin, paired_expand_predicate_builtin,
7812 paired_init_builtins): Delete.
7813 * config/rs6000/rs6000.h: Remove paired float support.
7814 * config/rs6000/rs6000.md: Remove paired float support.
7815 (move_from_CR_ov_bit): Delete.
7816 * config/rs6000/rs6000.opt (mpaired): Delete.
7817 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7818 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7820 2018-05-02 Richard Biener <rguenther@suse.de>
7823 * gimplify.c (gimplify_save_expr): When in SSA form allow
7824 SAVE_EXPRs to compute to SSA vars.
7826 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7829 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7830 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7831 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7832 clobber operands[2], instead use a new pseudo. Formatting fixes.
7834 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7836 PR tree-optimization/85586
7837 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7838 exit early for statements in the same group if the accesses are
7841 2018-05-02 Tom de Vries <tom@codesourcery.com>
7844 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7847 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7849 PR tree-optimization/85143
7850 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7852 2018-05-01 Tom de Vries <tom@codesourcery.com>
7855 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7856 not found" error message.
7858 2018-05-01 Tom de Vries <tom@codesourcery.com>
7861 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7862 * vec.c (test_ordered_remove_if): New function.
7863 (vec_c_tests): Call test_ordered_remove_if.
7864 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7865 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7866 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7867 VEC_ORDERED_REMOVE_IF.
7869 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7871 PR tree-optimization/82665
7872 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7873 pointer subtraction where arguments come from a memchr call.
7875 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7877 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7878 --push-state --as-needed and --pop-state instead of --as-needed and
7879 --no-as-needed if ld supports it.
7880 * configure: Regenerated.
7883 * doc/install.texi2html: Replace _002d with - and _002a with * in
7884 generated html files using sed.
7886 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7889 * gcc-rich-location.c (blank_line_before_p): New function.
7890 (use_new_line): New function.
7891 (gcc_rich_location::add_fixit_insert_formatted): New function.
7892 * gcc-rich-location.h
7893 (gcc_rich_location::add_fixit_insert_formatted): New function.
7895 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7897 * selftest.c (assert_streq): Rename "expected" and "actual" to
7898 "val1" and "val2". Extend NULL-handling to cover both inputs
7899 symmetrically, while still requiring both to be non-NULL for a pass.
7900 * selftest.h (assert_streq): Rename "expected" and "actual" to
7902 (ASSERT_EQ): Likewise.
7903 (ASSERT_EQ_AT): Likewise.
7904 (ASSERT_KNOWN_EQ): Likewise.
7905 (ASSERT_KNOWN_EQ_AT): Likewise.
7906 (ASSERT_NE): Likewise.
7907 (ASSERT_MAYBE_NE): Likewise.
7908 (ASSERT_MAYBE_NE_AT): Likewise.
7909 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7910 the assertion to pass.
7911 (ASSERT_STREQ_AT): Likewise.
7913 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7915 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7916 interaction with -pie.
7918 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7920 * selftest.h: Fix alphabetization of per-source-file selftest
7923 2018-04-30 Jason Merrill <jason@redhat.com>
7925 PR c++/61982 - dead stores to destroyed objects.
7926 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7929 2018-04-30 Jason Merrill <jason@redhat.com>
7931 * tree.c (build_clobber): New.
7932 * tree.h: Declare it.
7933 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7935 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7937 * diagnostic-show-locus.c (layout::layout): Update for
7938 location_get_source_line returning a char_span.
7939 (struct char_span): Move to input.h.
7940 (struct correction): Update for fields in char_span becoming
7942 (struct source_line): Update for location_get_source_line
7943 returning a char_span.
7944 (layout::print_line): Likewise.
7945 * edit-context.c (edited_file::print_content): Likewise.
7946 (edited_file::print_diff_hunk): Likewise.
7947 (edited_file::print_run_of_changed_lines): Likewise.
7948 (edited_file::get_num_lines): Likewise.
7949 (edited_line::edited_line): Likewise.
7950 * final.c (asm_show_source): Likewise.
7951 * input.c (location_get_source_line): Convert return type
7952 from const char * to char_span, losing the final "line_len"
7954 (dump_location_info): Update for the above.
7955 (get_substring_ranges_for_loc): Likewise. Use a char_span
7956 when handling the literal within the line.
7957 (test_reading_source_line): Update for location_get_source_line
7958 returning a char_span.
7959 * input.h (class char_span): Move here from
7960 diagnostic-show-locus.c, converting from a struct to a class.
7961 Make data members private.
7962 (char_span::operator bool): New.
7963 (char_span::length): New.
7964 (char_span::get_buffer): New.
7965 (char_span::operator[]): New.
7966 (char_span::subspan): Make const.
7967 (char_span::xstrdup): New.
7968 (location_get_source_line): Convert return type from const char *
7969 to char_span, losing the final "line_size" param.
7971 2018-04-30 Jan Hubicka <jh@suse.cz>
7973 * lto-wrapper.c (ltrans_priorities): New static var.
7974 (cmp_priority): New.
7975 (run_gcc): Read priorities and if doing parallel build order
7976 the Makefile by them.
7978 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7980 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7982 2018-04-30 Richard Biener <rguenther@suse.de>
7984 * tree-cfg.c (verify_address): Remove base argument, add
7985 flag whether to check TREE_ADDRESSABLE and do that.
7986 (verify_expr): Remove.
7987 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7988 (verify_gimple_assign_single): Likewise.
7989 (verify_gimple_switch): Likewise.
7990 (verify_expr_location_1): Dereference tp once. Add (disabled)
7991 piece from verify_expr.
7992 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7994 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7996 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7998 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8000 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8001 (small_data_pattern): Likewise.
8002 (arc_rewrite_small_data): Likewise.
8003 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8004 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8005 (get_symbol_alignment): New function.
8006 (legitimate_small_data_address_p): Likewise.
8007 (legitimate_scaled_address): Update, call
8008 legitimate_small_data_address_p.
8009 (output_sdata): New static variable.
8010 (arc_print_operand): Update how we handle small data operands.
8011 (arc_print_operand_address): Likewise.
8012 (arc_legitimate_address_p): Update, use
8013 legitimate_small_data_address_p.
8014 (arc_rewrite_small_data_p): Remove.
8015 (arc_rewrite_small_data_1): Likewise.
8016 (arc_rewrite_small_data): Likewise.
8017 (small_data_pattern): Likewise.
8018 (compact_sda_memory_operand): Update to use
8019 legitimate_small_data_address_p and get_symbol_alignment.
8020 (prepare_move_operands): Don't rewite sdata pattern.
8021 (prepare_extend_operands): Remove.
8022 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8024 (zero_extendqisi2): Likewise.
8025 (zero_extendhisi2): Likewise.
8026 (extendqihi2): Likewise.
8027 (extendqisi2): Likewise.
8028 (extendhisi2): Likewise.
8032 * config/arc/constraints.md (Usd): Change it to memory constraint.
8034 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8036 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8037 as source of std instructions.
8038 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8039 allow 6-bit constants as source for store instructions.
8040 (movdi_insn): Update instruction pattern to allow 6-bit constants
8041 as source for store instructions.
8043 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8045 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8047 2018-04-30 Nathan Sidwell <nathan@acm.org>
8048 Sandra Loosemore <sandra@codesourcery.com>
8050 * dumpfile.c (dump_open): Allow '-' for stdout.
8051 * doc/invoke.texi (Developer Options): Document dump filename
8052 determination early. Document stdin/stdout selection.
8054 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8056 Microblaze Target: PIC data text relative
8058 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8059 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8061 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8062 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8063 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8064 New addressing mode for data-text relative position indepenedent code.
8065 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8066 'ADDRESS_SYMBOLIC_TXT_REL'.
8067 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8068 (microblaze_legitimate_pic_operand): Exclude function calls from
8069 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8070 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8072 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8073 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8074 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8075 for 'address + offset'.
8076 (microblaze_expand_prologue): Add new function prologue call for
8078 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8079 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8080 table in case of TARGET_PIC_DATA_TEXT_REL.
8081 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8082 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8083 Add new macros 'UNSPEC_TEXT',
8084 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8085 + exclude function calls from 'UNSPEC_PLT' in case of data text
8087 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8088 new target hook for generating address diff vector tables in case of
8090 * doc/tm.texi : Regenerate.
8091 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8092 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8093 of addr diff vector generation.
8094 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8095 target hook definition.
8096 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8097 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8098 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8099 option for data text relative.
8101 2018-04-30 Richard Biener <rguenther@suse.de>
8103 * tree-chrec.h (evolution_function_is_constant_p): Remove
8105 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8107 2018-04-30 Richard Biener <rguenther@suse.de>
8110 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8112 2018-04-30 Richard Biener <rguenther@suse.de>
8114 PR tree-optimization/28364
8115 PR tree-optimization/85275
8116 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8117 copying first exit test.
8119 2018-04-28 Mark Wielaard <mark@klomp.org>
8121 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8123 (dwarf_AT): Handle DW_AT_addr_base.
8124 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8126 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8129 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8130 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8131 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8132 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8134 2018-04-28 Richard Biener <rguenther@suse.de>
8136 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8137 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8138 to reflect use. Only add interesting stmts.
8140 2018-04-27 Martin Jambor <mjambor@suse.cz>
8143 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8144 the jump function allows for passing through aggregate values.
8146 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8148 * input.h (in_system_header_at): Convert from macro to inline
8150 (from_macro_expansion_at): Likewise.
8151 (from_macro_definition_at): Likewise.
8153 2018-04-27 Jeff Law <law@redhat.com>
8155 * config.gcc: Mark tile* targets as deprecated/obsolete.
8157 2018-04-27 Richard Biener <rguenther@suse.de>
8159 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8162 2018-04-27 Richard Biener <rguenther@suse.de>
8164 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8166 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8168 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8169 with Yd constraint. Set "preferred_for_speed" attribute from
8170 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8172 (*movdi_internal): Ditto.
8173 (movti_interunit splitters): Remove
8174 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8175 (movdi_interunit splitters): Ditto.
8176 * config/i386/constraints.md (Ye): Remove.
8177 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8179 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8182 * config/aarch64/constraints.md (Usg): Limit to 31.
8185 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8187 PR tree-optimization/85529
8188 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8189 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8190 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8191 zero extension or masking of the MSB bit.
8192 (optimize_range_tests): Add FIRST_BB argument, pass it through
8193 to optimize_range_tests_var_bound.
8194 (maybe_optimize_range_tests, reassociate_bb): Adjust
8195 optimize_range_tests callers.
8197 2018-04-26 Richard Biener <rguenther@suse.de>
8198 Jakub Jelinek <jakub@redhat.com>
8200 * cgraph.h (symbol_table): Just declare debug method here.
8201 * symtab.c (symbol_table::debug): Define.
8203 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8205 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8207 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8209 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8210 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8211 (*movdi_internal): Substitute Yi and Yj constraint with x
8212 and Ym and Yn constraint with y constraint. Update "isa"
8213 attribute and set "preferred_for_speed" attribute from
8214 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8215 (*movsi_internal): Ditto.
8216 (*movdf_internal): Ditto.
8217 (*movsf_internal): Ditto.
8218 (*zero_extendsidi2): Ditto.
8219 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8220 (sse2_loadld): Ditto.
8221 (*vec_extract<ssevecmodelower>_0): Ditto.
8222 (*vec_extractv4si_0_zext_sse4): Ditto.
8223 (vec_concatv2di): Ditto.
8224 (*vec_dup<mode>): Ditto.
8225 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8226 * config/i386/constraints.md (Yi): Remove.
8231 2018-04-26 Nathan Sidwell <nathan@acm.org>
8233 * dumpfile.c (dump_open): New.
8234 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8235 (dump_finish): Detect stdio/stderr by value not name.
8237 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8239 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8241 2018-04-26 Tom de Vries <tom@codesourcery.com>
8244 * config/nvptx/nvptx.c (verify_neutering_jumps)
8245 (verify_neutering_labels): New function
8246 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8248 2018-04-26 Tom de Vries <tom@codesourcery.com>
8251 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8252 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8255 2018-04-26 Richard Biener <rguenther@suse.de>
8256 Tom de Vries <tom@codesourcery.com>
8259 * lto-streamer-out.c (output_function): Fixup loops if required to match
8260 discovery done in the reader.
8262 2018-04-26 Richard Biener <rguenther@suse.de>
8264 PR tree-optimization/85116
8265 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8266 have a loop exit from the single latch predecessor. Remove
8267 case of header with just condition.
8268 (ch_base::copy_headers): Exclude infinite loops from any
8270 (pass_ch::execute): Record exits.
8272 2018-04-26 Richard Biener <rguenther@suse.de>
8274 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8275 prologue cost vector and pass it to vect_get_load_cost.
8276 (vect_get_peeling_costs_all_drs): Likewise.
8277 (vect_peeling_hash_get_lowest_cost): Likewise.
8278 (vect_enhance_data_refs_alignment): Likewise.
8280 2018-04-26 Richard Biener <rguenther@suse.de>
8283 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8284 checking of integer<->pointer conversions.
8285 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8286 sign-/zero-extending pointer types.
8287 (expand_omp_for_static_chunk): Likewise.
8289 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8290 Jean Lee <xiaoyur347@gmail.com>
8292 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8293 (TARGET_ASAN_SHADOW_OFFSET): Define.
8294 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8295 true for -fsanitize=address.
8297 2018-04-25 Mark Wielaard <mark@klomp.org>
8299 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8302 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8304 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8305 than "alu", remove explicit "memory" and "imm_disp" attributes.
8306 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8309 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8310 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8311 gen_lowpart_no_emit.
8313 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8316 * config/i386/i386.c (ix86_expand_builtin): Change memory
8317 operand to XI, extend p0 to Pmode.
8318 * config/i386/i386.md: Change unspec volatile and operand
8319 1 mode to XI, change operand 0 mode to P.
8321 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8323 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8324 GET_MODE_MASK before any checking.
8325 (nds32_can_use_bset_p): Likewise.
8326 (nds32_can_use_btgl_p): Likewise.
8328 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8330 * config/nds32/nds32-doubleword.md: New define_split pattern for
8331 illegal register number.
8333 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8335 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8337 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8339 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8341 2018-04-25 Richard Biener <rguenther@suse.de>
8343 * lto-streamer.h (LTO_major_version): Bump to 8.
8345 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8347 * BASE-VER: Set to 9.0.0.
8349 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8351 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8352 in __abskf2 and __powikf2.
8354 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8357 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8358 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8359 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8360 Use the above on operand 2. Reindent.
8361 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8363 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8366 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8368 * config/i386/i386.opt (mcet): Removed.
8369 * doc/install.texi: Remove -mcet documentation.
8370 * doc/invoke.texi: Likewise.
8372 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8375 * doc/install.texi: Remove -mcet from bootstrap-cet.
8377 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8380 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8381 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8385 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8386 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8387 containing a CONST_VECTOR.
8389 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8391 * doc/install.texi: Update newlib dependency for nvptx.
8393 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8396 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8397 instead of INTVAL when shifting x left.
8399 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8401 PR tree-optimization/85478
8402 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8403 vect_grouped_store_supported for single element vectors.
8405 2018-04-24 Richard Biener <rguenther@suse.de>
8408 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8409 load cost increase to the case of non-constant step.
8411 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8414 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8415 destination if any_malformed_asm.
8417 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8420 * expr.c (store_field): In the bitfield case, if the value comes from
8421 a function call and is returned in registers by means of a PARALLEL,
8422 do not change the mode of the temporary unless BLKmode and VOIDmode.
8424 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8426 PR rtl-optimization/85423
8427 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8428 dependencies to debug insns when the previous insn is non-debug.
8430 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8432 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8433 enums into a single definition.
8434 (fls): Fix predicates and printing.
8437 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8439 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8440 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8441 and short u6 immediate.
8442 (check_if_valid_sleep_operand): Remove.
8443 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8445 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8447 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8448 flag_always_save_lp condition.
8449 * config/nds32/nds32.opt (malways-save-lp): New option.
8451 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8453 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8454 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8455 * config/nds32/nds32.h
8456 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8457 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8459 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8461 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8462 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8464 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8465 Chung-Ju Wu <jasonwucj@gmail.com>
8467 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8469 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8470 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8472 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8474 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8476 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8478 * config/nds32/nds32-protos.h (nds32_data_alignment,
8479 nds32_local_alignment): Declare.
8480 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8481 nds32_local_alignment): New functions.
8482 (TARGET_CONSTANT_ALIGNMENT): Define.
8483 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8485 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8487 * config/nds32/nds32.c
8488 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8489 (TARGET_MODES_TIEABLE_P): Likewise.
8491 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8493 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8496 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8497 Chung-Ju Wu <jasonwucj@gmail.com>
8499 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8500 for unaligned access.
8501 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8503 * config/nds32/nds32-intrinsic.md: Likewise.
8504 * config/nds32/nds32_intrinsic.h: Likewise.
8505 * config/nds32/nds32.h (nds32_builtins): Likewise.
8506 * config/nds32/nds32.opt (munaligned-access): New option.
8507 * config/nds32/nds32.c (nds32_asm_file_start): Display
8508 flag_unaligned_access status.
8510 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8512 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8513 -mno-relax is present.
8514 * config/riscv/linux.h (LINK_SPEC): Ditto.
8516 2018-04-20 Martin Sebor <msebor@redhat.com>
8519 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8521 (gimple_fold_builtin_stxcpy_chk): Same.
8522 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8524 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8527 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8528 __powikf2 when long double is IEEE 128-bit.
8530 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8532 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8533 step to make sure stack always aligned.
8535 2018-04-20 Carl Love <cel@us.ibm.com>
8538 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8539 size check for arg0.
8541 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8542 Tom de Vries <tom@codesourcery.com>
8545 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8546 Emit insns for calls too.
8547 (nvptx_find_par): Always look for worker-level predecessor insn.
8548 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8550 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8551 (nvptx_process_pars): Propagate frames for calls.
8553 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8556 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8558 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8559 (ix86_handle_option): Don't handle OPT_mibt.
8560 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8562 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8564 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8565 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8566 (ix86_target_macros): Define __CET__ with flag_cf_protection
8567 for -fcf-protection.
8568 * config/i386/i386.c (isa2_opts): Remove -mibt.
8569 * config/i386/i386.h (TARGET_IBT): Removed.
8570 (TARGET_IBT_P): Likewise.
8571 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8572 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8573 * config/i386/i386.opt (mcet): Update help message.
8576 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8577 -mcet as an alias for -mshstk.
8579 2018-04-20 Richard Biener <rguenther@suse.de>
8582 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8583 complexity by forcing a single use of the multiply operand.
8585 2018-04-20 Martin Jambor <mjambor@suse.cz>
8588 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8589 recursion dependency to only apply to non-clones.
8591 2018-04-20 Martin Jambor <mjambor@suse.cz>
8594 * ipa-cp.c (create_specialized_node): Check that clones of
8595 self-recursive edges exist during IPA-CP.
8597 2018-04-19 Toon Moene <toon@moene.org>
8599 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8602 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8604 PR tree-optimization/85467
8605 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8606 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8607 VECTOR_CST element to type.
8609 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8612 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8613 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8614 (builtin_longjmp): Likewise.
8615 (save_stack_nonlocal): New pattern.
8616 (restore_stack_nonlocal): Likewise.
8618 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8621 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8622 Replace ASM_OUTPUT_LABEL with fprintf.
8624 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8627 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8628 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8629 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8630 define __IBT__ and __SHSTK__ for -fcf-protection.
8631 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8633 (ix86_trampoline_init): Likewise.
8634 (x86_output_mi_thunk): Likewise.
8635 (ix86_notrack_prefixed_insn_p): Likewise.
8636 (ix86_option_override_internal): Don't disallow -fcf-protection.
8637 * config/i386/i386.md (rdssp<mode>): Also enable for
8639 (incssp<mode>): Likewise.
8640 (nop_endbr): Likewise.
8641 * config/i386/i386.opt (mcet): Change help message to built-in
8645 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8646 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8647 enable CET built-in functions.
8649 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8651 * common/config/i386/i386-common.c
8652 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8653 OPTION_MASK_ISA_MOVDIRI_UNSET,
8654 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8655 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8656 * config.gcc (movdirintrin.h): New header.
8657 * config/i386/cpuid.h (bit_MOVDIRI,
8658 bit_MOVDIR64B): New bits.
8659 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8661 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8662 (VOID, PVOID, PCVOID)): New function types.
8663 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8664 __builtin_ia32_directstoreu_u64,
8665 __builtin_ia32_movdir64b): New builtins.
8666 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8667 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8669 (ix86_valid_target_attribute_inner_p): Ditto.
8670 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8671 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8672 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8673 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8674 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8675 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8676 (movdiri<mode>, movdir64b_<mode>): New.
8677 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8678 * config/i386/immintrin.h: Include movdirintrin.h.
8679 * config/i386/movdirintrin.h: New file.
8680 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8682 2018-04-19 Richard Biener <rguenther@suse.de>
8685 * cfg.c (clear_bb_flags): When loop state says we have
8686 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8688 2018-04-19 Richard Biener <rguenther@suse.de>
8690 PR tree-optimization/84737
8691 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8692 copying restrict info.
8693 (vect_setup_realignment): Use it.
8694 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8695 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8696 the first DR to all generated stores.
8697 (vectorizable_load): Likewise for loads.
8699 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8701 PR tree-optimization/85446
8702 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8703 the integral and pointer types to have the same precision.
8705 * doc/install.texi: Document --disable-cet being the default and
8708 2018-04-18 Martin Liska <mliska@suse.cz>
8710 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8713 2018-04-18 Martin Liska <mliska@suse.cz>
8716 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8719 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8720 arguments if they are comparable.
8722 2018-04-18 Martin Liska <mliska@suse.cz>
8725 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8728 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8731 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8734 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8735 ENDBR after calling __morestack.
8737 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8740 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8741 by using gcc_base_ver to generate a gcc_driver_version, and use
8742 it when generating GCC_DRIVER_NAME.
8743 * configure: Regenerate.
8745 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8748 * config.gcc: Obsolete powerpc*-*-*spe*.
8750 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8753 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8754 (stabstr_D): Change type of unum from unsigned int to
8755 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8758 2018-04-17 Jim Wilson <jimw@sifive.com>
8761 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8762 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8763 Set arg_pointer_offset after using pretend_args_size.
8765 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8767 PR rtl-optimization/85431
8768 * dse.c (record_store): Ignore zero width stores.
8771 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8772 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8773 __builtin_stack_restore rather than after it.
8774 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8775 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8776 argument instead of virtual_dynamic_stack_rtx.
8778 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8780 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8782 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8783 Add note to error message to explain internal mapping of overloaded
8784 built-in function name to non-overloaded built-in function name.
8785 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8788 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8791 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8792 where the inputs overlap with the output.
8794 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8796 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8797 (=v, v) alternative and explicit "memory" attribute.
8798 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8799 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8801 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8802 "sselog1" type instead of "sselog".
8803 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8804 "sselog". Remove explicit "memory" attribute.
8805 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8806 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8808 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8809 "sselog1" type instead of "sselog". Remove explicit "memory"
8811 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8812 use "sselog1" type instead of "sselog". Remove explicit "memory"
8814 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8815 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8817 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8818 "sselog1" type instead of "sselog". Remove explicit "memory"
8820 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8821 use "sselog1" type instead of "sselog". Remove explicit "memory"
8825 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8828 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8831 2018-04-17 Martin Jambor <mjambor@suse.cz>
8834 * ipa-cp.c (create_specialized_node): Call
8835 expand_all_artificial_thunks if necessary.
8837 2018-04-17 Martin Liska <mliska@suse.cz>
8840 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8841 in message, remote space in between '_G' and '('.
8843 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8846 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8847 avx512f_vmcmp<mode>3<round_saeonly_name>,
8848 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8849 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8850 avx512f_rndscale<mode><round_saeonly_name>,
8851 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8852 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8853 Use %<iptr>2 instead of %2 for -masm=intel.
8854 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8855 avx512f_vcvttss2usi<round_saeonly_name>,
8856 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8858 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8859 avx512f_vcvttsd2usi<round_saeonly_name>,
8860 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8861 Use %q1 instead of %1 for -masm=intel.
8862 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8863 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8864 of %3 for -masm=intel.
8865 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8867 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8869 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8871 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8872 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8874 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8876 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8877 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8878 %g1 and one with %0 and %1.
8879 (avx512er_vmrcp28<mode><round_saeonly_name>,
8880 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8882 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8883 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8884 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8885 of %0 and %{%4%} for -masm=intel.
8886 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8887 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8888 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8889 order of %0 and %{%5%}%{z%} for -masm=intel.
8891 2018-04-17 Jan Hubicka <jh@suse.cz>
8894 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8896 2018-04-17 Martin Liska <mliska@suse.cz>
8899 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8900 for target_clone error message. Make default implementation
8901 clone to be a local declaration.
8902 (separate_attrs): Add new argument and check for an empty
8904 (expand_target_clones): Handle it.
8905 (ipa_target_clone): Make redirection just for target_clones
8908 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8909 Tom de Vries <tom@codesourcery.com>
8912 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8913 tiled basic blocks without omp continue statements.
8915 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8918 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8919 vec_extract expression as having side effects to make sure it gets
8922 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8925 * config/i386/i386.c (get_builtin_code_for_version): Check
8928 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8931 * gcc/config.gcc: Support "skylake".
8932 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8934 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8935 (processor_target_table): Add "skylake".
8936 (ix86_option_override_internal): Add "skylake".
8937 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8938 PROCESSOR_CANNONLAKE.
8939 (get_builtin_code_for_version): Fix priority for
8940 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8941 PROCESSOR_SKYLAKE-AVX512.
8942 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8943 (processor_type): Add PROCESSOR_SKYLAKE.
8945 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8946 Jason Merrill <jason@redhat.com>
8949 * convert.c (convert_to_integer_1): Use direct recursion for
8950 enumeral types and types with a precision less than the number
8951 of bits in their mode.
8953 2018-04-16 Julia Koval <julia.koval@intel.com>
8956 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8957 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8959 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8962 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8963 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8964 and -mno-direct-move.
8966 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8969 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8970 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8971 Ensure negative shifts result in {0}.
8973 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8975 PR rtl-optimization/79916
8976 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8977 regs (if any) to define how to gnerate SD moves when LRA is in
8980 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8982 PR rtl-optimization/85393
8983 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8984 * except.c (expand_dw2_landing_pad_for_region): Make static.
8985 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8986 a label and unconditional jump to old_bb, rather than
8987 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8990 PR rtl-optimization/85376
8991 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8992 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8993 instead of a specific value.
8995 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8996 Bin Cheng <bin.cheng@arm.com>
8998 PR tree-optimization/82965
8999 PR tree-optimization/83991
9000 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9001 by_profile_only parameter.
9002 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9003 information if the loop was predicted to iterate too many times.
9004 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9006 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9009 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9012 2018-04-13 Martin Liska <mliska@suse.cz>
9013 Jakub Jelinek <jakub@redhat.com>
9016 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9017 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9018 * builtins.c (expand_builtin_memory_copy_args): Use
9019 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9020 handle dest_addr == pc_rtx.
9022 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9025 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9026 asked to not generate direct moves.
9027 (fix_trunc<mode>si2_stfiwx): Similar.
9028 (fix_trunc<mode>si2_internal): Similar.
9030 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9033 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9034 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9035 lookup if dest in some wider mode is known to be const0_rtx and
9036 if so, record permanent equivalence for it to be ZERO_EXTEND of
9037 the narrower mode destination.
9039 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9041 * lto-streamer-out.c (output_function): Revert 259346.
9042 * omp-expand.c (expand_oacc_for): Likewise.
9044 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9046 PR rtl-optimization/85354
9047 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9048 * sel-sched.c (sel_global_init): ... here.
9050 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9053 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9054 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9055 mode for PE-COFF targets.
9056 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9057 (i386_pe_asm_lto_end): Likewise.
9058 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9059 (TARGET_ASM_LTO_END): Likewise.
9060 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9061 (i386_pe_asm_lto_start): New function.
9062 (i386_pe_asm_lto_end): Likewise.
9064 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9065 Richard Biener <rguenther@suse.de>
9068 * lto-streamer-out.c (output_function): Fix CFG loop state before
9070 * omp-expand.c (expand_oacc_for): Handle calls to internal
9071 functions like regular functions.
9073 2018-04-12 Richard Biener <rguenther@suse.de>
9076 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9077 for the early LTO debug to properly generate references to it
9078 during DIE emission. Do not re-use that for the skeleton for
9080 (dwarf2out_early_finish): Likewise.
9082 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9085 * config/i386/sse.md
9086 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9087 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9088 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9089 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9090 and output is a reg, avoid creating invalid lowpart subreg, but
9091 instead split into a 512-bit move. Don't split if not AVX512VL,
9092 input is xmm16+ reg and output is a mem.
9093 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9094 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9095 xmm16+ reg and output is a mem.
9097 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9099 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9100 also for flag_dwarf2_cfi_asm.
9102 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9104 PR rtl-optimization/85342
9105 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9106 a bool scalar var inside of the loop instead. Don't try to update
9107 recog_data.operand after failed apply_change_group.
9109 2018-04-12 Tom de Vries <tom@codesourcery.com>
9112 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9113 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9114 array with flexible array member as array without given dimension.
9115 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9116 argument for undefined param to true.
9118 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9121 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9122 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9123 from PowerPC section.
9124 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9125 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9126 help text that is too long.
9127 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9128 help text that is too long.
9129 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9130 help text that is too long.
9132 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9134 * config/alpha/alpha.md (stack_probe_internal): Rename
9135 from "probe_stack". Update all callers.
9137 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9139 PR rtl-optimization/84566
9140 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9141 sched_macro_fuse_insns.
9143 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9146 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9147 (compute_block_dependences): ... from here.
9149 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9151 PR tree-optimization/85331
9152 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9153 from int to HOST_WIDE_INT.
9155 2018-04-11 Martin Jambor <mjambor@suse.cz>
9158 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9159 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9160 not the same as the source val.
9161 (cgraph_edge_brings_value_p): New parameter.
9162 (gather_edges_for_value): Pass destination value to
9163 cgraph_edge_brings_value_p.
9164 (perhaps_add_new_callers): Likewise.
9165 (get_info_about_necessary_edges): Likewise and exclude values brought
9166 only by self-recursive edges.
9167 (create_specialized_node): Redirect only clones of self-calling edges.
9168 (+self_recursive_pass_through_p): New function.
9169 (find_more_scalar_values_for_callers_subset): Use it.
9170 (find_aggregate_values_for_callers_subset): Likewise.
9171 (known_aggs_to_agg_replacement_list): Removed.
9172 (decide_whether_version_node): Re-calculate known constants for all
9173 remaining context clones.
9175 2018-04-11 Richard Biener <rguenther@suse.de>
9178 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9179 from early DWARF output.
9180 (dwarf2out_early_finish): Output line info unconditionally into
9181 early DWARF and add reference to it.
9183 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9186 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9187 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9188 other than V2DFmode using iptr mode attribute.
9189 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9191 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9193 PR rtl-optimization/84659
9194 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9196 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9199 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9201 (output_loc_list): Pass address of a dummy size variable even in the
9202 locview handling loop.
9203 (index_location_lists): Add comment on why skip_loc_list_entry can't
9206 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9209 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9212 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9215 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9216 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9217 and -mstring-compare-inline-limit.
9219 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9222 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9223 for stack clash protection in a register whenever we need it to be in
9226 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9228 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9229 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9231 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9234 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9236 (mlong-double-): Ditto.
9237 * config/rs6000/sysv4.opt (msdata=): Ditto.
9238 (mtls-size=): Ditto.
9240 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9242 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9243 erroneous entries for
9244 "vector int vec_ldl (int, long int *)", and
9245 "vector unsigned int vec_ldl (int, unsigned long int *)".
9246 Add comments and entries for
9247 "vector bool char vec_ldl (int, bool char *)",
9248 "vector bool short vec_ldl (int, bool short *)",
9249 "vector bool int vec_ldl (int, bool int *)",
9250 "vector bool long long vec_ldl (int, bool long long *)",
9251 "vector pixel vec_ldl (int, pixel *)",
9252 "vector long long vec_ldl (int, long long *)",
9253 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9254 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9255 type tree bool_long_long_type_node and correct definition of
9256 bool_V2DI_type_node to make reference to this new type tree.
9257 (rs6000_mangle_type): Replace erroneous reference to
9258 bool_long_type_node with bool_long_long_type_node.
9259 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9260 comments to emphasize sign distinctions for char and int types and
9261 replace RS6000_BTI_bool_long constant with
9262 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9263 use of RS6000_BTI_pixel.
9264 (bool_long_type_node): Remove this macro definition.
9265 (bool_long_long_type_node): New macro definition
9267 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9269 PR rtl-optimization/85300
9270 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9271 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9272 simplify_unary_operation fails.
9274 2018-04-10 Martin Liska <mliska@suse.cz>
9276 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9277 cgraph_edge and ipa_ref.
9279 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9283 * config/i386/sse.md
9284 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9285 computation of the VEC_MERGE selector from mask.
9286 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9287 Fix decoding of the VEC_MERGE selector into mask.
9289 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9291 PR tree-optimization/85286
9292 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9294 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9296 * final.c (final_1): Set insn_last_address as well as
9297 insn_current_address.
9299 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9302 * explow.c (emit_stack_probe): Call validize_mem on memory location
9303 before passing it to gen_probe_stack. Create address operand and
9304 legitimize it for the probe_stack_address case.
9306 2018-04-09 Jan Hubicka <jh@suse.cz>
9309 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9310 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9311 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9313 (free_lang_data): Rebuild type inheritance graph.
9315 2018-04-09 Martin Sebor <msebor@redhat.com>
9317 * invoke.texi (-finline-small-functions): Mention other optimization
9319 (-findirect-inlining, -fpartial-inlining): Same.
9320 (-finline-functions-called-once): Same.
9321 (-freorder-blocks-and-partition): Same.
9323 2018-04-09 Jan Hubicka <jh@suse.cz>
9326 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9327 jumps; choose last target that matches the criteria (i.e.
9328 no partition changes for non-crossing jumps).
9329 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9330 support for redirecting crossing jumps to non-crossing.
9332 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9334 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9335 also for naked functions.
9337 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9339 * config/arc/arc.md (add_shift): New pattern.
9340 (add_shift2): Likewise.
9341 (sub_shift): Likewise.
9342 (sub_shift_cmp0_noout): Likewise.
9343 (compare_si_ashiftsi): Likewise.
9344 (xbfu_cmp0_noout): New combine pattern.
9345 (xbfu_cmp0"): Likewise.
9346 (movsi_set_cc_insn): Place the predicable variant first.
9347 (commutative_binary_cmp0_noout): Remove clobber.
9348 (commutative_binary_cmp0): New pattern.
9349 (noncommutative_binary_cmp0): Likewise.
9350 (noncommutative_binary_cmp0_noout): Likewise.
9351 (noncommutative_binary_comparison_result_used): Removed.
9352 (rsub_cmp0): New pattern.
9353 (rsub_cmp0_noout): Likewise.
9354 (extzvsi): Changed, keep only meaningful variants.
9355 (SQH, SEZ): New iterators.
9356 (SQH_postfix): New mode attribute.
9357 (SEZ_prefix): New code attribute.
9358 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9359 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9360 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9362 (noncommutative_operator): Check the availability of barrel
9365 2018-04-09 Richard Biener <rguenther@suse.de>
9367 PR tree-optimization/85284
9368 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9369 Only use the niter constraining form of simple_iv when the exit
9372 2018-04-09 Tom de Vries <tom@codesourcery.com>
9375 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9376 (define_expand "*memory_barrier"): New define_expand.
9377 (define_insn "memory_barrier"): New insn.
9379 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9381 PR rtl-optimization/80463
9382 PR rtl-optimization/83972
9383 PR rtl-optimization/83480
9385 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9386 correct producer for the insn.
9387 (tidy_control_flow): Fixup seqnos in case of debug insns.
9389 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9391 PR rtl-optimization/83913
9393 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9394 different sched-times when merging exprs.
9396 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9398 PR rtl-optimization/83962
9400 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9401 tidy_fallthru_edge and tidy_control_flow.
9403 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9405 PR rtl-optimization/83530
9407 * sel-sched.c (force_next_insn): New global variable.
9408 (remove_insn_for_debug): When force_next_insn is true, also leave only
9409 next insn in the ready list.
9410 (sel_sched_region): When the region wasn't scheduled, make another pass
9411 over it with force_next_insn set to 1.
9413 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9415 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9417 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9418 for interrupt control.
9419 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9420 functions for interrupt control.
9421 * config/nds32/nds32-intrinsic.md: Likewise.
9422 * config/nds32/nds32_intrinsic.h: Likewise.
9423 * config/nds32/nds32.h (nds32_builtins): Likewise.
9425 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9427 * config/nds32/nds32.c (nds32_init_machine_status,
9428 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9429 strict_aligned_p field.
9430 (nds32_expand_to_rtl_hook): New function.
9431 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9432 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9434 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9435 Chung-Ju Wu <jasonwucj@gmail.com>
9437 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9438 * config/nds32/nds32-n7.md: New file.
9439 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9440 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9442 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9443 * config/nds32/nds32.md (pipeline_model): Add n7.
9444 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9445 * config/nds32/pipelines.md: Include n7 settings.
9447 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9448 Chung-Ju Wu <jasonwucj@gmail.com>
9450 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9451 * config/nds32/nds32-e8.md: New file.
9452 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9453 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9455 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9456 * config/nds32/nds32.md (pipeline_model): Add e8.
9457 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9458 * config/nds32/pipelines.md: Include e8 settings.
9460 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9461 Chung-Ju Wu <jasonwucj@gmail.com>
9463 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9464 * config/nds32/nds32-n8.md: New file.
9465 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9466 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9468 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9469 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9470 * config/nds32/nds32.md (pipeline_model): Add n8.
9471 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9472 * config/nds32/pipelines.md: Include n8 settings.
9474 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9475 Chung-Ju Wu <jasonwucj@gmail.com>
9477 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9478 * config/nds32/nds32-n9-2r1w.md: New file.
9479 * config/nds32/nds32-n9-3r2w.md: New file.
9480 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9481 nds32_register_ports): New or modify for cpu n9.
9482 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9484 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9485 * config/nds32/nds32-utils.c: New file.
9486 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9487 TARGET_MUL_SLOW): Define.
9488 * config/nds32/nds32.md (pipeline_model): New attribute.
9489 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9490 New options that support cpu n9.
9491 * config/nds32/pipelines.md: Include n9 settings.
9492 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9494 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9496 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9497 information if necessary.
9498 (output_cond_branch_compare_zero): Likewise.
9499 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9500 (nds32_target_alignment): Refine for alignment.
9501 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9502 (FUNCTION_BOUNDARY): Modify.
9503 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9505 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9507 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9509 * config/nds32/constants.md (unspec_volatile_element): Add values for
9510 TLB operation and data prefetch.
9511 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9512 functions for TLB operation and data prefetch.
9513 * config/nds32/nds32-intrinsic.md: Likewise.
9514 * config/nds32/nds32_intrinsic.h: Likewise.
9515 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9516 (nds32_print_operand): Likewise.
9517 * config/nds32/nds32.h (nds32_builtins): Likewise.
9519 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9520 Andrew Pinski <pinsika@gcc.gnu.org>
9523 * match.pd: Use constant_boolean_node of correct type instead of
9524 boolean_true_node or boolean_false_node for simplifying
9525 pointer comparisons to zero.
9527 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9529 PR tree-optimization/80021
9530 * tree.c (verify_type_variant): Make error call in verify_variant_match
9531 translatable and remove final full stop.
9533 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9535 * config/nds32/constants.md (unspec_volatile_element): Add
9536 UNSPEC_VOLATILE_EH_RETURN.
9537 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9538 nds32_output_stack_pop): Support dwarf exception handling process.
9539 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9540 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9541 exception handling process.
9542 (nds32_compute_stack_frame): Likewise.
9543 (nds32_return_addr_rtx): Likewise.
9544 (nds32_initial_elimination_offset): Likewise.
9545 (nds32_expand_prologue): Likewise.
9546 (nds32_expand_epilogue): Likewise.
9547 (nds32_dynamic_chain_address): New function.
9548 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9550 (DYNAMIC_CHAIN_ADDRESS): Define.
9551 (EH_RETURN_DATA_REGNO): Define.
9552 (EH_RETURN_STACKADJ_RTX): Define.
9553 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9554 patterns for dwarf exception handling.
9556 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9558 * config/nds32/nds32.h: Clean up obsolete macros.
9560 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9562 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9563 Add enum values for particular instructions.
9564 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9565 particular intrinsic functions.
9566 * config/nds32/nds32-intrinsic.md: Likewise.
9567 * config/nds32/nds32_intrinsic.h: Likewise.
9568 * config/nds32/nds32.h (nds32_builtins): Likewise.
9569 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9570 (btst, ave): New patterns for particular instructions.
9572 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9574 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9575 Add enum values for atomic load/store and memory sync.
9576 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9578 * config/nds32/nds32-intrinsic.md: Likewise.
9579 * config/nds32/nds32_intrinsic.h: Likewise.
9580 * config/nds32/nds32.h (nds32_builtins): Likewise.
9582 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9584 PR tree-optimization/85257
9585 * fold-const.c (native_encode_vector): If not all elts could fit
9586 and off is -1, return 0 rather than offset.
9587 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9588 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9589 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9590 adjust buffer in native_interpret_expr call.
9592 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9594 * config/nds32/constants.md (unspec_volatile_element): Add cache
9595 control enum values.
9596 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9597 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9598 * config/nds32/nds32.c (nds32_cctl_names): New.
9599 (nds32_print_operand): Handle cache control register names.
9600 * config/nds32/nds32.h (nds32_builtins): New enum values.
9601 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9603 * config/nds32/nds32.md (type): Add mmu.
9604 * config/nds32/pipelines.md (simple_insn): Add mmu.
9606 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9608 * config/nds32/nds32.md (type): Remove call.
9609 * config/nds32/pipelines.md (simple_insn): Likewise.
9611 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9613 * config/nds32/constants.md (unspec_volatile_element): Add
9614 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9615 UNSPEC_VOLATILE_FMFCFG.
9616 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9617 description for fmfcfg and fmfcsr.
9618 (bdesc_1arg): Add fmtcsr.
9619 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9620 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9621 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9622 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9623 unspec_fmfcfg): New patterns.
9624 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9625 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9626 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9627 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9628 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9629 __nds32__fmfcfg): Define.
9631 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9633 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9634 intrinsic register names.
9635 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9636 intrinsic register enum values and macros.
9638 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9640 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9641 for load/store addressing form.
9642 (nds32_print_operand_address): Likewise.
9644 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9647 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9648 based on LABEL_REF. Remove useless assertion.
9649 (pic_address_needs_scratch): Fix formatting.
9650 (sparc_legitimize_pic_address): Minor tweaks.
9651 (sparc_delegitimize_address): Adjust assertion accordingly.
9652 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9653 into symbolic_operand.
9654 (movsi_high_pic_label_ref): Likewise.
9655 (movsi_lo_sum_pic_label_ref): Likewise.
9656 (movdi_pic_label_ref): Likewise.
9657 (movdi_high_pic_label_ref): Likewise.
9658 (movdi_lo_sum_pic_label_ref): Likewise.
9660 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9662 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9663 custom LIB_SPEC setup.
9665 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9666 Kito Cheng <kito.cheng@gmail.com>
9668 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9669 * config/riscv/freebsd.h: New.
9671 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9673 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9674 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9677 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9678 Kito Cheng <kito.cheng@gmail.com>
9680 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9681 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9682 * config/nds32/nds32-protos.h (nds32_output_call,
9683 nds32_output_return): Declare.
9684 * config/nds32/nds32.md: Refine all the call and return patterns.
9686 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9689 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9690 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9692 PR rtl-optimization/84872
9693 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9694 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9697 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9699 * expr.c (copy_blkmode_to_reg): Revert 254862.
9700 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9702 2018-04-06 Richard Biener <rguenther@suse.de>
9705 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9706 after seeing a component reference with an adjacent field. Treat
9707 refs to arrays at struct end of external decls similar to
9708 refs to unconstrained commons.
9710 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9713 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9714 look through SAVE_EXPRs with non-side-effects argument. Adjust
9716 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9719 2018-04-06 Richard Biener <rguenther@suse.de>
9722 * alias.c (find_base_term): New wrapper around find_base_term
9723 unwinding CSELIB_VAL_PTR changes.
9724 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9727 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9729 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9731 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9732 constant definitions.
9733 ("nop"): lr 0,0 -> nopr r0
9734 ("nop_lr0", "nop_lr1"): New insn definitions.
9736 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9738 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9739 NDS32_V3PUSH_AVAILABLE_P macro.
9741 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9742 Chung-Ju Wu <jasonwucj@gmail.com>
9744 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9745 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9746 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9747 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9748 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9749 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9750 * config/nds32/constraints.md: New constraints and checking for hard
9751 float configuration.
9752 * config/nds32/iterators.md: New mode iterator and attribute for hard
9753 float configuration.
9754 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9756 * config/nds32/nds32-fpu.md: New file.
9757 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9758 deal with hard float code generation.
9759 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9761 (abi_type, float_reg_number): New enum type.
9762 * config/nds32/nds32-predicates.c: New predicates for hard float.
9763 * config/nds32/nds32-protos.h: Declare functions for hard float.
9764 * config/nds32/nds32.c: Implementation for hard float configuration.
9765 * config/nds32/nds32.h: Definitions for hard float configuration.
9766 * config/nds32/nds32.md: Include hard float machine description and
9767 modify patterns for hard float configuration.
9768 * config/nds32/nds32.opt: New options for hard float configuration.
9769 * config/nds32/predicates.md: New predicates for hard float
9772 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9774 * common/config/nds32/nds32-common.c
9775 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9777 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9780 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9781 CONSTRUCTOR_ELT (ctor, ...)->value.
9783 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9786 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9788 2018-04-05 Tom de Vries <tom@codesourcery.com>
9791 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9794 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9795 Kito Cheng <kito.cheng@gmail.com>
9797 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9798 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9799 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9800 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9802 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9803 Kito Cheng <kito.cheng@gmail.com>
9805 * config/nds32/constraints.md (Ufe): New memory constraint.
9806 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9807 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9808 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9810 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9811 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9813 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9815 * config/nds32/nds32.md: Use optimize_size in the condition for
9816 alu-shift instructions.
9818 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9820 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9822 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9824 * config/nds32/nds32.md (negsi2): Refine pattern.
9826 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9827 Chung-Ju Wu <jasonwucj@gmail.com>
9829 * config/nds32/iterators.md (shift_rotate): New code iterator.
9830 (shift): New code attribute.
9831 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9832 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9833 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9834 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9835 bit-wise operations.
9836 (andsi3, *andsi3): Ditto.
9837 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9838 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9839 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9840 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9841 nds32_ior_operand, nds32_xor_operand): New predicates.
9843 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9845 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9846 (addsi3, subsi3): ... this.
9848 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9850 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9852 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9854 * config/nds32/nds32.md: Adjust indention.
9856 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9858 * config/nds32/nds32.md (feature): New attribute.
9860 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9862 * config/nds32/nds32.md (subtype): New attribute.
9864 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9867 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9868 expansion to perform a bitwise AND of the argument followed by a
9869 boolean negation of the result.
9871 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9873 PR rtl-optimization/84878
9874 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9875 the basic block. Assert the use reference is not artificial and that
9876 it has an associated insn.
9878 2018-04-04 Michael Matz <matz@suse.de>
9880 * builtins.c (compute_objsize): Pass correct operand
9881 to array_at_struct_end_p.
9883 2018-04-04 Richard Biener <rguenther@suse.de>
9886 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9887 from contexts for DINFO_LEVEL_TERSE and below.
9889 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9891 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9892 resiter_operand condition.
9893 * config/nds32/nds32.md (*move<mode>): Ditto.
9895 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9896 Monk Chiang <sh.chiang04@gmail.com>
9898 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9900 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9902 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9904 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9905 Kito Cheng <kito.cheng@gmail.com>
9907 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9908 nds32_cond_code_str, output_cond_branch,
9909 output_cond_branch_compare_zero, nds32_expand_cbranch,
9910 nds32_expand_cstore, nds32_expand_movcc,
9911 nds32_output_cbranchsi4_equality_zero,
9912 nds32_output_cbranchsi4_equality_reg,
9913 nds32_output_cbranchsi4_equality_reg_or_const_int,
9914 nds32_output_cbranchsi4_greater_less_zero: New functions.
9915 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9916 nds32_expand_cstore, nds32_expand_movcc,
9917 nds32_output_cbranchsi4_equality_zero,
9918 nds32_output_cbranchsi4_equality_reg,
9919 nds32_output_cbranchsi4_equality_reg_or_const_int,
9920 nds32_output_cbranchsi4_greater_less_zero): Declare.
9921 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9922 nds32_rimm11s_operand): New predicates.
9923 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9924 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9927 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9929 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9930 * config/nds32/nds32.md: Ditto.
9931 * config/nds32/pipelines.md: Ditto.
9933 2018-04-04 Richard Biener <rguenther@suse.de>
9935 PR tree-optimization/85168
9936 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9937 propagating abnormals.
9939 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9941 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9943 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9944 Kito Cheng <kito.cheng@gmail.com>
9946 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9947 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9948 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9949 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9950 * config/nds32/nds32.md (sibcall_internal): New.
9951 (sibcall_register): Remove.
9952 (sibcall_immediate): Remove.
9953 (sibcall_value_internal): New.
9954 (sibcall_value_register): Remove.
9955 (sibcall_value_immediate): Remove.
9956 * config/nds32/predicates.md (nds32_general_register_operand): New.
9957 (nds32_call_address_operand): New.
9959 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9961 PR rtl-optimization/85167
9962 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9963 bb_defs if *split_p, instead preinitialize it to NULL.
9965 PR tree-optimization/85156
9966 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9967 evaluating the argument multiple times.
9969 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9971 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9973 (_mm_cvtpd_ps): Likewise.
9974 (_mm_cvttpd_epi32): Likewise.
9975 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9976 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9977 vector, pixel, and bool following altivec.h include.
9979 2018-04-03 Martin Sebor <msebor@redhat.com>
9981 * doc/extend.texi (Common Function Attributes): Clarify.
9982 (const attribute): Likewise.
9983 (pure attribute): Likewise.
9985 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9988 * config/i386/i386.c (ix86_expand_vector_set): Use
9989 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9991 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9993 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9994 instructions when changing rounding bits to preserve precision bits
9995 in the x87 control word.
9997 2018-04-03 Martin Liska <mliska@suse.cz>
9999 PR tree-optimization/82491
10000 * rtl.h (strip_offset_and_add): Replace += suboffset with
10001 poly_uint64 () + suboffset.
10003 2018-03-29 Martin Liska <mliska@suse.cz>
10004 Martin Jambor <mjambor@suse.cz>
10007 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10008 param_type is not an integral or pointer type.
10010 2018-04-03 Richard Biener <rguenther@suse.de>
10012 * sese.h (recompute_all_dominators): Remove.
10014 2018-04-02 Martin Sebor <msebor@redhat.com>
10016 * doc/invoke.texi (-Wrestrict): Fix typos.
10018 2018-04-02 Jim Wilson <jimw@sifive.com>
10020 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10021 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10022 (<optab>di3, <optab>si3_extend): Likewise.
10023 (<optab>si3_mask, <optab>si3_mask_1): New.
10024 (<optab>di3_mask, <optab>di3_mask_1): New.
10025 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10026 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10027 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10029 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10031 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10034 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10036 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10037 (nds32_canonicalize_comparison): New function.
10039 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10040 Kito Cheng <kito.cheng@gmail.com>
10041 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10043 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10044 * config/nds32/constants.md (unspec_volatile_element): Add
10045 UNSPEC_VOLATILE_RELAX_GROUP.
10046 * config/nds32/nds32-relax-opt.c: New file.
10047 * config/nds32/nds32-predicates.c
10048 (nds32_symbol_load_store_p): New function.
10049 * config/nds32/nds32-protos.h
10050 (nds32_symbol_load_store_p): Declare function.
10051 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10052 * config/nds32/nds32.c
10053 (nds32_register_pass): New function to register pass.
10054 (nds32_register_passes): New function to register passes.
10055 * config/nds32/nds32.md (relax_group): New pattern.
10056 * config/nds32/nds32.opt (mrelax-hint): New option.
10057 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10059 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10061 * config/nds32/t-nds32: Modify files dependency.
10063 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10065 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10066 (PROFILE_HOOK): Define its implementation.
10068 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10070 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10071 type and 32-bit size.
10073 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10075 PR middle-end/85090
10076 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10077 (V_128_256): New mode iterator.
10078 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10079 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10080 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10082 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10083 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10085 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10088 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10089 NaN inputs correctly.
10091 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10094 * config/rs6000/vsx.md (??r): New mode attribute.
10095 (*vsx_mov<mode>_64bit): Use it.
10096 (*vsx_mov<mode>_32bit): Likewise.
10098 2018-03-30 Martin Sebor <msebor@redhat.com>
10100 PR tree-optimization/84818
10101 * builtins.c (check_access): Use warning_n.
10103 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10106 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10108 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10111 2018-03-30 Julia Koval <julia.koval@intel.com>
10114 * x86-tune.def (movx, partial_reg_dependency): Enable for
10117 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10119 PR inline-asm/84985
10120 * lra-constraints.c (process_alt_operands): Move setting
10121 this_alternative_matches below.
10123 2018-03-29 Martin Liska <mliska@suse.cz>
10126 * doc/invoke.texi: Document how LTO works with debug info.
10127 Describe auto-load support of binutils. Mention 'x86-64'
10128 as valid option value of -march option.
10130 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10132 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10135 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10136 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10137 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10140 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10143 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10144 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10145 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10146 for RS6000_BTM_POWERPC64.
10147 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10148 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10149 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10152 (DIVDEU): Likewise.
10154 2018-03-28 Carl Love <cel@us.ibm.com>
10157 2017-09-27 Carl Love <cel@us.ibm.com>
10159 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10160 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10161 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10164 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10166 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10167 instead of __vector bool.
10168 (_mm_max_pu8): Likewise.
10169 (_mm_min_pi16): Likewise.
10171 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10174 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10175 (DIVWEUO): Likewise.
10176 (DIVDEO): Likewise.
10177 (DIVDEUO): Likewise.
10178 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10179 DIVWEUO and DIVDEUO.
10180 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10181 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10182 (div_extend): Likewise.
10183 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10185 (__builtin_divweuo): Likewise.
10186 (__builtin_divdeo): Likewise.
10187 (__builtin_divdeuo): Likewise.
10189 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10192 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10193 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10195 PR tree-optimization/82004
10196 * gimple-match-head.c (optimize_pow_to_exp): New function.
10197 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10198 Don't fold to exp if optimize_pow_to_exp is false.
10200 2018-03-28 Martin Liska <mliska@suse.cz>
10203 * calls.c (initialize_argument_information): Fix trailing space.
10204 * common.opt: Fix typo and provide better explanation for
10205 -fsanitize-coverage option.
10206 * config/i386/i386.opt: Fix typo.
10208 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10209 Martin Liska <mliska@suse.cz>
10212 * gimplify.c (asan_poison_variable): Don't do the check for
10213 gimplify_omp_ctxp here.
10214 (gimplify_decl_expr): Do it here.
10215 (gimplify_target_expr): Likewise.
10217 2018-03-28 Martin Liska <mliska@suse.cz>
10220 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10221 chkp_type_bounds_count if MPX is not enabled.
10223 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10225 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10227 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10230 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10231 function to create the function decl for complex long double
10232 multiply and divide for -mabi=ieeelongdouble.
10233 (init_float128_ieee): Call it.
10235 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10238 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10239 -fcf-protection=branch -mibt.
10240 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10242 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10245 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10246 arm_word_relocations.
10248 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10251 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10252 extern array declarations.
10254 2018-03-27 Richard Biener <rguenther@suse.de>
10256 PR middle-end/84067
10257 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10258 explicit single_use checks.
10260 2018-03-27 Richard Biener <rguenther@suse.de>
10262 PR tree-optimization/85082
10263 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10266 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10268 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10269 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10270 Turn on fasynchronous-unwind-tables and funwind-tables.
10272 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10275 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10276 (*bmi_blsr_<mode>_ccz): Ditto.
10278 2018-03-26 Tom de Vries <tom@codesourcery.com>
10280 PR tree-optimization/85063
10281 * omp-general.c (offloading_function_p): New function. Factor out
10283 * omp-offload.c (pass_omp_target_link::gate): ... here.
10284 * omp-general.h (offloading_function_p): Declare.
10285 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10286 with attribute omp declare target for offloading functions.
10288 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10290 PR tree-optimization/84005
10291 * tree-data-ref.h (get_base_for_alignment): Declare.
10292 * tree-data-ref.c (get_base_for_alignment_1): New function.
10293 (get_base_for_alignment): Likewise.
10294 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10295 get_base_for_alignment to find a suitable base object, instead
10296 of always using drb->base_address.
10298 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10300 PR inline-asm/85022
10301 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10302 known size by default.
10304 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10306 PR inline-asm/85030
10307 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10308 and non BLKmode operands.
10310 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10313 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10314 Clean up attributes.
10316 2018-03-23 Richard Biener <rguenther@suse.de>
10319 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10320 we are going to emit early debug for LTO.
10322 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10324 PR inline-asm/85034
10325 * function.c (match_asm_constraints_1): Don't optimize if input
10326 doesn't satisfy general_operand predicate for output's mode.
10328 PR inline-asm/85022
10329 * alias.c (write_dependence_p): Don't require for x_canonicalized
10330 non-VOIDmode if x has VOIDmode.
10333 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10334 just don't try to optimize it rather than assert it never happens.
10336 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10338 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10339 macro expansions for definition of ST_INTERNAL_<mode> and
10340 LD_INTERNAL_<mode> builtins.
10341 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10343 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10345 (altivec_expand_st_builtin): Likewise.
10346 (altivec_expand_builtin): Remove calls to deleted functions.
10347 (rs6000_address_for_altivec): Delete this function.
10348 * config/rs6000/vector.md: Remove expands for
10349 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10351 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10354 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10355 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10356 re-computing once computed.
10357 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10358 (arm_init_machine_status): Initialize
10359 machine->static_chain_stack_bytes.
10361 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10364 * doc/extend.texi: Add four new prototypes for vec_ld.
10365 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10366 definitions for more logical presentation.
10367 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10368 entries for V1TI variants of __builtin_altivec_ld builtin.
10369 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10370 handling of V1TI variant of LVX icode pattern.
10371 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10372 (rs6000_gimple_fold_builtin): Likewise.
10373 (altivec_init_builtins): Add code to define
10374 __builtin_altivec_lvx_v1ti function.
10376 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10378 PR inline-asm/84941
10379 * function.c (match_asm_constraints_1): Don't do the optimization
10380 if input isn't a REG, SUBREG, MEM or constant.
10382 2018-03-22 Tom de Vries <tom@codesourcery.com>
10384 PR tree-optimization/84956
10385 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10386 bb_has_abnormal_pred.
10388 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10391 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10392 DECL_INITIAL (decl) to decl at the end.
10393 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10394 adjust the comment.
10396 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10398 * doc/extend.texi (__builtin_tgmath): Document when complex
10399 integer types are treated as _Complex _Float64.
10401 2018-03-21 Tom de Vries <tom@codesourcery.com>
10403 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10405 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10407 PR tree-optimization/84960
10408 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10409 if it is ENTRY block, move them into single succ of ENTRY in that case.
10411 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10413 PR tree-optimization/84811
10414 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10415 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10416 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10417 (known_subrange_p): Update accordingly. Cast each value involved
10418 in the size comparison, rather than casting the result of the
10421 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10423 PR tree-optimization/84982
10424 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10425 by flipping the least significant bit rather than all bits from
10426 bitpos to bitpos + bitsize - 1.
10428 2018-03-21 Nathan Sidwell <nathan@acm.org>
10430 * doc/extend.texi (Deprecated Features): Remove mention of
10431 long-deleted deprecations.
10433 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10436 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10437 * configure: Regenerate.
10439 2018-03-21 Tom de Vries <tom@codesourcery.com>
10441 PR tree-optimization/83126
10442 * tree-parloops.c (num_phis): New function.
10443 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10445 2018-03-21 Nathan Sidwell <nathan@acm.org>
10447 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10448 mention anon-struct/union members and trailing attributes.
10450 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10452 PR tree-optimization/84969
10453 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10454 builtin memset partitions if they set different rhs values.
10456 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10458 PR rtl-optimization/84989
10459 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10460 VEC_DUPLICATE with scalar result mode.
10462 2018-03-21 Martin Liska <mliska@suse.cz>
10465 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10466 not intended return statement.
10468 2018-03-21 Martin Liska <mliska@suse.cz>
10471 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10472 (chkp_find_bound_slots_1): Limit number of iterations.
10474 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10477 * Minor grammar fixes for x86 options.
10479 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10482 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10483 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10485 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10488 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10489 (altivec_lvx_<mode>_1op): Likewise.
10490 (altivec_stvx_<mode>_2op): Likewise.
10491 (altivec_stvx_<mode>_1op): Likewise.
10492 (altivec_lvx_<VM2:mode>): New define_expand.
10493 (altivec_stvx_<VM2:mode>): Likewise.
10494 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10495 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10496 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10497 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10498 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10499 (rs6000_gen_lvx): Likewise.
10500 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10501 (altivec_expand_stv_builtin): Likewise.
10502 (altivec_expand_builtin): Likewise.
10503 * config/rs6000/vector.md: Likewise.
10505 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10508 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10511 2018-03-20 Richard Biener <rguenther@suse.de>
10514 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10515 sign-conversions as zero, fall back to standard scalar_stmt
10518 2018-03-20 Martin Liska <mliska@suse.cz>
10521 * predict.c (rebuild_frequencies): Handle case when we have
10522 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10524 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10527 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10528 flag_section_anchors.
10529 * varasm.c (use_blocks_for_decl_p): Remove hack for
10530 dw2_force_const_mem.
10533 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10535 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10536 be created, use lowpart_subreg of operands[0] rather than operands[0]
10538 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10539 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10540 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10541 and n constraint instead of aarch64_shift_imm_di and Usd.
10542 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10543 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10545 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10548 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10549 to favor GPR over NEON registers.
10550 (<shift>di3_neon): Likewise.
10552 2018-03-20 Tom de Vries <tom@codesourcery.com>
10555 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10556 (nvptx_process_pars): Emit bar.sync asap and alap.
10558 2018-03-20 Tom de Vries <tom@codesourcery.com>
10561 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10562 seen_label if seen_label is already set.
10564 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10567 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10568 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10569 Use 1U instead of 1. Formatting fixes.
10572 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10573 instead of TREE_TYPE (s1) for the return value.
10575 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10577 PR tree-optimization/84946
10578 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10579 bitsize + bitsize in poly_uint64 rather than poly_int64.
10582 * dwarf2asm.c: Include fold-const.c.
10583 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10584 of decl rather than decl itself.
10586 PR rtl-optimization/84643
10587 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10589 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10592 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10593 calling assemble_variable.
10595 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10598 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10599 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10601 2018-03-19 Jim Wilson <jimw@sifive.com>
10604 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10605 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10606 (riscv_first_stack_step): Likewise.
10607 (riscv_option_override): Use STACK_BOUNDARY instead of
10608 MIN_STACK_BOUNDARY.
10609 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10610 MIN_STACK_BOUNDARY.
10611 (BIGGEST_ALIGNMENT): Set to 128.
10612 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10613 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10616 2018-03-19 Richard Biener <rguenther@suse.de>
10618 PR tree-optimization/84933
10619 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10620 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10622 2018-03-19 Richard Biener <rguenther@suse.de>
10624 PR tree-optimization/84859
10625 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10626 (cond_if_else_store_replacement): Perform sinking operation on
10627 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10628 Generalize what a BB with a single eligible store is.
10630 2018-03-19 Richard Biener <rguenther@suse.de>
10632 PR tree-optimization/84929
10633 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10634 chrec_is_positive against non-chrec arg.
10636 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10639 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10641 2018-03-18 Martin Liska <mliska@suse.cz>
10643 PR rtl-optimization/84635
10644 * regrename.c (build_def_use): Use matches_mode only when
10647 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10649 PR tree-optimization/84913
10650 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10651 vectorize chains of COND_EXPRs.
10653 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10655 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10657 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10659 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10661 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10663 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10665 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10666 Kito Cheng <kito.cheng@gmail.com>
10668 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10669 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10670 (nds32_adjust_reg_alloc_order): New function.
10671 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10673 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10675 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10676 nds32_print_operand, nds32_print_operand_address): Use
10677 HOST_WIDE_INT_PRINT_DEC instead.
10679 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10681 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10683 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10686 * config/i386/i386.c (initial_ix86_tune_features,
10687 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10688 unsigned long long.
10689 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10690 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10691 rather than 1u << ix86_tune. Formatting fix.
10692 (ix86_option_override_internal): Change ix86_arch_mask from
10693 unsigned int to unsigned HOST_WIDE_INT, initialize to
10694 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10695 (ix86_function_specific_restore): Likewise.
10697 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10700 * postreload.c (reload_combine_recognize_pattern): Perform
10701 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10702 truncate_int_for_mode the result for the destination's mode.
10705 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10706 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10708 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10711 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10714 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10715 regno_allocno_class_array and sorted_pseudos.
10716 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10717 insns where regno is used.
10719 2018-03-16 Martin Liska <mliska@suse.cz>
10722 * multiple_target.c (create_dispatcher_calls): Redirect
10723 reference in the symbol table.
10725 2018-03-16 Martin Liska <mliska@suse.cz>
10728 * multiple_target.c (create_dispatcher_calls): Redirect also
10731 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10735 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10736 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10737 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10739 2018-03-16 Julia Koval <julia.koval@intel.com>
10741 * doc/invoke.texi (Skylake Server): Add CLWB.
10742 Cannonlake): Remove CLWB.
10744 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10746 PR tree-optimization/84841
10747 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10749 (FLOAT_ONE_CONST_TYPE): Define.
10750 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10751 (sort_by_operand_rank): Put entries with higher constant_type last
10752 rather than first to match comments.
10754 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10756 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10759 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10762 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10763 check and use address of target always.
10765 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10768 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10769 (indirect_thunk_bnd_needed): Likewise.
10770 (indirect_thunks_used): Likewise.
10771 (indirect_thunks_bnd_used): Likewise.
10772 (indirect_return_needed): New.
10773 (indirect_return_bnd_needed): Likewise.
10774 (output_indirect_thunk_function): Add a bool argument for
10776 (output_indirect_thunk_function): Don't generate alias for
10777 function return thunk.
10778 (ix86_code_end): Call output_indirect_thunk_function to generate
10779 function return thunks.
10780 (ix86_output_function_return): Set indirect_return_bnd_needed
10781 and indirect_return_needed instead of indirect_thunk_bnd_needed
10782 and indirect_thunk_needed.
10784 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10786 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10787 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10788 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10790 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10791 Paul Hua <paul.hua.gm@gmail.com>
10794 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10796 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10798 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10799 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10800 resp. SFmode cases.
10802 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10805 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10806 instead of GET_MODE_SIZE when comparing Units.
10808 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10811 * varasm.c (hash_section): Return an unchangeble hash value
10812 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10813 Return !aarch64_can_use_per_function_literal_pools_p ().
10815 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10818 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10819 rather than address of cmode as last argument to prepare_cmp_insn.
10821 2018-03-15 Julia Koval <julia.koval@intel.com>
10823 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10824 F_AVX512VNNI, F_AVX512BITALG): New.
10826 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10829 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10830 insn for floating-point loads and stores.
10832 2018-03-14 Carl Love <cel@us.ibm.com>
10834 * config/rs6000/rs6000-c.c: Add macro definitions for
10835 ALTIVEC_BUILTIN_VEC_PERMXOR.
10836 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10837 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10838 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10839 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10841 * config/doc/extend.texi: Add prototypes for vec_permxor.
10843 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10846 * diagnostic-show-locus.c (class layout_point): Convert m_line
10847 from int to linenum_type.
10848 (line_span::comparator): Use linenum "compare" function when
10849 comparing line numbers.
10850 (test_line_span): New function.
10851 (layout_range::contains_point): Convert param "row" from int to
10853 (layout_range::intersects_line_p): Likewise.
10854 (layout::will_show_line_p): Likewise.
10855 (layout::print_source_line): Likewise.
10856 (layout::should_print_annotation_line_p): Likewise.
10857 (layout::print_annotation_line): Likewise.
10858 (layout::print_leading_fixits): Likewise.
10859 (layout::annotation_line_showed_range_p): Likewise.
10860 (struct line_corrections): Likewise for field m_row.
10861 (line_corrections::line_corrections): Likewise for param "row".
10862 (layout::print_trailing_fixits): Likewise.
10863 (layout::get_state_at_point): Likewise.
10864 (layout::get_x_bound_for_row): Likewise.
10865 (layout::print_line): Likewise.
10866 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10867 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10868 * input.c (selftest::test_linenum_comparisons): New function.
10869 (selftest::input_c_tests): Call it.
10870 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10871 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10872 * selftest.h (ASSERT_GT): New macro.
10873 (ASSERT_GT_AT): New macro.
10874 (ASSERT_LT): New macro.
10875 (ASSERT_LT_AT): New macro.
10877 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10879 PR rtl-optimization/84780
10880 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10882 2018-03-14 Martin Liska <mliska@suse.cz>
10884 * tree.c (record_node_allocation_statistics): Use
10885 get_stats_node_kind.
10886 (get_stats_node_kind): New function extracted from
10887 record_node_allocation_statistics.
10888 (free_node): Use get_stats_node_kind.
10890 2018-03-14 Richard Biener <rguenther@suse.de>
10892 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10893 that the value-set of ANTIC_IN doesn't grow.
10896 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10898 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10899 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10901 2018-03-14 Julia Koval <julia.koval@intel.com>
10903 * config.gcc (icelake-client, icelake-server): New.
10905 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10906 (initial_ix86_arch_features): Ditto.
10907 (PTA_SKYLAKE): Add SGX.
10908 (PTA_ICELAKE): Remove.
10909 (PTA_ICELAKE_CLIENT): New.
10910 (PTA_ICELAKE_SERVER): New.
10911 (ix86_option_override_internal): Split up icelake on icelake client and
10913 (get_builtin_code_for_version): Ditto.
10914 (fold_builtin_cpu): Ditto.
10915 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10916 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10917 * config/i386/i386.h (processor_type): Ditto.
10918 * doc/invoke.texi: Ditto.
10920 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10923 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10924 INTEGER_CST offset, add it together with bitpos / 8 and
10925 sign extend based on POINTER_SIZE.
10929 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10932 * config/i386/constraints.md (Yc): New register constraint.
10933 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10934 Use Yc constraint for alternative 2 of operand 0. Remove
10935 preferred_for_speed attribute.
10937 2018-03-14 Richard Biener <rguenther@suse.de>
10939 PR tree-optimization/84830
10940 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10941 with the old one to avoid oscillations.
10943 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10946 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10948 (assign_by_spills): Return a flag of reload assignment failure.
10949 Do not process the reload assignment failures. Do not spill other
10950 reload pseudos if they has the same reg class. Update n if
10952 (lra_assign): Add a return arg. Set up from the result of
10953 assign_by_spills call.
10954 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10955 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10956 usage_insns if it is not NULL.
10957 (spill_hard_reg_in_range): New function.
10958 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10959 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10960 function prototypes.
10961 (lra_assign): Change prototype.
10962 * lra.c (lra): Add code to deal with fails by splitting hard reg
10965 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10967 * config/riscv/riscv.opt (mrelax): New option.
10968 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10969 "norelax" when riscv_mrelax is disabled.
10970 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10972 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10975 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10976 reassociation for int modes.
10978 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10980 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10981 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10983 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10984 * config/aarch64/aarch64-sve.md
10985 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10986 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10987 (*extend<mode><Vwide>2): Rename to...
10988 (aarch64_sve_extend<mode><Vwide>2): ...this.
10989 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10990 renaming the old pattern to...
10991 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10993 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10994 define_expand, renaming the old pattern to...
10995 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10996 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10997 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10998 account when deciding which SVE instruction the optab should use.
10999 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11001 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11003 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11004 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11005 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11006 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11007 (tlsdesc_small_<mode>): Turn a define_expand and use
11008 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11009 (tlsdesc_small_advsimd_<mode>): ...this.
11010 (tlsdesc_small_sve_<mode>): New pattern.
11012 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11014 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11015 (UNSPEC_UMUL_HIGHPART): New constants.
11016 (MUL_HIGHPART): New int iteraor.
11017 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11018 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11020 (*<su>mul<mode>3_highpart): New define_insn.
11022 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11025 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11028 2018-03-13 Martin Liska <mliska@suse.cz>
11031 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11033 (sem_item_optimizer::~sem_item_optimizer): Release it.
11034 (sem_item_optimizer::merge_classes): Register variable aliases.
11035 (sem_item_optimizer::fixup_pt_set): New function.
11036 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11037 * ipa-icf.h: Declare new variables and functions.
11039 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11041 PR middle-end/84834
11042 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11043 integer_pow2p@2 and test integer_pow2p in condition.
11044 (A < 0 ? C : 0): Similarly for @1.
11046 PR middle-end/84831
11047 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11048 characters starting at p contain '\0' character, don't look beyond
11052 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11053 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11056 * reg-stack.c (change_stack): Change update_end var from int to
11057 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11058 also call set_block_for_insn on the newly added insns and rescan.
11061 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11062 on the last operand.
11065 * tree.c (stabilize_reference_1): Return save_expr (e) for
11066 STATEMENT_LIST even if it doesn't have side-effects.
11068 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11070 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11072 2018-03-12 Renlin Li <renlin.li@arm.com>
11074 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11075 aarch64_output_scalar_simd_mov_immediate.
11077 2018-03-12 Martin Sebor <msebor@redhat.com>
11079 PR tree-optimization/83456
11080 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11081 for perfectly overlapping calls to memcpy.
11082 (gimple_fold_builtin_memory_chk): Same.
11083 (gimple_fold_builtin_strcpy): Handle no-warning.
11084 (gimple_fold_builtin_stxcpy_chk): Same.
11085 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11087 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11089 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11090 parameter. Use it for SFmode.
11091 (rs6000_function_arg_advance_1): Adjust.
11092 (rs6000_function_arg): Adjust.
11093 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11095 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11097 PR rtl-optimization/84169
11098 PR rtl-optimization/84780
11099 * combine.c (can_combine_p): Check for a 2-insn combination whether
11100 the destination register is used between the two insns, too.
11102 2018-03-12 Richard Biener <rguenther@suse.de>
11104 PR tree-optimization/84803
11105 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11106 for refs DR analysis didn't process.
11108 2018-03-12 Richard Biener <rguenther@suse.de>
11110 PR tree-optimization/84777
11111 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11112 force-vectorize loops ignore whether we are optimizing for size.
11114 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11116 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11117 (TARGET_MD_ASM_ADJUST): Define.
11119 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11120 Kito Cheng <kito.cheng@gmail.com>
11121 Chung-Ju Wu <jasonwucj@gmail.com>
11123 * config/nds32/nds32.c (nds32_compute_stack_frame,
11124 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11125 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11126 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11127 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11128 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11129 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11130 * config/nds32/nds32.md (prologue, epilogue): Use macro
11131 NDS32_V3PUSH_AVAILABLE_P to do checking.
11133 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11136 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11137 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11138 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11139 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11140 addition of most attributes on !orig_type_die or the attribute not
11141 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11143 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11144 Chung-Ju Wu <jasonwucj@gmail.com>
11146 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11147 __NDS32_VH__ macro.
11148 * config/nds32/nds32.opt (mvh): New option.
11150 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11151 Chung-Ju Wu <jasonwucj@gmail.com>
11153 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11155 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11156 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11159 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11160 Chung-Ju Wu <jasonwucj@gmail.com>
11162 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11164 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11165 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11167 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11168 Kito Cheng <kito.cheng@gmail.com>
11169 Chung-Ju Wu <jasonwucj@gmail.com>
11171 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11172 UNSPEC_FFMISM and UNSPEC_FLMISM.
11173 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11174 for ffb, ffmism and flmism.
11175 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11176 (unspec_ffmism): Ditto.
11177 (unspec_flmism): Ditto.
11178 (nds32_expand_builtin_impl): Check if string extension is available.
11179 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11180 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11182 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11185 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11188 * lra-assigns.c (assign_by_spills): Return a flag of reload
11189 assignment failure. Do not process the reload assignment
11190 failures. Do not spill other reload pseudos if they has the same
11192 (lra_assign): Add a return arg. Set up from the result of
11193 assign_by_spills call.
11194 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11195 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11196 usage_insns if it is not NULL.
11197 (spill_hard_reg_in_range): New function.
11198 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11199 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11200 function prototypes.
11201 (lra_assign): Change prototype.
11202 * lra.c (lra): Add code to deal with fails by splitting hard reg
11205 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11208 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11210 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11213 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11214 (dw_val_node): Add val_symbolic_view.
11215 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11216 (symview_upper_bound): New.
11217 (new_line_info_table): Initialize symviews_since_reset.
11218 (dwarf2out_source_line): Count symviews_since_reset and set
11219 symview_upper_bound.
11220 (dw_val_equal_p): Handle symview.
11221 (add_AT_symview): New.
11222 (print_dw_val): Handle symview.
11223 (attr_checksum, attr_checksum_ordered): Likewise.
11224 (same_dw_val_p, size_of_die): Likewise.
11225 (value_format, output_die): Likewise.
11226 (add_high_low_attributes): Use add_AT_symview for entry_view.
11227 (dwarf2out_finish): Reset symview_upper_bound, clear
11230 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11233 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11234 Add strict argument and use it.
11235 (rs6000_split_multireg_move): Update for new strict argument.
11236 (mem_operand_gpr): Disallow all non-offsettable addresses.
11237 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11239 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11242 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11243 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11244 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11247 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11248 decl, use remap_type if we want to use the type.
11250 2018-03-09 Martin Sebor <msebor@redhat.com>
11252 PR tree-optimization/84526
11253 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11255 (builtin_access::generic_overlap): Be prepared to handle non-array
11258 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11260 PR rtl-optimization/84682
11261 * lra-constraints.c (process_address_1): Check is_address flag
11262 for address constraints.
11263 (process_alt_operands): Likewise.
11264 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11265 preprocess_constraints.
11266 * recog.h (preprocess_constraints): Add oploc parameter.
11268 * recog.c (preprocess_constraints): Test address_operand for
11269 CT_ADDRESS constraints.
11271 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11274 * lra-assigns.c (assign_by_spills): Return a flag of reload
11275 assignment failure. Do not process the reload assignment
11276 failures. Do not spill other reload pseudos if they has the same
11278 (lra_assign): Add a return arg. Set up from the result of
11279 assign_by_spills call.
11280 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11281 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11282 usage_insns if it is not NULL.
11283 (spill_hard_reg_in_range): New function.
11284 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11285 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11286 function prototypes.
11287 (lra_assign): Change prototype.
11288 * lra.c (lra): Add code to deal with fails by splitting hard reg
11291 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11294 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11295 Accept complain bool parameter. Only emit errors if it is true.
11296 (arm_parse_cpu_option_name): Likewise.
11297 (arm_target_thumb_only): Adjust callers of the above.
11298 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11299 prototype to take a default true bool parameter.
11300 (arm_parse_arch_option_name): Likewise.
11302 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11303 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11307 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11308 * configure: Regenerate.
11309 * configure.ac ("linker --version-script option"): New.
11310 ("linker soname option"): New.
11312 2018-03-09 Richard Biener <rguenther@suse.de>
11314 PR tree-optimization/84775
11315 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11316 immediate uses of predicate stmts and mark them modified.
11319 PR tree-optimization/84178
11320 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11322 (version_loop_for_if_conversion): Delay update_ssa call.
11323 (tree_if_conversion): Delay update_ssa until after predicate
11326 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11329 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11330 when the function accesses prior frames.
11332 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11335 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11336 gen_llsym, otherwise call maybe_gen_llsym.
11338 PR inline-asm/84742
11339 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11340 has ',' character inside of it.
11342 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11345 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11346 as clobbering CC_REGNUM.
11348 2018-03-08 Richard Biener <rguenther@suse.de>
11350 PR middle-end/84552
11351 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11352 (follow_copies_to_constant): Do not follow SSA names registered
11355 2018-03-08 Richard Biener <rguenther@suse.de>
11357 PR tree-optimization/84178
11358 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11360 (version_loop_for_if_conversion): Delay update_ssa call.
11361 (tree_if_conversion): Delay update_ssa until after predicate
11364 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11366 PR tree-optimization/84178
11367 * tree-if-conv.c (release_bb_predicate): Remove the
11368 the assertion that the stmts have NULL use_ops.
11369 Discard the statements, asserting that they haven't
11370 yet been added to a BB.
11372 2018-03-08 Richard Biener <rguenther@suse.de>
11374 PR tree-optimization/84746
11375 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11376 (phi_translate): Pass in destination ANTIC_OUT set.
11377 (phi_translate_1): Likewise. For a simplified result lookup
11378 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11379 (phi_translate_set): Adjust.
11380 (do_pre_regular_insertion): Likewise.
11381 (do_pre_partial_partial_insertion): Likewise.
11383 2018-03-08 Martin Liska <mliska@suse.cz>
11385 PR gcov-profile/84735
11386 * doc/gcov.texi: Document usage of profile files.
11387 * gcov-io.h: Document changes in the format.
11389 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11393 * dwarf2out.c (struct dw_line_info_table): Update comments for
11395 (FORCE_RESET_NEXT_VIEW): New.
11396 (FORCE_RESETTING_VIEW_P): New.
11397 (RESETTING_VIEW_P): Check for -1 too.
11398 (ZERO_VIEW_P): Likewise.
11399 (new_line_info_table): Force-reset next view.
11400 (dwarf2out_begin_function): Likewise.
11401 (dwarf2out_source_line): Simplify zero_view_p initialization.
11402 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11403 view directly. Omit view when omitting .loc at line 0.
11405 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11407 PR tree-optimization/84740
11408 * tree-switch-conversion.c (process_switch): Call build_constructors
11409 only if info.phi_count is non-zero.
11411 PR tree-optimization/84739
11412 * tree-tailcall.c (find_tail_calls): Check call arguments against
11413 DECL_ARGUMENTS (current_function_decl) rather than
11414 DECL_ARGUMENTS (func) when checking for tail recursion.
11416 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11418 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11419 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11420 Volker Reichelt's entry and add entries for people that perform
11421 GCC fuzzy testing and report numerous bugs.
11423 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11426 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11427 readonly data in sdata, if that is disabled.
11428 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11429 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11430 -mreadonly-in-sdata option.
11432 2018-03-07 Martin Sebor <msebor@redhat.com>
11434 PR tree-optimization/84468
11435 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11436 basic block when looking for nul assignment.
11438 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11441 * except.h (output_function_exception_table): Adjust prototype.
11442 * except.c (output_function_exception_table): Remove FNNAME parameter
11443 and add SECTION parameter. Ouput one part of the table at a time.
11444 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11445 the first part of the exception table and emit unwind directives.
11446 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11447 (i386_pe_seh_cold_init): Likewise.
11448 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11449 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11450 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11451 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11452 * config/i386/winnt.c: Include except.h.
11453 (struct seh_frame_state): Add reg_offset, after_prologue and
11454 in_cold_section fields.
11455 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11456 (i386_pe_seh_cold_init): New function.
11457 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11458 to seh->in_cold_section.
11459 (seh_emit_push): Record the offset of the push.
11460 (seh_emit_save): Record the offet of the save.
11461 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11462 Test seh->after_prologue to disregard the epilogue.
11463 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11464 (i386_pe_end_cold_function): New function.
11466 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11469 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11470 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11473 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11475 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11476 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11478 PR middle-end/84723
11479 * multiple_target.c: Include tree-inline.h and intl.h.
11480 (expand_target_clones): Diagnose and fail if node->definition and
11481 !tree_versionable_function_p (node->decl).
11483 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11485 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11487 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11488 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11489 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11491 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11494 * combine.c (try_combine): Use reg_or_subregno instead of handling
11495 just paradoxical SUBREGs and REGs.
11497 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11499 * config/arc/arc.c (arc_finalize_pic): Remove function.
11500 (arc_must_save_register): We use single base PIC register, remove
11501 checks to save/restore the PIC register.
11502 (arc_expand_prologue): Likewise.
11503 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11505 (arc_verify_short): Likewise.
11506 (arc_attr_type): Likewise.
11507 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11508 (walk_stores): Likewise.
11509 (arc_address_cost): Make it static.
11510 (arc_verify_short): Likewise.
11511 (branch_dest): Likewise.
11512 (arc_attr_type): Likewise.
11513 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11514 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11515 (arc_final_prescan_insn): Remove inserting the nops due to
11516 hardware hazards. It is done in reorg step.
11517 (insn_length_variant_t): Remove.
11518 (insn_length_parameters_t): Likewise.
11519 (arc_insn_length_parameters): Likewise.
11520 (arc_get_insn_variants): Likewise.
11521 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11523 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11525 PR inline-asm/84683
11526 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11529 PR tree-optimization/84687
11530 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11532 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11534 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11536 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11537 Rename to ppc_speculation_barrier.
11538 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11539 __builtin_ppc_speculation_barrier.
11541 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11544 * combine.c (combine_simplify_rtx): Don't try to simplify if
11545 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11548 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11550 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11551 to 32 bytes when compiling for POWER9.
11553 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11556 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11557 regparm >= 3 with no arg reg available also for calls with
11558 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11561 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11563 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11565 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11568 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11570 2018-03-05 Richard Biener <rguenther@suse.de>
11572 PR tree-optimization/84486
11573 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11574 When inserting a __builtin_assume_aligned call set the LHS
11575 SSA name alignment info accordingly.
11577 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11579 PR tree-optimization/84114
11580 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11581 Avoid reassociation of FLOAT_MODE addition.
11583 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11585 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11586 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11587 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11588 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11589 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11590 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11591 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11593 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11594 __builtin_ia32_wbinvd): New builtins.
11595 (SPECIAL_ARGS2): New.
11596 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11597 (SPECIAL_ARGS2): New.
11598 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11599 (ix86_valid_target_attribute_inner_p): Ditto.
11600 (ix86_init_mmx_sse_builtins): Add special_args2.
11601 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11602 TARGET_WBNOINVD_P): New.
11603 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11604 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11605 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11606 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11607 * config/i386/pconfigintrin.h: New file.
11608 * config/i386/wbnoinvdintrin.h: Ditto.
11609 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11611 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11613 2018-03-05 Richard Biener <rguenther@suse.de>
11615 PR tree-optimization/84670
11616 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11618 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11619 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11620 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11621 doesn't grow if all successors (recursively) were visited at least
11624 2018-03-05 Richard Biener <rguenther@suse.de>
11626 PR tree-optimization/84650
11627 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11628 if executed in the loop pipeline.
11630 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11632 * doc/configfiles.texi (Configuration Files): Move info about
11633 conditionalizing $target-protos.h to...
11634 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11635 differs from $target-protos.h.
11637 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11638 Chung-Ju Wu <jasonwucj@gmail.com>
11640 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11641 * config/nds32/nds32-multiple.md (setmemsi): Define.
11642 * config/nds32/nds32-memory-manipulation.c
11643 (nds32_gen_dup_4_byte_to_word_value): New.
11644 (emit_setmem_word_loop): New.
11645 (emit_setmem_byte_loop): New.
11646 (nds32_expand_setmem_loop): New.
11647 (nds32_expand_setmem_loop_v3m): New.
11648 (nds32_expand_setmem_unroll): New.
11649 (nds32_expand_setmem): New.
11651 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11652 Chung-Ju Wu <jasonwucj@gmail.com>
11654 * config/nds32/nds32-memory-manipulation.c
11655 (nds32_emit_load_store): New.
11656 (nds32_emit_post_inc_load_store): New.
11657 (nds32_emit_mem_move): New.
11658 (nds32_emit_mem_move_block): New.
11659 (nds32_expand_movmemsi_loop_unknown_size): New.
11660 (nds32_expand_movmemsi_loop_known_size): New.
11661 (nds32_expand_movmemsi_loop): New.
11662 (nds32_expand_movmemsi_unroll): New.
11663 (nds32_expand_movmemqi): Rename ...
11664 (nds32_expand_movmemsi): ... to this.
11665 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11666 (movmemsi): ... to this.
11667 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11668 (nds32_expand_movmemsi): ... to this.
11670 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11671 Monk Chiang <sh.chiang04@gmail.com>
11672 Chung-Ju Wu <jasonwucj@gmail.com>
11674 * config/nds32/nds32-protos.h
11675 (nds32_expand_load_multiple): New arguments.
11676 (nds32_expand_store_multiple): Ditto.
11677 (nds32_valid_multiple_load_store): Rename ...
11678 (nds32_valid_multiple_load_store_p): ... to this.
11679 * config/nds32/nds32-memory-manipulation.c
11680 (nds32_expand_load_multiple): Refine implementation.
11681 (nds32_expand_store_multiple): Ditto.
11682 * config/nds32/nds32-multiple.md
11683 (load_multiple): Update nds32_expand_load_multiple interface.
11684 (store_multiple): Update nds32_expand_store_multiple interface.
11685 * config/nds32/nds32-predicates.c
11686 (nds32_valid_multiple_load_store): Rename ...
11687 (nds32_valid_multiple_load_store_p): ... to this and refine
11689 * config/nds32/predicates.md
11690 (nds32_load_multiple_and_update_address_operation): New predicate.
11691 (nds32_store_multiple_and_update_address_operation): New predicate.
11693 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11694 Chung-Ju Wu <jasonwucj@gmail.com>
11696 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11697 (combo): New attribute.
11698 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11700 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11702 * config/nds32/nds32.opt: Change -mcmodel= default value.
11704 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11705 Monk Chiang <sh.chiang04@gmail.com>
11706 Chung-Ju Wu <jasonwucj@gmail.com>
11708 * config/nds32/constants.md (unspec_element): New enum.
11709 * config/nds32/constraints.md (Umw): New constraint.
11710 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11711 * config/nds32/nds32-intrinsic.md: Likewise.
11712 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11713 (nds32_valid_smw_lwm_base_p): New.
11714 (nds32_output_smw_single_word): New.
11715 (nds32_output_lmw_single_word): New.
11716 (nds32_expand_unaligned_load): New.
11717 (nds32_expand_unaligned_store): New.
11718 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11719 (nds32_output_smw_single_word): Declare.
11720 (nds32_output_lmw_single_word): Declare.
11721 (nds32_expand_unaligned_load): Declare.
11722 (nds32_expand_unaligned_store): Declare.
11723 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11724 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11725 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11726 NDS32_BUILTIN_UASTORE_DW.
11727 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11730 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11731 Kito Cheng <kito.cheng@gmail.com>
11732 Chung-Ju Wu <jasonwucj@gmail.com>
11734 * config/nds32/nds32-intrinsic.c
11735 (nds32_expand_builtin_null_ftype_reg): Delete.
11736 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11737 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11738 (nds32_read_argument): New.
11739 (nds32_legitimize_target): Ditto.
11740 (nds32_legitimize_argument): Ditto.
11741 (nds32_check_constant_argument): Ditto.
11742 (nds32_expand_unop_builtin): Ditto.
11743 (nds32_expand_unopimm_builtin): Ditto.
11744 (nds32_expand_binop_builtin): Ditto.
11745 (nds32_builtin_decl_impl): Ditto.
11746 (builtin_description): Ditto.
11747 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11748 (nds32_init_builtins_impl): Ditto.
11749 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11750 (nds32_builtin_decl): New.
11751 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11752 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11754 2018-03-02 Jeff Law <law@redhat.com>
11756 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11757 (redundant_insn, fill_simple_delay_slots): Likewise.
11758 (fill_slots_from_thread): Likewise.
11759 * resource.c (mark_referenced_resources): Likewise.
11760 (mark_set_resources, find_dead_or_set_registers): Likewise.
11762 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11764 * substring-locations.h (format_warning_va): Formatting fix for
11765 ATTRIBUTE_GCC_DIAG.
11766 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11768 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11769 * substring-locations.c: Include intl.h.
11770 (format_warning_va): Turned into small wrapper around
11771 format_warning_n_va, renamed to ...
11772 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11773 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11775 (format_warning_at_substring_n): New function.
11776 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11777 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11778 format_warning_at_substring with just a shorter name instead of
11779 const function pointer.
11780 (fmtwarn_n): New function.
11781 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11782 appropriate, get rid of all the fmtstr temporaries, move conditionals
11783 with G_() wrapped string literals directly into fmtwarn arguments,
11784 cast dir.len to (int), formatting fixes.
11786 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11788 * doc/invoke.texi: Remove "Cilk Plus" references.
11790 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11791 Richard Biener <rguenther@suse.de>
11794 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11795 for error or warning attributes if CALL_FROM_THUNK_P is set.
11798 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11801 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11802 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11805 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11806 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11808 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11809 instead of -1U in last predictors element's probability member.
11811 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11814 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11815 arguments if they are comparable.
11817 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11819 PR tree-optimization/84634
11820 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11821 masks and masked_loop_p with a single loop_masks, making sure it's
11822 null for bb vectorization.
11824 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11826 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11827 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11828 loop->force_vectorize to check whether there is no alias.
11830 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11833 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11835 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11837 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11838 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11839 instead of a loop around prev_real_insn.
11840 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11843 PR inline-asm/84625
11844 * config/i386/i386.c (ix86_print_operand): Use conditional
11845 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11848 2018-03-02 Richard Biener <rguenther@suse.de>
11850 PR tree-optimization/84427
11851 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11852 (bitmap_set_subtract_values): Rewrite to handle multiple
11855 (prune_clobbered_mems): Likewise.
11856 (phi_translate): Take edge instead of pred/phiblock.
11857 (phi_translate_1): Likewise.
11858 (phi_translate_set): Likewise. Insert all translated
11859 exprs for a value into the set, keeping possibly multiple
11860 expressions per value.
11861 (compute_antic_aux): Adjust for phi_translate changes.
11862 When intersecting union the expressions and prune those
11863 not in the final value set, keeping possibly multiple
11864 expressions per value. Do not use value-insertion
11865 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11866 all expressions. Add verification that the value-sets
11867 only shrink during iteration.
11868 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11869 (do_pre_regular_insertion): Likewise.
11870 (do_pre_partial_partial_insertion): Likewise.
11872 2018-03-02 Richard Biener <rguenther@suse.de>
11875 * config/darwin.c (saved_debug_info_level): New static global.
11876 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11877 (darwin_asm_lto_end): Restore debug info generation settings.
11879 2018-03-01 Martin Liska <mliska@suse.cz>
11882 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11883 volatile arguments.
11885 2018-03-01 Richard Biener <rguenther@suse.de>
11888 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11889 type annotation with LTO when debug was disabled at compile-time.
11891 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11893 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11895 (mips_final_postscan_insn): Likewise.
11897 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11899 PR rtl-optimization/84528
11900 * alias.c (init_alias_target): Add commentary.
11901 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11902 a unique base value if the frame pointer is not eliminated
11903 to the stack pointer.
11905 2018-03-01 Tom de Vries <tom@codesourcery.com>
11907 PR rtl-optimization/83327
11908 * lra-int.h (hard_regs_spilled_into): Declare.
11909 * lra.c (hard_regs_spilled_into): Define.
11910 (init_reg_info): Init hard_regs_spilled_into.
11911 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11912 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11913 (process_bb_lives): Handle hard_regs_spilled_into.
11914 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11915 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11917 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11919 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11920 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11921 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11922 * config/rs6000/aix72.h: New file.
11924 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11926 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11927 instead of warning_at with conditional singular and plural messages
11931 * stor-layout.c (update_alignment_for_field): For
11932 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11933 && !DECL_PACKED (field), do the alignment update, just use
11934 only desired_align instead of MAX (type_align, desired_align)
11936 (place_field): Don't do known_align < desired_align handling
11937 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11938 is non-NULL, instead do it after rli->prev_field handling and
11939 only if not within a bitfield word. For DECL_PACKED (field)
11940 use type_align of BITS_PER_UNIT.
11942 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11944 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11945 superfluous parentheses and trailing spaces.
11947 2018-02-28 Richard Biener <rguenther@suse.de>
11949 PR tree-optimization/84584
11950 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11951 SCoPs with fake exit edge.
11953 2018-02-28 Martin Liska <mliska@suse.cz>
11956 * timevar.c (timer::print): Fix format to properly print 100%
11959 2018-02-28 Richard Biener <rguenther@suse.de>
11961 PR middle-end/84607
11962 * genmatch.c (capture_info::walk_match): Do not mark
11963 captured expressions without operands as expr_p given
11964 they act more like predicates and should be subject to
11965 "lost tail" side-effect preserving.
11967 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11969 PR rtl-optimization/81611
11970 * auto-inc-dec.c (attempt_change): Move dead note from
11971 mem_insn if it's the next use of regno
11972 (find_address): Take address use of reg holding
11973 non-incremented value. Add parm to limit search to the named
11975 (merge_in_block): Attempt to use a mem insn that is the next
11976 use of the original regno.
11978 2018-02-27 Martin Sebor <msebor@redhat.com>
11981 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11982 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11984 2018-02-27 Martin Sebor <msebor@redhat.com>
11986 PR translation/84207
11987 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11988 n argument to unsigned HOST_WIDE_INT.
11989 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11990 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11991 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11992 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11994 2018-02-27 Richard Biener <rguenther@suse.de>
11996 PR tree-optimization/84512
11997 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11998 Do not use the estimate returned from record_stmt_cost for
11999 the scalar iteration cost but sum properly using add_stmt_cost.
12001 2018-02-27 Richard Biener <rguenther@suse.de>
12003 PR tree-optimization/84466
12004 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12005 Adjust last change to less strictly validate use operands.
12007 2018-02-27 Martin Liska <mliska@suse.cz>
12009 PR gcov-profile/84548
12010 * gcov.c (process_file): Allow partial overlap and consider it
12011 also as group functions.
12012 (output_lines): Properly calculate range of lines for a group.
12014 2018-02-27 Martin Liska <mliska@suse.cz>
12016 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12017 'ggc' suffixes. Change first column width.
12018 (timer::print): Fix formatting of the column.
12020 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12022 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12023 preserve inline entry blocks for the sake of debug inline
12024 entry point markers alone.
12025 (remove_unused_locals): Suggest in comments a better place to
12026 force the preservation of inline entry blocks that are
12027 otherwise unused, but do not preserve them.
12029 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12031 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12033 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12036 * config/i386/constraints.md (Bs): Replace
12037 ix86_indirect_branch_register with
12038 TARGET_INDIRECT_BRANCH_REGISTER.
12040 * config/i386/i386.md (indirect_jump): Likewise.
12041 (tablejump): Likewise.
12042 (*sibcall_memory): Likewise.
12043 (*sibcall_value_memory): Likewise.
12044 Peepholes of indirect call and jump via memory: Likewise.
12045 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12046 (*sibcall_value_GOT_32): Likewise.
12047 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12048 (GOT_memory_operand): Likewise.
12049 (call_insn_operand): Likewise.
12050 (sibcall_insn_operand): Likewise.
12051 (GOT32_symbol_operand): Likewise.
12052 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12054 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12056 PR rtl-optimization/83496
12057 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12058 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12059 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12060 redundant insn, if any.
12061 (relax_delay_slots): Likewise.
12062 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12064 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12066 PR tree-optimization/83965
12067 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12068 that grouped statements are part of a reduction chain. Return
12069 true if the statement is not marked as a reduction itself but
12070 is part of a group.
12071 (vect_recog_dot_prod_pattern): Don't check whether the statement
12072 is part of a group here.
12073 (vect_recog_sad_pattern): Likewise.
12074 (vect_recog_widen_sum_pattern): Likewise.
12076 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12079 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12081 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12084 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12086 (ix86_output_indirect_function_return): New prototype.
12087 (ix86_split_simple_return_pop_internal): Likewise.
12088 * config/i386/i386.c (indirect_return_via_cx): New.
12089 (indirect_return_via_cx_bnd): Likewise.
12090 (indirect_thunk_name): Handle return va CX_REG.
12091 (output_indirect_thunk_function): Create alias for
12092 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12093 (ix86_output_indirect_jmp): Remove the bool argument.
12094 (ix86_output_indirect_function_return): New function.
12095 (ix86_split_simple_return_pop_internal): Likewise.
12096 * config/i386/i386.md (*indirect_jump): Don't pass false
12097 to ix86_output_indirect_jmp.
12098 (*tablejump_1): Likewise.
12099 (simple_return_pop_internal): Change it to define_insn_and_split.
12100 Call ix86_split_simple_return_pop_internal to split it for
12101 -mfunction-return=.
12102 (simple_return_indirect_internal): Call
12103 ix86_output_indirect_function_return instead of
12104 ix86_output_indirect_jmp.
12106 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12109 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12110 memset and value initialization afterwards.
12112 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12114 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12116 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12119 * common/config/aarch64/aarch64-common.c
12120 (aarch_option_optimization_table[]): Switch
12121 off fomit-frame-pointer
12123 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12124 Chung-Ju Wu <jasonwucj@gmail.com>
12126 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12128 (store_multiple): Ditto.
12130 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12132 * config.gcc: Add --with-cpu support for nds32 target.
12133 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12134 * config/nds32/nds32.opt: Add -mcpu= option.
12136 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12138 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12139 isel=yes): Warn for these deprecated options.
12141 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12143 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12144 ISA_2_5_MASKS_EMBEDDED.
12146 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12148 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12149 p->max as pointers rather than using iterative_hash_expr.
12151 2018-02-23 Carl Love <cel@us.ibm.com>
12153 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12154 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12156 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12157 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12158 P8V_BUILTIN_VEC_VUNSIGNED2.
12160 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12163 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12164 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12166 * lra-constraints.c (curr_insn_transform): Set up
12167 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12169 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12170 * lra-eliminations.c (spill_pseudos): Ditto.
12171 (process_insn_for_elimination): Ditto.
12172 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12173 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12174 LRA_NON_CLOBBERED_ALT.
12176 2018-02-22 Martin Sebor <msebor@redhat.com>
12178 PR tree-optimization/84480
12179 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12180 to maybe_diag_stxncpy_trunc. Call it.
12181 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12182 from gimple_fold_builtin_strcpy. Print inlining stack.
12183 (handle_builtin_stxncpy): Print inlining stack.
12184 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12186 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12189 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12190 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12191 and -fcheck-pointer-bounds are used together.
12192 (indirect_thunk_prefix): New enum.
12193 (indirect_thunk_need_prefix): New function.
12194 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12195 "_nt" instead of "_bnd" for NOTRACK prefix.
12196 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12197 (output_indirect_thunk_function): Likewise.
12199 (ix86_code_end): Update output_indirect_thunk_function calls.
12200 (ix86_output_indirect_branch_via_reg): Replace
12201 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12202 (ix86_output_indirect_branch_via_push): Likewise.
12203 (ix86_output_function_return): Likewise.
12204 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12205 incompatible with -fcf-protection=branch and
12206 -fcheck-pointer-bounds.
12208 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12211 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12212 Change gcc_assert call to output_operand_lossage.
12214 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12216 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12218 2018-02-22 DJ Delorie <dj@redhat.com>
12219 Sebastian Perta <sebastian.perta@renesas.com>
12220 Oleg Endo <olegendo@gcc.gnu.org>
12222 * config/rx/rx.c (rx_rtx_costs): New function.
12223 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12225 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12227 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12229 2018-02-22 Martin Liska <mliska@suse.cz>
12232 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12233 Add "native" as a possible value.
12235 2018-02-22 Martin Liska <mliska@suse.cz>
12238 * config/i386/i386.c (ix86_option_override_internal):
12239 Add "native" as a possible value for -march and -mtune.
12241 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12244 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12245 to all type variants.
12247 PR tree-optimization/84503
12248 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12249 width as info->bitpos + info->bitsize - start.
12250 (merged_store_group::merge_overlapping): Simplify width computation.
12251 (check_no_overlap): New function.
12252 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12253 start + width and last_order of the group, fail if check_no_overlap
12255 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12256 to group if check_no_overlap fails.
12258 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12260 * config/rs6000/altivec.md: Delete contraint arguments to
12261 define_expand, define_split, and define_peephole2, and in
12262 define_insn_and_split if always unused.
12263 * config/rs6000/darwin.md: Ditto.
12264 * config/rs6000/dfp.md: Ditto.
12265 * config/rs6000/rs6000.md: Ditto.
12266 * config/rs6000/sync.md: Ditto.
12267 * config/rs6000/vector.md: Ditto.
12268 * config/rs6000/vsx.md: Ditto.
12270 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12272 * config/rs6000/altivec.md: Write output control strings as braced
12273 blocks instead of double-quoted strings.
12274 * config/rs6000/darwin.md: Ditto.
12275 * config/rs6000/rs6000.md: Ditto.
12276 * config/rs6000/vector.md: Ditto.
12277 * config/rs6000/vsx.md: Ditto.
12279 2018-02-21 Jason Merrill <jason@redhat.com>
12281 PR c++/84314 - ICE with templates and fastcall attribute.
12282 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12284 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12286 * ipa-cp.c (determine_versionability): Fix comment typos.
12288 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12291 * ipa-cp.c (determine_versionability): Do not version functions caling
12294 2018-02-21 Martin Liska <mliska@suse.cz>
12297 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12298 Add "native" as a possible value.
12299 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12300 the macro when native cpu detection is available.
12302 2018-02-21 Martin Liska <mliska@suse.cz>
12305 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12306 Add "native" as a possible value.
12307 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12308 when native cpu detection is available.
12310 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12311 Martin Sebor <msebor@redhat.com>
12313 PR tree-optimization/84478
12314 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12316 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12317 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12318 support which is conservatively correct, for 2 only stay conservative
12319 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12320 argument to the 2 argument get_range_strlen, adjust 6 arg
12321 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12323 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12324 (gimple_fold_builtin_strlen): Pass true as last argument to
12327 2018-02-20 Martin Sebor <msebor@redhat.com>
12329 PR middle-end/84095
12330 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12331 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12332 (builtin_memref::builtin_memref): Factor out parts into
12333 set_base_and_offset and call it.
12335 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12337 PR middle-end/84406
12338 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12339 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12340 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12341 search at the associated MODE_INT.
12343 2018-02-20 Jeff Law <law@redhat.com>
12345 PR middle-end/82123
12346 PR tree-optimization/81592
12347 PR middle-end/79257
12348 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12349 for range data rather than using global data.
12350 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12351 range data rather than using global data.
12352 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12353 pass it to children as needed.
12354 (struct directive::fmtresult): Similarly.
12355 (struct directive::set_width): Similarly.
12356 (struct directive::set_precision): Similarly.
12357 (format_integer, format_directive, parse_directive): Similarly.
12358 (format_none): Accept unnamed vr_values parameter.
12359 (format_percent, format_floating, format_character): Similarly.
12360 (format_string, format_plain): Similarly.
12361 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12362 the EVRP range analyzer for range data rather than using global data.
12363 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12364 gimple-ssa-evrp-analyze.h
12365 (class sprintf_dom_walker): Add after_dom_children member function.
12366 Add evrp_range_analyzer member.
12367 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12368 range analyzer as needed.
12369 (sprintf_dom_walker::after_dom_children): New member function.
12370 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12372 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12373 (evrp_range_analyzer::pop_to_marker): Likewise.
12375 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12377 PR tree-optimization/84419
12378 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12379 with the required type if its current type is compatible but
12382 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12384 PR middle-end/82004
12385 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12386 after vectorization.
12388 2018-02-20 Martin Liska <mliska@suse.cz>
12391 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12392 possible values if we don't have a hint.
12394 2018-02-20 Martin Liska <mliska@suse.cz>
12398 * final.c (shorten_branches): Build align_tab array with one
12400 * opts.c (finish_options): Add alignment option limit check.
12401 (MAX_CODE_ALIGN): Likewise.
12402 (MAX_CODE_ALIGN_VALUE): Likewise.
12403 * doc/invoke.texi: Document maximum allowed option value for
12404 all -falign-* options.
12406 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12409 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12410 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12411 * var-tracking.c (emit_note_insn_var_location): Remove all references
12412 to NOTE_INSN_CALL_ARG_LOCATION.
12413 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12414 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12415 Use copy_rtx_if_shared.
12416 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12417 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12418 (dwarf2out_var_location): Remove handling of
12419 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12421 * final.c (final_scan_insn): Remove all references to
12422 NOTE_INSN_CALL_ARG_LOCATION.
12423 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12424 before dumping final insns.
12425 * except.c (emit_note_eh_region_end): Remove all references to
12426 NOTE_INSN_CALL_ARG_LOCATION.
12427 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12428 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12429 * config/arc/arc.c (hwloop_optimize): Likewise.
12430 * config/arm/arm.c (create_fix_barrier): Likewise.
12431 * config/s390/s390.c (s390_chunkify_start): Likewise.
12432 * config/sh/sh.c (find_barrier): Likewise.
12433 * config/i386/i386.c (rest_of_insert_endbranch,
12434 ix86_seh_fixup_eh_fallthru): Likewise.
12435 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12436 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12437 * config/frv/frv.c (frv_function_prologue): Likewise.
12438 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12440 (note_outside_basic_block_p): Remove all references to
12441 NOTE_INSN_CALL_ARG_LOCATION.
12442 * gengtype.c (adjust_field_rtx_def): Likewise.
12443 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12445 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12446 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12449 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12452 PR tree-optimization/84452
12453 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12454 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12457 2018-02-19 Martin Liska <mliska@suse.cz>
12460 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12462 2018-02-19 Martin Liska <mliska@suse.cz>
12463 Richard Sandiford <richard.sandiford@linaro.org>
12465 PR tree-optimization/82491
12466 * gimple-fold.c (get_base_constructor): Make earlier bail out
12469 2018-02-19 Carl Love <cel@us.ibm.com>
12471 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12472 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12474 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12475 P8V_BUILTIN_VEC_NEG.
12477 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12479 * config/rl78/rl78.md (movdf): New define expand.
12481 2018-02-19 Martin Liska <mliska@suse.cz>
12484 * doc/invoke.texi: Fix typo.
12485 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12487 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12489 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12490 handle rs6000_single_float and rs6000_double_float specially for
12493 2018-02-16 Jeff Law <law@redhat.com>
12495 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12496 (pop_regs): Use it.
12498 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12501 * ipa-inline.c (inline_small_functions): Fix a typo.
12503 2018-02-16 Nathan Sidwell <nathan@acm.org>
12505 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12507 2018-02-16 Carl Love <cel@us.ibm.com>
12509 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12510 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12511 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12512 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12513 expansion to P8V_BUILTIN_VEC_FLOAT2.
12515 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12517 PR rtl-optimization/70023
12518 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12519 src_regno into account.
12521 2018-02-16 Carl Love <cel@us.ibm.com>
12523 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12524 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12525 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12526 * config/rs6000/rs6000.c: Remove case statements for
12527 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12528 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12529 and P9V_BUILTIN_VEC_VINSERT4B.
12530 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12531 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12532 * config/rs6000/vsx.md:
12533 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12536 2018-02-16 Carl Love <cel@us.ibm.com>
12538 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12540 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12542 * config/rs6000/rs6000-c.c: Add the definitions for
12543 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12544 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12545 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12546 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12547 definition for insert4b and define insn *insert3b_internal.
12548 * doc/extend.texi: Add documentation for vec_extract4b.
12550 2018-02-16 Nathan Sidwell <nathan@acm.org>
12552 * doc/extend.texi (Backwards Compatibility): Mention friend
12553 injection. Note for-scope is deprecated.
12554 * doc/invoke.texi (-ffriend-injection): Deprecate.
12556 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12558 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12559 that moved to I2, also allow destinations that are a paradoxical
12560 subreg (instead of a normal reg).
12562 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12565 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12568 2018-02-16 Richard Biener <rguenther@suse.de>
12570 PR tree-optimization/84037
12571 PR tree-optimization/84016
12573 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12574 Adjust vec_construct for the fact we need additional higher latency
12575 128bit inserts for AVX256 and AVX512 vector builds.
12576 (ix86_add_stmt_cost): Scale vector construction cost for
12579 2018-02-16 Richard Biener <rguenther@suse.de>
12581 PR tree-optimization/84417
12582 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12583 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12584 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12586 2018-02-16 Martin Liska <mliska@suse.cz>
12589 * internal-fn.def (ASAN_CHECK): Set proper flags.
12590 (ASAN_MARK): Likewise.
12592 2018-02-16 Julia Koval <julia.koval@intel.com>
12594 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12595 from PTA_CANNONLAKE.
12597 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12600 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12601 Use ++iter rather than iter++ for std::list iterators.
12602 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12603 defer deleting them until all nodes in the forest are processed. Do
12604 free even leaf nodes. Change to_process into auto_vec.
12607 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12608 * vec.h (vec_default_construct): Use memset instead of placement new
12609 if BROKEN_VALUE_INITIALIZATION is defined.
12610 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12611 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12614 PR rtl-optimization/83723
12615 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12616 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12617 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12619 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12621 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12623 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12625 PR rtl-optimization/81443
12626 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12627 from inner REGs to paradoxical SUBREGs.
12629 2018-02-16 Richard Biener <rguenther@suse.de>
12631 PR tree-optimization/84399
12632 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12633 For operands we can analyze at their definition make sure we can
12634 analyze them at each use as well.
12636 2018-02-16 Richard Biener <rguenther@suse.de>
12638 PR tree-optimization/84190
12639 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12640 volatile accesses if the decl isn't volatile.
12642 2018-02-15 Jason Merrill <jason@redhat.com>
12644 PR c++/84314 - ICE with templates and fastcall attribute.
12645 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12646 TYPE_CANONICAL on an existing type.
12648 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12650 PR tree-optimization/84383
12651 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12652 dstoff nor call operand_equal_p if dstbase is NULL.
12654 PR tree-optimization/84334
12655 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12656 also a CONSTANT_CLASS_P, punt.
12658 2018-02-14 Jim Wilson <jimw@sifive.com>
12660 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12661 first SMALL_OPERAND check. New local min_second_step. Move assert
12662 to where locals are set. Add TARGET_RVC support.
12663 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12665 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12667 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12669 2018-02-14 Martin Sebor <msebor@redhat.com>
12671 PR tree-optimization/83698
12672 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12673 arrays constrain the offset range to their bounds.
12674 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12675 (builtin_access::overlap): Avoid setting the size of overlap if it's
12677 (maybe_diag_overlap): Also consider arrays when deciding what values
12678 of offsets to include in diagnostics.
12680 2018-02-14 Martin Sebor <msebor@redhat.com>
12683 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12684 that correspond to the kind of a declaration.
12686 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12689 * config/pa/pa.md: Load address of PIC label using the linkage table
12690 if the label is nonlocal.
12692 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12694 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12695 warning message if user requests -maltivec=be.
12696 * doc/invoke.texi: Document deprecation of -maltivec=be.
12698 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12701 * config/rs6000/rs6000-c.c: Update definitions for
12702 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12703 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12705 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12708 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12709 add _get_ssp intrinsics. Remove argument from
12710 __builtin_ia32_rdssp[d|q].
12711 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12712 * config/i386/i386-builtin.def: Remove argument from
12713 __builtin_ia32_rdssp[d|q].
12714 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12715 ix86_expand_special_args_builtin for _rdssp[d|q].
12716 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12717 Clear register before usage.
12718 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12719 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12721 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12723 PR tree-optimization/84357
12724 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12725 operand 1 of an ARRAY_REF too.
12727 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12730 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12731 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12733 (set_of_reg): New struct.
12734 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12735 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12736 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12738 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12739 Split into bitclr, bitset, bitinvert patterns if appropriate.
12740 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12741 use rx_fuse_in_memory_bitop.
12742 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12743 to named insn, correct maximum insn length.
12745 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12748 * machmode.def: Define a complex mode for PARTIAL_INT.
12749 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12751 * doc/rtl.texi: Document CSPImode.
12752 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12754 (msp430_hard_regno_nregs_with_padding): Likewise.
12756 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12759 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12761 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12763 PR rtl-optimization/84169
12764 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12765 we generated a parallel as new i3 and we split that to new i2 and i3
12766 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12767 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12768 those to i2, not i1. Partially rewrite this scan code.
12770 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12773 * stor-layout.c (place_field): For variable length fields, adjust
12774 offset_align afterwards not just based on the field's alignment,
12775 but also on the size.
12777 PR middle-end/84309
12778 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12779 of exps and logs in the use_exp2 case.
12781 2018-02-13 Jeff Law <law@redhat.com>
12783 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12784 entry for "vector".
12786 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12789 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12793 * common.opt (gas-loc-support, gas-locview-support): New.
12794 (ginline-points, ginternal-reset-location-views): New.
12795 * doc/invoke.texi: Document them. Use @itemx where intended.
12796 (gvariable-location-views): Adjust.
12797 * target.def (reset_location_view): New.
12798 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12799 (TARGET_RESET_LOCATION_VIEW): New.
12800 * doc/tm.texi: Rebuilt.
12801 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12802 (dwarf2out_default_as_locview_support): New.
12803 (output_asm_line_debug_info): Use option variables.
12804 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12805 (output_loc_list): Likewise.
12806 (add_high_low_attributes): Check option variables.
12807 Don't output entry view attribute in strict mode.
12808 (gen_inlined_subroutine_die): Check option variables.
12809 (dwarf2out_inline_entry): Likewise.
12810 (init_sections_and_labels): Likewise.
12811 (dwarf2out_early_finish): Likewise.
12812 (maybe_reset_location_view): New, from...
12813 (dwarf2out_var_location): ... here. Call it.
12814 * debug.h (dwarf2out_default_as_loc_support): Declare.
12815 (dwarf2out_default_as_locview_support): Declare.
12816 * hooks.c (hook_int_rtx_insn_0): New.
12817 * hooks.h (hook_int_rtx_insn_0): Declare.
12818 * toplev.c (process_options): Take -gas-loc-support and
12819 -gas-locview-support from dwarf2out. Enable
12820 -gvariable-location-views by default only with locview
12821 assembler support. Enable -ginternal-reset-location-views by
12822 default only if the target defines the corresponding hook.
12823 Enable -ginline-points by default if location views are
12824 enabled; force it disabled if statement frontiers are
12826 * tree-inline.c (expand_call_inline): Check option variables.
12827 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12829 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12831 PR tree-optimization/84321
12832 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12833 handling. Also check whether the anti-range contains any values
12834 that satisfy the mask; switch to a VR_RANGE if not.
12836 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12839 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12841 2018-02-13 Martin Jambor <mjambor@suse.cz>
12844 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12845 of call statements, also set location of a load to a temporary.
12847 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12849 * config/rl78/rl78.c (add_vector_labels): New function.
12850 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12851 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12852 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12853 which checks that no arguments are passed.
12854 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12855 * doc/extend.texi: Documentation for the new attribute.
12857 2018-02-13 Andreas Schwab <schwab@suse.de>
12859 * config/riscv/linux.h (CPP_SPEC): Define.
12861 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12864 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12865 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12866 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12867 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12868 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12869 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12870 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12871 temporarily for AES and PCLMUL builtins.
12873 PR tree-optimization/84339
12874 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12875 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12878 PR middle-end/84309
12879 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12880 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12881 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12883 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12885 * omp-simd-clone.h: New file.
12886 * omp-simd-clone.c: Include omp-simd-clone.h.
12887 (expand_simd_clones): No longer static.
12888 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12889 cgraph.h and omp-simd-clone.h.
12890 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12891 (vect_recog_widen_shift_pattern): Formatting fix.
12892 (vect_pattern_recog_1): Don't check optab for calls.
12895 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12896 operands[2] into a REG before using gen_lowpart on it.
12898 2018-02-12 Jeff Law <law@redhat.com>
12901 * config/sh/sh.c (find_barrier): Consider a sibling call
12904 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12905 successfully back substituting a reg.
12907 2018-02-12 Richard Biener <rguenther@suse.de>
12909 PR tree-optimization/84037
12910 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12911 parameter, move visited init to caller.
12912 (vect_slp_analyze_operations): Separate cost from validity
12913 check, initialize visited once for all instances.
12914 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12916 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12918 (vect_model_promotion_demotion_cost): Likewise.
12919 (vectorizable_bswap): Guard cost modeling with !slp_node
12920 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12922 (vectorizable_call): Likewise.
12923 (vectorizable_conversion): Likewise.
12924 (vectorizable_assignment): Likewise.
12925 (vectorizable_shift): Likewise.
12926 (vectorizable_operation): Likewise.
12927 (vectorizable_store): Likewise.
12928 (vectorizable_load): Likewise.
12929 (vectorizable_condition): Likewise.
12930 (vectorizable_comparison): Likewise.
12932 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12935 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12936 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12937 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12939 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12941 PR middle-end/83665
12942 * params.def (inline-min-speedup): Increase from 8 to 15.
12943 (max-inline-insns-auto): Decrease from 40 to 30.
12944 * ipa-split.c (consider_split): Add some buffer for function to
12945 be considered inlining candidate.
12946 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12949 2018-02-12 Richard Biener <rguenther@suse.de>
12951 PR tree-optimization/84037
12952 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12953 matched stmts if we cannot swap the non-matched ones.
12955 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12957 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12958 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12959 _mm_maskz_scalef_round_ss): New intrinsics.
12960 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12961 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12962 __builtin_ia32_scalefss_round): Remove.
12963 (__builtin_ia32_scalefsd_mask_round,
12964 __builtin_ia32_scalefss_mask_round): New intrinsics.
12965 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12966 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12967 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12968 "<round_constraint>")): Changed to ...
12969 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12970 "<round_scalar_constraint>")): ... this.
12971 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12972 %0, %1, %2<round_op3>}"): Changed to ...
12973 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12974 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12975 %2<round_scalar_mask_op3>}"): ... this.
12976 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12978 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12980 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12981 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12982 (_mm_maskz_sqrt_round_ss): New intrinsics.
12983 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12984 (__builtin_ia32_sqrtsd_mask_round)
12985 (__builtin_ia32_sqrtss_mask_round): New builtins.
12986 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12987 (__builtin_ia32_sqrtss_round): Remove.
12988 (__builtin_ia32_sqrtsd_mask_round)
12989 (__builtin_ia32_sqrtss_mask_round): New builtins.
12990 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12991 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12992 ((match_operand:VF_128 1 "vector_operand"
12993 "xBm,<round_constraint>")): Changed to ...
12994 ((match_operand:VF_128 1 "vector_operand"
12995 "xBm,<round_scalar_constraint>")): ... this.
12996 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12997 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12998 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12999 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13000 %<iptr>1<round_scalar_mask_op3>}): ... this.
13001 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13002 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13004 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13007 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13008 Cast vec_cmpeq result to correct type.
13009 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13010 Cast vec_cmpgt result to correct type.
13012 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13014 * final.c (final_scan_insn_1): Renamed from...
13015 (final_scan_insn): ... this. New wrapper, to recover
13016 seen from the outermost call in recursive ones.
13017 * config/sparc/sparc.c (output_return): Drop seen from call.
13018 (output_sibcall): Likewise.
13019 * config/visium/visium.c (output_branch): Likewise.
13021 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13023 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13026 2018-02-10 Alan Modra <amodra@gmail.com>
13029 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13030 Specify LR as an input.
13032 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13035 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13036 remove_member_access_dummy_vars): New functions.
13037 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13038 lower_omp_1, execute_lower_omp): Use them.
13040 PR rtl-optimization/84308
13041 * shrink-wrap.c (spread_components): Release todo vector.
13043 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13045 PR rtl-optimization/57193
13046 * ira-color.c (struct allocno_color_data): Add member
13047 conflict_allocno_hard_prefs.
13048 (update_conflict_allocno_hard_prefs): New.
13049 (bucket_allocno_compare_func): Add a preference based on
13050 conflict_allocno_hard_prefs.
13051 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13052 (color_allocnos): Remove a dead code. Initiate
13053 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13055 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13058 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13059 constraint from =wa to wa. Avoid a subreg on the output operand,
13060 instead use a pseudo and subreg it in a move.
13061 (p9_xxbrd_<mode>): Changed to ...
13062 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13063 (p9_xxbrd_v2df): New expander.
13064 (p9_xxbrw_<mode>): Changed to ...
13065 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13066 (p9_xxbrw_v4sf): New expander.
13068 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13070 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13072 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13075 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13076 multiply in 32-bit mode.
13077 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13078 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13081 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13083 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13084 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13085 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13086 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13088 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13091 * dwarf2out.c (is_trivial_indirect_ref): New function.
13092 (dwarf2out_late_global_decl): Do not generate a location
13093 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13094 and that are not defined in the current unit.
13096 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13098 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13099 instead of a libcall for UNORDERED.
13101 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13104 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13105 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13107 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13110 * config/s390/s390.c (s390_set_current_function): Invoke
13111 s390_indirect_branch_settings also if fndecl didn't change.
13113 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13115 * config/rs6000/rs6000.md (blockage): Set length to zero.
13117 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13119 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13121 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13124 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13125 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13129 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13130 PARALLEL incoming that failed vt_get_decl_and_offset check.
13132 PR middle-end/84237
13133 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13134 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13136 (get_variable_section): For decls in named .bss* sections pass true as
13137 second argument to bss_initializer_p.
13139 2018-02-09 Marek Polacek <polacek@redhat.com>
13140 Jakub Jelinek <jakub@redhat.com>
13143 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13144 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13145 Sync some changes from cxx_fold_indirect_ref.
13147 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13149 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13151 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13152 (BLOCK_INLINE_ENTRY_LABEL): New.
13153 (dwarf2out_var_location): Disregard inline entry markers.
13154 (inline_entry_data): New struct.
13155 (inline_entry_data_hasher): New hashtable type.
13156 (inline_entry_data_hasher::hash): New.
13157 (inline_entry_data_hasher::equal): New.
13158 (inline_entry_data_table): New variable.
13159 (add_high_low_attributes): Add DW_AT_entry_pc and
13160 DW_AT_GNU_entry_view attributes if a pending entry is found
13161 in inline_entry_data_table. Add old entry_pc attribute only
13162 if debug nonbinding markers are disabled.
13163 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13164 markers are enabled.
13165 (block_within_block_p, dwarf2out_inline_entry): New.
13166 (dwarf2out_finish): Check that no entries remained in
13167 inline_entry_data_table.
13168 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13169 (final_scan_insn, notice_source_line): Likewise.
13170 (rest_of_clean_state): Skip inline entry markers.
13171 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13173 * gimple.c (gimple_build_debug_inline_entry): New.
13174 * gimple.h (enum gimple_debug_subcode): Add
13175 GIMPLE_DEBUG_INLINE_ENTRY.
13176 (gimple_build_debug_inline_entry): Declare.
13177 (gimple_debug_inline_entry_p): New.
13178 (gimple_debug_nonbind_marker_p): Adjust.
13179 * insn-notes.def (INLINE_ENTRY): New.
13180 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13181 inline entry marker notes.
13182 (print_insn): Likewise.
13183 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13184 (INSN_DEBUG_MARKER_KIND): Likewise.
13185 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13186 * tree-inline.c (expand_call_inline): Build and insert
13187 debug_inline_entry stmt.
13188 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13189 inline entry blocks early, if nonbind markers are enabled.
13190 (dump_scope_block): Dump fragment info.
13191 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13192 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13193 (gimple_build_debug_inline_entry): New.
13194 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13195 Enable/disable inline entry points too.
13196 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13197 (DEBUG_INSN): Describe inline entry markers.
13199 * common.opt (gvariable-location-views): New.
13200 (gvariable-location-views=incompat5): New.
13201 * config.in: Rebuilt.
13202 * configure: Rebuilt.
13203 * configure.ac: Test assembler for view support.
13204 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13205 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13206 * dwarf2out.c (var_loc_view): New typedef.
13207 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13208 (dwarf2out_locviews_in_attribute): New.
13209 (dwarf2out_locviews_in_loclist): New.
13210 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13211 (enum dw_line_info_opcode): Add LI_adv_address.
13212 (struct dw_line_info_table): Add view.
13213 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13214 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13215 (zero_view_p): New variable.
13216 (ZERO_VIEW_P): New macro.
13217 (output_asm_line_debug_info): New.
13218 (struct var_loc_node): Add view.
13219 (add_AT_view_list, AT_loc_list): New.
13220 (add_var_loc_to_decl): Add view param. Test it against last.
13221 (new_loc_list): Add view params. Record them.
13222 (AT_loc_list_ptr): Handle loc and view lists.
13223 (view_list_to_loc_list_val_node): New.
13224 (print_dw_val): Handle dw_val_class_view_list.
13225 (size_of_die): Likewise.
13226 (value_format): Likewise.
13227 (loc_list_has_views): New.
13228 (gen_llsym): Set vl_symbol too.
13229 (maybe_gen_llsym, skip_loc_list_entry): New.
13230 (dwarf2out_maybe_output_loclist_view_pair): New.
13231 (output_loc_list): Output view list or entries too.
13232 (output_view_list_offset): New.
13233 (output_die): Handle dw_val_class_view_list.
13234 (output_dwarf_version): New.
13235 (output_compilation_unit_header): Use it.
13236 (output_skeleton_debug_sections): Likewise.
13237 (output_rnglists, output_line_info): Likewise.
13238 (output_pubnames, output_aranges): Update version comments.
13239 (output_one_line_info_table): Output view numbers in asm comments.
13240 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13241 Call maybe_gen_llsym.
13242 (loc_list_from_tree_1): Adjust.
13243 (add_AT_location_description): Create view list attribute if
13244 needed, check it's absent otherwise.
13245 (convert_cfa_to_fb_loc_list): Adjust.
13246 (maybe_emit_file): Call output_asm_line_debug_info for test.
13247 (dwarf2out_var_location): Reset views as needed. Precompute
13248 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13249 attribute. Set view.
13250 (new_line_info_table): Reset next view.
13251 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13252 (dwarf2out_source_line): Likewise. Output view resets and labels to
13253 the assembler, or select appropriate line info opcodes.
13254 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13255 (optimize_string_length): Catch it. Adjust.
13256 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13257 dw_val_class_view_list, and remove it if no longer needed.
13258 (hash_loc_list): Hash view numbers.
13259 (loc_list_hasher::equal): Compare them.
13260 (optimize_location_lists): Check whether a view list symbol is
13261 needed, and whether the locview attribute is present, and
13262 whether they match. Remove the locview attribute if no longer
13264 (index_location_lists): Call skip_loc_list_entry for test.
13265 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13266 Use output_dwarf_version.
13267 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13268 (struct dw_val_node): Add val_view_list.
13269 * final.c (SEEN_NEXT_VIEW): New.
13270 (set_next_view_needed): New.
13271 (clear_next_view_needed): New.
13272 (maybe_output_next_view): New.
13273 (final_start_function): Rename to...
13274 (final_start_function_1): ... this. Take pointer to FIRST,
13275 add SEEN parameter. Emit param bindings in the initial view.
13276 (final_start_function): Reintroduce SEEN-less interface.
13277 (final): Rename to...
13278 (final_1): ... this. Take SEEN parameter. Output final pending
13279 next view at the end.
13280 (final): Reintroduce seen-less interface.
13281 (final_scan_insn): Output pending next view before switching
13282 sections or ending a block. Mark the next view as needed when
13283 outputting variable locations. Notify debug backend of section
13284 changes, and of location view changes.
13285 (rest_of_handle_final): Adjust.
13286 * toplev.c (process_options): Autodetect value for debug variable
13287 location views option. Warn on incompat5 without -gdwarf-5.
13288 * doc/invoke.texi (gvariable-location-views): New.
13289 (gvariable-location-views=incompat5): New.
13290 (gno-variable-location-views): New.
13292 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13294 PR tree-optimization/84136
13295 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13296 that the result of find_edge is non-NULL.
13298 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13301 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13302 storing integer register in SImode. Fix cost of 256 and 512
13303 byte aligned SSE register store.
13305 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13307 * config/i386/i386.c (ix86_multiplication_cost): Fix
13308 multiplication cost for TARGET_AVX512DQ.
13310 2018-02-08 Marek Polacek <polacek@redhat.com>
13312 PR tree-optimization/84238
13313 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13316 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13318 PR tree-optimization/84265
13319 * tree-vect-stmts.c (vectorizable_store): Don't treat
13320 VMAT_CONTIGUOUS accesses as grouped.
13321 (vectorizable_load): Likewise.
13323 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13325 PR tree-optimization/81635
13326 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13327 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13328 (test_round_for_mask): New functions.
13329 (wide_int_cc_tests): Call test_round_for_mask.
13330 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13331 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13332 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13333 range returned by get_range_info.
13335 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13338 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13339 * symtab.c: Include builtins.h
13340 (symtab_node::output_to_lto_symbol_table_p): Move here
13341 from lto-streamer-out.c:output_symbol_p.
13342 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13343 (output_symbol_p): Move all logic to symtab.c
13344 (produce_symtab): Update.
13346 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13348 * config/s390/s390-opts.h (enum indirect_branch): Define.
13349 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13350 (s390_indirect_branch_via_thunk)
13351 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13352 (enum s390_indirect_branch_type): Define.
13353 * config/s390/s390.c (struct s390_frame_layout, struct
13354 machine_function): Remove.
13355 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13356 (indirect_branch_table_label_no, indirect_branch_table_name):
13358 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13359 (enum s390_indirect_branch_option): Define.
13360 (s390_return_addr_from_memory): New function.
13361 (s390_handle_string_attribute): New function.
13362 (s390_attribute_table): Add new attribute handler.
13363 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13364 (s390_indirect_branch_via_thunk): New function.
13365 (s390_indirect_branch_via_inline_thunk): New function.
13366 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13367 sibling call optimization for non z10 compiles.
13368 (s390_emit_call): Force indirect branch target to be a single
13369 register. Add r1 clobber for non-z10 compiles.
13370 (s390_emit_epilogue): Emit return jump via return_use expander.
13371 (s390_reorg): Handle JUMP_INSNs as execute targets.
13372 (s390_option_override_internal): Perform validity checks for the
13373 new command line options.
13374 (s390_indirect_branch_attrvalue): New function.
13375 (s390_indirect_branch_settings): New function.
13376 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13377 (s390_output_indirect_thunk_function): New function.
13378 (s390_code_end): Implement target hook.
13379 (s390_case_values_threshold): Implement target hook.
13380 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13382 * config/s390/s390.h (struct s390_frame_layout)
13383 (struct machine_function): Move here from s390.c.
13384 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13385 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13386 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13387 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13388 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13389 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13390 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13391 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13392 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13393 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13394 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13395 (mnemonic attribute): Add values which aren't recognized
13397 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13398 pattern for branch conversion. Fix mnemonic attribute.
13399 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13400 indirect branch via thunk if requested.
13401 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13402 ("*indirect_jump"): Disable for branch conversion using out of
13404 ("indirect_jump_via_thunk<mode>_z10")
13405 ("indirect_jump_via_thunk<mode>")
13406 ("indirect_jump_via_inlinethunk<mode>_z10")
13407 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13408 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13409 ("casesi_jump_via_inlinethunk<mode>_z10")
13410 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13411 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13412 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13413 ("*indirect2_jump"): Disable for branch conversion.
13414 ("casesi_jump"): Turn into expander and expand patterns for branch
13416 ("return_use"): New expander.
13417 ("*return"): Emit return via thunk and rename it to ...
13418 ("*return<mode>"): ... this one.
13419 * config/s390/s390.opt: Add new options and and enum for the
13422 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13424 * lra-constraints.c (match_reload): Unconditionally use
13425 gen_lowpart_SUBREG, rather than selecting between that
13426 and equivalent gen_rtx_SUBREG code.
13428 2018-02-08 Richard Biener <rguenther@suse.de>
13430 PR tree-optimization/84233
13431 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13432 changed flag instead of boguously re-using phi_inserted.
13434 2018-02-08 Martin Jambor <mjambor@suse.cz>
13436 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13437 static local variables.
13439 2018-02-08 Richard Biener <rguenther@suse.de>
13441 PR tree-optimization/84278
13442 * tree-vect-stmts.c (vectorizable_store): When looking for
13443 smaller vector types to perform grouped strided loads/stores
13444 make sure the mode is supported by the target.
13445 (vectorizable_load): Likewise.
13447 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13449 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13450 Increase LDP/STP opportunities by adding adjacent callee-saves.
13452 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13454 PR rtl-optimization/84068
13455 PR rtl-optimization/83459
13456 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13458 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13460 PR tree-optimization/84224
13461 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13462 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13463 non-zero arguments.
13465 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13468 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13469 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13470 count, remove one USE.
13472 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13474 * doc/install.texi (Configuration): Document the
13475 --with-long-double-format={ibm,ieee} PowerPC configuration
13479 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13480 Convert from define_expand to be define_insn_and_split. Rework
13481 float/double/_Float128 conversions to QI/HI/SImode to work with
13482 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13483 conversions to QI/HImode types did a store and then a load to
13484 truncate the value. For conversions to VSX registers, don't split
13485 the insn, instead emit the code directly. Use the code iterator
13486 any_fix to combine signed and unsigned conversions.
13487 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13488 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13489 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13490 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13491 (fix_<mode>di2_hw): Likewise.
13492 (fixuns_<mode>di2_hw): Likewise.
13493 (fix_<mode>si2_hw): Likewise.
13494 (fixuns_<mode>si2_hw): Likewise.
13495 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13496 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13497 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13498 fix<uns>_trunc<SFDF:mode>si2_p8.
13499 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13500 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13501 (fix<uns>_<mode>_mem): Likewise.
13502 (fctiw<u>z_<mode>_mem): Likewise.
13503 (fix<uns>_<mode>_mem): Likewise.
13504 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13505 the register allocator from doing a direct move to the GPRs to do
13506 a store, and instead use the ISA 3.0 store byte/half-word from
13507 vector register instruction. For IEEE 128-bit floating point,
13508 also optimize stores of 32-bit ints.
13509 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13511 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13513 * genextract.c (push_pathstr_operand): New function to support
13515 (walk_rtx): Call push_pathstr_operand.
13516 (print_path): Support [a-zA-Z].
13518 2018-02-07 Richard Biener <rguenther@suse.de>
13520 PR tree-optimization/84037
13521 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13522 (cse_and_gimplify_to_preheader): Declare.
13523 (vect_get_place_in_interleaving_chain): Likewise.
13524 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13526 (_loop_vec_info::~_loop_vec_info): Delete it.
13527 (cse_and_gimplify_to_preheader): New function.
13528 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13529 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13530 (vectorizable_load): Likewise. For grouped stores always base
13531 the IV on the first element.
13532 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13533 condition before gimplifying.
13535 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13537 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13538 *DIV_EXPR and *MOD_EXPR.
13540 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13543 * config/i386/i386.c (ix86_option_override_internal): Mask out
13544 the CF_SET bit when checking -fcf-protection.
13546 2018-02-07 Tom de Vries <tom@codesourcery.com>
13549 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13552 2018-02-07 Richard Biener <rguenther@suse.de>
13554 PR tree-optimization/84204
13555 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13558 PR tree-optimization/84205
13559 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13560 special-case isl_ast_op_zdiv_r.
13562 PR tree-optimization/84223
13563 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13564 Only add conditions from within the region.
13565 (gather_bbs::after_dom_children): Adjust.
13567 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13570 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13571 * config/avr/avr.md: Only post-reload split REG-REG moves if
13572 either register is GENERAL_REG_P.
13574 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13576 PR tree-optimization/84235
13577 * tree-ssa-scopedtables.c
13578 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13579 if the subtraction is performed in floating point type where NaNs are
13580 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13581 build 1. Formatting fix.
13583 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13586 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13587 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13588 and skip it regardless of bb boundaries. Use CALL_P macro,
13589 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13590 unnecessarily, formatting fix.
13592 2018-02-06 Michael Collison <michael.collison@arm.com>
13594 * config/arm/thumb2.md:
13595 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13596 (*thumb_mov_notscc): Ditto.
13598 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13601 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13602 unsigned_fix, not "s".
13604 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13606 * configure.ac (gcc_fn_eh_frame_ro): New function.
13607 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13608 correct .eh_frame permissions.
13609 * configure: Regenerate.
13611 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13613 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13614 irrelevant options.
13616 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13618 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13619 Display warning message for -mno-speculate-indirect-jumps.
13621 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13623 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13625 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13627 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13629 PR tree-optimization/84225
13630 * tree-eh.c (find_trapping_overflow): Only call
13631 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13633 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13636 * config/i386/i386.c: Reimplement the check of possible options
13637 -mibt/-mshstk conbination. Change error messages.
13638 * doc/invoke.texi: Fix a typo: remove extra '='.
13640 2018-02-06 Marek Polacek <polacek@redhat.com>
13642 PR tree-optimization/84228
13643 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13645 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13648 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13649 emitted arch directives.
13650 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13651 __ARM_FEATURE_COPROC before changing architectures.
13653 2018-02-06 Richard Biener <rguenther@suse.de>
13655 * config/i386/i386.c (print_reg): Fix typo.
13656 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13658 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13660 * configure: Regenerate.
13662 2018-02-05 Martin Sebor <msebor@redhat.com>
13664 PR tree-optimization/83369
13665 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13668 2018-02-05 Martin Liska <mliska@suse.cz>
13670 * doc/invoke.texi: Cherry-pick upstream r323995.
13672 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13674 * ira.c (ira_init_register_move_cost): Adjust comment.
13676 2018-02-05 Martin Liska <mliska@suse.cz>
13678 PR gcov-profile/84137
13679 * doc/gcov.texi: Fix typo in documentation.
13681 2018-02-05 Martin Liska <mliska@suse.cz>
13683 PR gcov-profile/83879
13684 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13685 using dlopen functionality.
13687 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13689 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13690 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13691 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13692 _mm_maskz_range_round_ss): New intrinsics.
13693 (__builtin_ia32_rangesd128_round)
13694 (__builtin_ia32_rangess128_round): Remove.
13695 (__builtin_ia32_rangesd128_mask_round,
13696 __builtin_ia32_rangess128_mask_round): New builtins.
13697 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13698 __builtin_ia32_rangess128_round): Remove.
13699 (__builtin_ia32_rangesd128_mask_round,
13700 __builtin_ia32_rangess128_mask_round): New builtins.
13701 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13702 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13703 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13704 "<round_saeonly_constraint>")): Changed to ...
13705 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13706 "<round_saeonly_scalar_constraint>")): ... this.
13707 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13708 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13709 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13710 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13711 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13713 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13715 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13717 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13718 Remove all values except native, 8540 and 8548.
13720 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13722 * config/i386/i386.c (ix86_output_function_return): Pass
13723 INVALID_REGNUM, instead of -1, as invalid register number to
13724 indirect_thunk_name and output_indirect_thunk.
13726 2018-02-02 Julia Koval <julia.koval@intel.com>
13728 * config.gcc: Add -march=icelake.
13729 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13730 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13731 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13732 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13733 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13734 (processor_target_table): Add icelake.
13735 (ix86_option_override_internal): Handle new PTAs.
13736 (get_builtin_code_for_version): Handle icelake.
13737 (M_INTEL_COREI7_ICELAKE): New.
13738 (fold_builtin_cpu): Handle icelake.
13739 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13740 * doc/invoke.texi: Add -march=icelake.
13742 2018-02-02 Julia Koval <julia.koval@intel.com>
13744 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13745 to wide_int_bitmask.
13746 * wide-int-bitmask.h: New.
13748 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13751 * config/i386/i386.md: Replace Pmode with word_mode in
13752 builtin_setjmp_setup and builtin_longjmp to support x32.
13754 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13758 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13760 (rs6000_supported_cpu_names): New static variable.
13761 (linux_cpu_translation_table): Likewise.
13762 (elf_platform) <cpu>: Define new static variable and use it.
13763 Translate kernel AT_PLATFORM name to canonical name if needed.
13764 Error if platform name is unknown.
13766 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13769 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13771 2018-02-01 Jeff Law <law@redhat.com>
13774 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13775 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13776 the scratch if RELEASE_VIA_POP is false.
13777 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13778 If we have to save a temporary register, decrement SIZE appropriately.
13779 Pass new arguments to release_scratch_register_on_entry.
13780 (ix86_adjust_stack_and_probe): Likewise.
13781 (ix86_emit_probe_stack_range): Pass new arguments to
13782 release_scratch_register_on_entry.
13784 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13786 PR rtl-optimization/84157
13787 * combine.c (change_zero_ext): Use REG_P predicate in
13788 front of HARD_REGISTER_P predicate.
13790 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13792 * config/avr/avr.c (avr_option_override): Move disabling of
13793 -fdelete-null-pointer-checks to...
13794 * common/config/avr/avr-common.c (avr_option_optimization_table):
13797 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13799 PR tree-optimization/81635
13800 * tree-data-ref.c (split_constant_offset_1): For types that
13801 wrap on overflow, try to use range info to prove that wrapping
13804 2018-02-01 Renlin Li <renlin.li@arm.com>
13807 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13808 TAILCALL_ADDR_REGS.
13809 (aarch64_register_move_cost): Likewise.
13810 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13811 TAILCALL_ADDR_REGS.
13812 (REG_CLASS_NAMES): Likewise.
13813 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13814 TAILCALL_ADDR_REGS. Remove IP registers.
13815 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13817 2018-02-01 Richard Biener <rguenther@suse.de>
13819 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13820 for specifying RPO order and allow NULL for that.
13821 * domwalk.c (dom_walker::dom_walker): Likewise.
13822 (dom_walker::walk): Handle NULL RPO order.
13823 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13825 (rewrite_update_dom_walker): Likewise.
13826 (mark_def_dom_walker): Likewise.
13828 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13830 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13831 (aarch64_maybe_expand_sve_subreg_move): Declare.
13832 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13833 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13835 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13836 that are semantically a reverse operation.
13837 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13838 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13839 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13841 (aarch64_can_change_mode_class): For big-endian, forbid changes
13842 between two SVE modes if they have different element sizes.
13844 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13846 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13847 the TImode handling for big-endian targets.
13849 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13851 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13852 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13854 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13855 Remove BSWAP handing for big-endian targets and use the form of
13856 LD1RQ appropariate for the mode.
13858 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13860 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13861 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13862 duplicated element.
13864 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13867 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13868 check for operands that need to go through aarch64_sve_reload_be.
13870 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13872 PR tree-optimization/81661
13873 PR tree-optimization/84117
13874 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13875 * tree-eh.c: Include gimplify.h.
13876 (find_trapping_overflow, replace_trapping_overflow,
13877 rewrite_to_non_trapping_overflow): New functions.
13878 * tree-vect-loop.c: Include tree-eh.h.
13879 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13880 * tree-data-ref.c: Include tree-eh.h.
13881 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13883 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13885 PR rtl-optimization/84123
13886 * combine.c (change_zero_ext): Check if hard register satisfies
13887 can_change_dest_mode before calling gen_lowpart_SUBREG.
13889 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13892 * ira.c (ira_init_register_move_cost): Remove assert.
13894 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13896 PR rtl-optimization/84071
13897 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13898 * doc/tm.texi: Regenerate.
13900 2018-01-31 Richard Biener <rguenther@suse.de>
13902 PR tree-optimization/84132
13903 * tree-data-ref.c (analyze_miv_subscript): Properly
13904 check whether evolution_function_is_affine_multivariate_p
13905 before calling gcd_of_steps_may_divide_p.
13907 2018-01-31 Julia Koval <julia.koval@intel.com>
13910 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13911 * config/i386/i386.md (rdpid_rex64) New.
13912 (rdpid): Make 32bit only.
13914 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13917 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13918 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13920 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13923 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13925 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13927 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13929 PR rtl-optimization/84071
13930 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13931 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13933 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13935 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13936 (arc_attribute_table): Add 'aux' attribute.
13937 (arc_in_small_data_p): Consider aux like variables.
13938 (arc_is_aux_reg_p): New function.
13939 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13940 (arc_get_aux_arg): New function.
13941 (prepare_move_operands): Handle aux-register access.
13942 (arc_handle_aux_attribute): New function.
13943 * doc/extend.texi (ARC Variable attributes): Add subsection.
13945 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13947 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13948 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13949 (arc_attribute_table): Add 'uncached' attribute.
13950 (arc_print_operand): Print '.di' flag for uncached memory
13952 (arc_in_small_data_p): Do not consider for small data the uncached
13954 (arc_is_uncached_mem_p): New function.
13955 * config/arc/predicates.md (compact_store_memory_operand): Check
13956 for uncached memory accesses.
13957 (nonvol_nonimm_operand): Likewise.
13958 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13960 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13963 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13964 falign-loops=): Add Optimization flag.
13966 2018-01-30 Jeff Law <law@redhat.com>
13969 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13970 INT_REGISTERS_SAVED. Check it prior to calling
13971 get_scratch_register_on_entry.
13972 (ix86_adjust_stack_and_probe): Similarly.
13973 (ix86_emit_probe_stack_range): Similarly.
13974 (ix86_expand_prologue): Corresponding changes.
13976 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13979 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13980 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13982 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13985 * lra-constraints.c (curr_insn_transform): Process AND in the
13988 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13990 PR rtl-optimization/83986
13991 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13992 dependence against last_pending_memory_flush in addition to
13993 pending_jump_insns.
13995 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13997 PR tree-optimization/81611
13998 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14001 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14004 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14007 2018-01-30 Richard Biener <rguenther@suse.de>
14008 Jakub Jelinek <jakub@redhat.com>
14010 PR tree-optimization/84111
14011 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14012 inner loops added during recursion, as they don't have up-to-date
14015 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14018 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14019 (can_inline_edge_by_limits_p): ... here.
14020 (can_early_inline_edge_p, check_callers,
14021 update_caller_keys, update_callee_keys, recursive_inlining,
14022 add_new_edges_to_heap, speculation_useful_p,
14023 inline_small_functions,
14024 inline_small_functions, flatten_function,
14025 inline_to_all_callers_1): Update.
14027 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14029 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14032 2018-01-30 Richard Biener <rguenther@suse.de>
14034 PR tree-optimization/83008
14035 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14036 invariant and constant vector uses in stmts when they need
14037 more than one stmt.
14039 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14042 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14043 * configure: Regenerate.
14045 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14047 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14049 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14050 Use gen_rtx_REG rather than gen_lowpart.
14052 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14054 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14055 rather than 0 when creating partial subregs.
14057 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14059 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14062 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14065 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14066 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14067 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14068 flags. This restores the settings used before the 2017-07-24.
14069 Turning off pre increment/decrement/modify allows IVOPTS to
14070 optimize DF/SF loops where the index is an int.
14072 2018-01-29 Richard Biener <rguenther@suse.de>
14073 Kelvin Nilsen <kelvin@gcc.gnu.org>
14076 * tree-vect-stmts.c (vectorizable_call): Don't call
14077 targetm.vectorize_builtin_md_vectorized_function if callee is
14080 2018-01-22 Carl Love <cel@us.ibm.com>
14082 * doc/extend.tex: Fix typo in second arg in
14083 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14085 2018-01-29 Richard Biener <rguenther@suse.de>
14087 PR tree-optimization/84086
14088 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14089 (flush_ssaname_freelist): When SSA names were released reset
14090 the SCEV hash table.
14092 2018-01-29 Richard Biener <rguenther@suse.de>
14094 PR tree-optimization/84057
14095 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14096 removed paths when removing edges.
14098 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14100 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14101 -mfunction-return=@var{choice}.
14103 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14105 PR diagnostic/84034
14106 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14107 Handle CR like TAB.
14108 (layout::print_source_line): Likewise.
14109 (test_get_line_width_without_trailing_whitespace): Add test cases.
14111 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14113 PR middle-end/84040
14114 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14117 2018-01-26 Jim Wilson <jimw@sifive.com>
14119 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14121 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14124 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14126 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14127 and CMP + SUB-immediate -> SUBS.
14129 2018-01-26 Martin Sebor <msebor@redhat.com>
14131 PR tree-optimization/83896
14132 * tree-ssa-strlen.c (get_string_len): Rename...
14133 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14134 Avoid assuming length is constant.
14135 (handle_char_store): Use HOST_WIDE_INT for string length.
14137 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14140 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14141 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14143 2018-01-26 Richard Biener <rguenther@suse.de>
14145 PR rtl-optimization/84003
14146 * dse.c (record_store): Only record redundant stores when
14147 the earlier store aliases at least all accesses the later one does.
14149 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14151 PR rtl-optimization/83985
14152 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14153 REG_CFA_RESTORE insns.
14154 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14155 REG_CFA_RESTORE insns here.
14158 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14159 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14161 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14163 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14164 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14165 (arc_init): Likewise.
14166 (arc_override_options): Likewise.
14167 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14169 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14171 * config/arc/arc.h (TARGET_DBNZ): Define.
14172 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14173 properly set the tune attribute.
14174 (dbnz): Use TARGET_DBNZ guard.
14175 * config/arc/arc.opt (mtune): Add core3 option.
14177 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14179 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14180 recognize new pic like addresses.
14181 (arc_delegitimize_address): Clean up.
14183 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14185 * config/arc/arc-arches.def: Option mrf16 valid for all
14187 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14188 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14189 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14190 * config/arc/arc-tables.opt: Regenerate.
14191 * config/arc/arc.c (arc_conditional_register_usage): Handle
14192 reduced register file case.
14193 (arc_file_start): Set must have build attributes.
14194 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14195 mrf16 option value.
14196 * config/arc/arc.opt (mrf16): Add new option.
14197 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14198 * config/arc/genmultilib.awk: Handle new mrf16 option.
14199 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14200 * config/arc/t-multilib: Regenerate.
14201 * doc/invoke.texi (ARC Options): Document mrf16 option.
14203 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14205 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14206 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14207 (arc_attribute_table): Add 'secure_call' attribute.
14208 (arc_print_operand): Print secure call operand.
14209 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14211 (arc_is_secure_call_p): New function. * config/arc/arc.md
14212 (call_i): Add support for sjli instruction.
14213 (call_value_i): Likewise.
14214 * config/arc/constraints.md (Csc): New constraint.
14216 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14217 John Eric Martin <John.Martin@emmicro-us.com>
14219 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14220 * config/arc/arc.c (_arc_jli_section): New struct.
14221 (arc_jli_section): New type.
14222 (rc_jli_sections): New static variable.
14223 (arc_handle_jli_attribute): New function.
14224 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14225 (arc_file_end): New function.
14226 (TARGET_ASM_FILE_END): Define.
14227 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14228 (arc_add_jli_section): New function.
14229 (jli_call_scan): Likewise.
14230 (arc_reorg): Call jli_call_scan.
14231 (arc_output_addsi): Remove 'S' from printing asm operand.
14232 (arc_is_jli_call_p): New function.
14233 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14235 (movhi_insn): Likewise.
14236 (movsi_insn): Likewise.
14237 (movsi_set_cc_insn): Likewise.
14238 (loadqi_update): Likewise.
14239 (load_zeroextendqisi_update): Likewise.
14240 (load_signextendqisi_update): Likewise.
14241 (loadhi_update): Likewise.
14242 (load_zeroextendhisi_update): Likewise.
14243 (load_signextendhisi_update): Likewise.
14244 (loadsi_update): Likewise.
14245 (loadsf_update): Likewise.
14246 (movsicc_insn): Likewise.
14247 (bset_insn): Likewise.
14248 (bxor_insn): Likewise.
14249 (bclr_insn): Likewise.
14250 (bmsk_insn): Likewise.
14251 (bicsi3_insn): Likewise.
14252 (cmpsi_cc_c_insn): Likewise.
14253 (movsi_ne): Likewise.
14254 (movsi_cond_exec): Likewise.
14255 (clrsbsi2): Likewise.
14256 (norm_f): Likewise.
14266 (call_i): Remove 'S' asm letter, add jli instruction.
14267 (call_value_i): Likewise.
14268 * config/arc/arc.op (mjli-always): New option.
14269 * config/arc/constraints.md (Cji): New constraint.
14270 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14272 (subsf3_fpx): Likewise.
14273 (mulsf3_fpx): Likewise.
14274 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14276 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14277 function attrbutes.
14278 * doc/invoke.texi (ARC): Document mjli-always option.
14280 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14282 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14283 avoid addition with 0 and use incw and decw where possible.
14285 2018-01-26 Richard Biener <rguenther@suse.de>
14287 PR tree-optimization/81082
14288 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14289 association if it requires casting to unsigned.
14290 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14291 from fold_plusminus_mult_expr to catch important cases late when
14292 range info is available.
14294 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14296 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14297 * configure.ac (hidden_linkonce): New test.
14298 * configure: Regenerate.
14299 * config.in: Regenerate.
14301 2018-01-26 Julia Koval <julia.koval@intel.com>
14303 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14304 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14305 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14306 _mm_mask_bitshuffle_epi64_mask): Fix type.
14307 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14308 USI_FTYPE_V4DI_V4DI_USI): Remove.
14309 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14310 __builtin_ia32_vpshufbitqmb256_mask,
14311 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14312 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14313 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14315 2018-01-26 Alan Modra <amodra@gmail.com>
14318 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14319 UNSPEC_VBPERMQ. Sort other unspecs.
14321 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14323 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14325 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14327 PR middle-end/83055
14328 * predict.c (drop_profile): Do not push/pop cfun; update also
14330 (handle_missing_profiles): Fix logic looking for zero profiles.
14332 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14334 PR middle-end/83977
14335 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14336 on functions with #pragma omp declare simd or functions with simd
14338 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14339 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14340 Remove trailing \n from warning_at calls.
14342 2018-01-25 Tom de Vries <tom@codesourcery.com>
14345 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14346 for neutered workers.
14348 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14351 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14352 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14354 2018-01-24 Jeff Law <law@redhat.com>
14357 * i386.c (get_probe_interval): Move to earlier point.
14358 (ix86_compute_frame_layout): If -fstack-clash-protection and
14359 the frame is larger than the probe interval, then use pushes
14360 to save registers rather than reg->mem moves.
14361 (ix86_expand_prologue): Remove conditional for int_registers_saved
14364 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14367 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14368 min/max for never referenced object.
14370 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14372 PR middle-end/83977
14373 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14375 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14376 attributes from DECL_ATTRIBUTES (decl) without affecting
14377 DECL_ATTRIBUTES (current_function_decl).
14378 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14379 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14381 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14383 PR tree-optimization/83979
14384 * fold-const.c (fold_comparison): Use constant_boolean_node
14385 instead of boolean_{true,false}_node.
14387 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14389 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14392 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14394 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14395 Simplify the clause that sets the length attribute.
14396 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14397 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14398 clause that sets the length attribute.
14399 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14401 2018-01-24 Tom de Vries <tom@codesourcery.com>
14404 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14405 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14406 Add strict parameter.
14407 (prevent_branch_around_nothing): Insert dummy insn between branch to
14408 label and label with no ptx insn inbetween.
14409 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14411 2018-01-24 Tom de Vries <tom@codesourcery.com>
14414 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14415 for neutered threads in warp.
14416 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14418 2018-01-24 Richard Biener <rguenther@suse.de>
14420 PR tree-optimization/83176
14421 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14424 2018-01-24 Richard Biener <rguenther@suse.de>
14426 PR tree-optimization/82819
14427 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14428 code generating pluses that are no-ops in the target precision.
14430 2018-01-24 Richard Biener <rguenther@suse.de>
14432 PR middle-end/84000
14433 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14435 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14437 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14438 to merge probabilities.
14439 * predict.c (probably_never_executed): Also mark as cold functions
14440 with global 0 profile and guessed local profile.
14441 * profile-count.c (profile_probability::combine_with_count): New
14443 * profile-count.h (profile_probability::operator*,
14444 profile_probability::operator*=, profile_probability::operator/,
14445 profile_probability::operator/=): Reduce precision to adjusted
14446 and set value to guessed on contradictory divisions.
14447 (profile_probability::combine_with_freq): Remove.
14448 (profile_probability::combine_wiht_count): Declare.
14449 (profile_count::force_nonzero):: Set to adjusted.
14450 (profile_count::probability_in):: Set quality to adjusted.
14451 * tree-ssa-tail-merge.c (replace_block_by): Use
14452 combine_with_count.
14454 2018-01-23 Andrew Waterman <andrew@sifive.com>
14455 Jim Wilson <jimw@sifive.com>
14457 * config/riscv/riscv.c (riscv_stack_boundary): New.
14458 (riscv_option_override): Set riscv_stack_boundary. Handle
14459 riscv_preferred_stack_boundary_arg.
14460 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14461 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14462 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14463 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14464 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14465 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14467 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14470 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14471 of struct ix86_frame.
14472 (ix86_expand_epilogue): Likewise. Add a local variable for
14473 the reg_save_offset field in struct ix86_frame.
14475 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14477 PR tree-optimization/82604
14478 * tree-loop-distribution.c (enum partition_kind): New enum item
14479 PKIND_PARTIAL_MEMSET.
14480 (partition_builtin_p): Support above new enum item.
14481 (generate_code_for_partition): Ditto.
14482 (compute_access_range): Differentiate cases that equality can be
14483 proven at all loops, the innermost loops or no loops.
14484 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14485 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14486 (finalize_partitions, distribute_loop): Don't fuse partition of
14487 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14488 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14489 parloop is enabled.
14491 2018-01-23 Martin Liska <mliska@suse.cz>
14493 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14494 order to ignore the predictor.
14495 (PRED_POLYMORPHIC_CALL): Likewise.
14496 (PRED_RECURSIVE_CALL): Likewise.
14498 2018-01-23 Martin Liska <mliska@suse.cz>
14500 * tree-profile.c (tree_profiling): Print function header to
14501 aware reader which function we are working on.
14502 * value-prof.c (gimple_find_values_to_profile): Do not print
14503 not interesting value histograms.
14505 2018-01-23 Martin Liska <mliska@suse.cz>
14507 * profile-count.h (enum profile_quality): Add
14508 profile_uninitialized as the first value. Do not number values
14509 as they are zero based.
14510 (profile_count::verify): Update sanity check.
14511 (profile_probability::verify): Likewise.
14513 2018-01-23 Nathan Sidwell <nathan@acm.org>
14515 * doc/invoke.texi (ffor-scope): Deprecate.
14517 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14519 PR tree-optimization/83510
14520 * domwalk.c (set_all_edges_as_executable): New function.
14521 (dom_walker::dom_walker): Convert bool param
14522 "skip_unreachable_blocks" to enum reachability. Move setup of
14523 edge flags to set_all_edges_as_executable and only do it when
14524 reachability is REACHABLE_BLOCKS.
14525 * domwalk.h (enum dom_walker::reachability): New enum.
14526 (dom_walker::dom_walker): Convert bool param
14527 "skip_unreachable_blocks" to enum reachability.
14528 (set_all_edges_as_executable): New decl.
14529 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14530 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14532 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14533 but converting true to REACHABLE_BLOCKS.
14534 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14536 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14537 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14538 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14540 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14541 if check_all_array_refs will be called.
14543 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14545 * tree.c (selftest::test_location_wrappers): Add more test
14548 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14550 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14551 (selftest::test_bit_in_range): Likewise.
14553 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14556 * doc/sourcebuild.texi (vect_float): Say that the selector
14557 only describes the situation when -funsafe-math-optimizations is on.
14558 (vect_float_strict): Document.
14560 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14562 PR tree-optimization/83965
14563 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14564 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14565 instead of checking only for a reduction.
14566 (vect_recog_widen_sum_pattern): Likewise.
14568 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14570 * predict.c (probably_never_executed): Only use precise profile info.
14571 (compute_function_frequency): Skip after inlining hack since we now
14572 have quality checking.
14574 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14576 * profile-count.h (profile_probability::very_unlikely,
14577 profile_probability::unlikely, profile_probability::even): Set
14578 precision to guessed.
14580 2018-01-23 Richard Biener <rguenther@suse.de>
14582 PR tree-optimization/83963
14583 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14584 Properly terminate dominator walk when crossing the exit edge not
14585 when visiting its source block.
14587 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14590 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14591 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14593 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14595 PR tree-optimization/83957
14596 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14597 semicolon after for body surrounded by braces.
14599 PR tree-optimization/83081
14600 * profile-count.h (profile_probability::split): New method.
14601 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14602 Use profile_probability::split.
14603 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14604 when splitting a single conditional jump into 2.
14606 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14608 PR tree-optimization/69452
14609 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14612 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14614 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14615 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14616 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14618 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14620 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14622 * config/rl78/rl78.md (movdi): New define_expand.
14623 * config/rl78/rl78.c (rl78_split_movdi): New function.
14625 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14628 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14630 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14631 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14632 128-bit to produce an UNSPEC move to get the double word with the
14633 signbit and then a shift directly to do signbit.
14634 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14635 implementation with a new version that just does either a direct
14636 move or a regular move. Move memory interface to separate insns.
14637 Move insns so they are next to the expander.
14638 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14639 with signbit move. Split big and little endian case.
14640 (signbit<mode>2_dm_mem_le): Likewise.
14641 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14642 (signbit<mode>2_dm2): Likewise.
14644 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14646 * config/rl78/rl78.md (anddi3): New define_expand.
14648 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14650 * config/rl78/rl78.md (umindi3): New define_expand.
14652 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14654 * config/rl78/rl78.md (smindi3): New define_expand.
14656 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14658 * config/rl78/rl78.md (smaxdi3): New define_expand.
14660 2018-01-22 Carl Love <cel@us.ibm.com>
14662 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14663 LVX_V1TI): Add macro expansion.
14664 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14665 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14666 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14667 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14668 Change check to determine if the instruction is a byte reversing
14669 entry. Fix typo in comment.
14670 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14671 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14672 Add def_builtin calls for new builtins.
14673 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14674 Add define_insn expansion.
14676 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14678 * config/rl78/rl78.md (umaxdi3): New define_expand.
14680 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14682 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14683 for non-QImode registers.
14685 2018-01-22 Richard Biener <rguenther@suse.de>
14687 PR tree-optimization/83963
14688 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14689 including the loop exit block.
14690 (scop_detection::merge_sese): Likewise.
14691 (scop_detection::add_scop): Do it here instead.
14693 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14695 * doc/sourcebuild.texi (arm_softfloat): Document.
14697 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14700 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14701 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14702 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14704 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14705 David Edelsohn <dje.gcc@gmail.com>
14708 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14709 Change "crset eq" to "crset 2".
14710 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14711 (*call_indirect_aix<mode>_nospec): Likewise.
14712 (*call_value_indirect_aix<mode>_nospec): Likewise.
14713 (*call_indirect_elfv2<mode>_nospec): Likewise.
14714 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14715 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14716 change assembly output from . to $.
14717 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14718 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14719 (*tablejump<mode>_internal1_nospec): Likewise.
14721 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14724 * config/sh/sh_optimize_sett_clrt.cc:
14725 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14727 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14729 PR tree-optimization/83940
14730 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14731 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14732 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14733 use it to pass back the definition type.
14734 (vect_check_store_rhs): Likewise rhs_dt_out.
14735 (vect_build_gather_load_calls): Add a mask_dt argument and use
14736 it instead of a call to vect_is_simple_use.
14737 (vectorizable_store): Update calls to vect_check_load_store_mask
14738 and vect_check_store_rhs. Use the dt returned by the latter instead
14739 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14740 instead of calls to vect_is_simple_use. Pass the scalar rather
14741 than the vector operand to vect_is_simple_use when handling
14742 second and subsequent copies of an rhs value.
14743 (vectorizable_load): Update calls to vect_check_load_store_mask
14744 and vect_build_gather_load_calls. Use the cached mask_dt and
14745 gs_info.offset_dt instead of calls to vect_is_simple_use.
14747 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14749 PR middle-end/83945
14750 * tree-emutls.c: Include gimplify.h.
14751 (lower_emutls_2): New function.
14752 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14753 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14754 it before further processing.
14757 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14758 UINTVAL (trueop1) instead of INTVAL (op1).
14760 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14764 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14765 INCOMING_FRAME_SP_OFFSET if not defined.
14766 (scan_trace): Add ENTRY argument. If true and
14767 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14768 emit a note to adjust the CFA offset.
14769 (create_cfi_notes): Adjust scan_trace callers.
14770 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14771 INCOMING_FRAME_SP_OFFSET in the CIE.
14772 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14773 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14775 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14776 * doc/tm.texi: Regenerated.
14778 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14780 PR rtl-optimization/83147
14781 * lra-constraints.c (remove_inheritance_pseudos): Use
14782 lra_substitute_pseudo_within_insn.
14784 2018-01-19 Tom de Vries <tom@codesourcery.com>
14785 Cesar Philippidis <cesar@codesourcery.com>
14788 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14790 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14793 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14794 spaces for function labels.
14796 2018-01-19 Martin Liska <mliska@suse.cz>
14798 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14799 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14800 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14801 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14802 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14803 (PRED_CONST_RETURN): Change from 69 to 65.
14804 (PRED_NULL_RETURN): Change from 91 to 71.
14805 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14806 (PRED_LOOP_GUARD): Change from 66 to 73.
14808 2018-01-19 Martin Liska <mliska@suse.cz>
14810 * predict.c (predict_insn_def): Add new assert.
14811 (struct branch_predictor): Change type to signed integer.
14812 (test_prediction_value_range): Amend test to cover
14813 PROB_UNINITIALIZED.
14814 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14815 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14816 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14817 (PRED_LOOP_IV_COMPARE): Likewise.
14818 * predict.h (PROB_UNINITIALIZED): Define new constant.
14820 2018-01-19 Martin Liska <mliska@suse.cz>
14822 * predict.c (dump_prediction): Add new format for
14823 analyze_brprob.py script which is enabled with -details
14825 * profile-count.h (precise_p): New function.
14827 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14829 PR tree-optimization/83922
14830 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14831 there are no statements that need masking.
14832 (vect_active_double_reduction_p): New function.
14833 (vect_analyze_loop_operations): Use it when handling phis that
14834 are not in the loop header.
14836 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14838 PR tree-optimization/83914
14839 * tree-vect-loop.c (vectorizable_induction): Don't convert
14840 init_expr or apply the peeling adjustment for inductions
14841 that are nested within the vectorized loop.
14843 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14845 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14848 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14852 * function.h (gimplify_parameters): Add gimple_seq * argument.
14853 * function.c: Include gimple.h and options.h.
14854 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14855 for the added local temporaries if needed.
14856 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14857 if there are any parameter cleanups, wrap whole body into a
14858 try/finally with the cleanups.
14860 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14863 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14864 Use GET_MODE_CLASS for scalar floating point.
14866 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14870 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14871 Fix call of call_cgraph_insertion_hooks.
14873 2018-01-18 Martin Sebor <msebor@redhat.com>
14875 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14877 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14880 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14883 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14886 * common.opt: (-ffile-prefix-map): New option.
14887 * opts.c (common_handle_option): Defer it.
14888 * opts-global.c (handle_common_deferred_options): Handle it.
14889 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14890 * file-prefix-map.h: New file.
14891 (remap_debug_filename, add_debug_prefix_map): ...here.
14892 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14893 * final.c (debug_prefix_map, add_debug_prefix_map
14894 remap_debug_filename): Move to...
14895 * file-prefix-map.c: New file.
14896 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14897 generalize, get rid of alloca(), use strrchr() instead of strchr().
14898 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14899 Implement in terms of add_prefix_map().
14900 (remap_macro_filename, remap_debug_filename): Implement in term of
14902 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14903 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14904 * dbxout.c: Include file-prefix-map.h.
14905 * varasm.c: Likewise.
14906 * vmsdbgout.c: Likewise.
14907 * xcoffout.c: Likewise.
14908 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14909 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14910 * doc/invoke.texi (-ffile-prefix-map): Document.
14911 (-fdebug-prefix-map): Update description.
14913 2018-01-18 Martin Liska <mliska@suse.cz>
14915 * config/i386/i386.c (indirect_thunk_name): Document that also
14917 (output_indirect_thunk): Document why both instructions
14918 (pause and lfence) are generated.
14920 2018-01-18 Richard Biener <rguenther@suse.de>
14922 PR tree-optimization/83887
14923 * graphite-scop-detection.c
14924 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14925 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14926 (scop_detection::merge_sese): Re-implement with a flood-fill
14927 algorithm that properly finds a SESE region if it exists.
14929 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14932 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14933 pointer_diff optimizations use view_convert instead of convert.
14935 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14937 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14938 Generate different code for -mno-speculate-indirect-jumps.
14939 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14940 (*call_indirect_aix<mode>): Disable for
14941 -mno-speculate-indirect-jumps.
14942 (*call_indirect_aix<mode>_nospec): New define_insn.
14943 (*call_value_indirect_aix<mode>): Disable for
14944 -mno-speculate-indirect-jumps.
14945 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14946 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14947 -mno-speculate-indirect-jumps.
14948 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14950 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14952 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14953 long double type, set the flags for noting the default long double
14954 type, even if we don't pass or return a long double type.
14956 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14959 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14962 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14964 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14965 support for merge[hl].
14966 (fold_mergehl_helper): New helper function.
14967 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14968 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14969 (altivec_vmrglw_direct): Add xxmrglw insn.
14971 2018-01-17 Andrew Waterman <andrew@sifive.com>
14973 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14974 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14976 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14979 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14980 call the lto_location_cache before reading the
14981 DECL_SOURCE_LOCATION of the types.
14983 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14984 Richard Sandiford <richard.sandiford@linaro.org>
14986 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14987 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14988 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14989 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14990 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14992 * config/aarch64/constraints.md (aarch64_movti_operand):
14994 * config/aarch64/predicates.md (Uti): Add new constraint.
14996 2018-01-17 Carl Love <cel@us.ibm.com>
14998 * config/rs6000/vsx.md (define_expand xl_len_r,
14999 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15000 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15002 (define_expand, define_insn): Move the shift left from the
15003 define_insn to the define_expand for lxvl and stxvl instructions.
15004 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15005 and XL_LEN_R definitions to PURE.
15007 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15009 * config/i386/i386.c (indirect_thunk_name): Declare regno
15010 as unsigned int. Compare regno with INVALID_REGNUM.
15011 (output_indirect_thunk): Ditto.
15012 (output_indirect_thunk_function): Ditto.
15013 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15014 in the call to output_indirect_thunk_function.
15016 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15018 PR middle-end/83884
15019 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15020 rather than the size of inner_type to determine the stack slot size
15021 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15023 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15026 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15029 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15031 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15032 endian Linux systems to optionally enable multilibs for selecting
15033 the long double type if the user configured an explicit type.
15034 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15035 have no long double multilibs if not defined.
15036 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15037 warn if the user used -mabi={ieee,ibm}longdouble and we built
15038 multilibs for long double.
15039 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15040 appropriate multilib option.
15041 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15043 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15044 for building long double multilibs.
15045 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15047 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15049 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15052 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15054 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15057 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15060 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15063 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15065 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15066 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15068 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15070 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15071 different rtl trees depending on TARGET_64BIT.
15072 (rs6000_gen_lvx): Likewise.
15074 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15076 * config/visium/visium.md (nop): Tweak comment.
15077 (hazard_nop): Likewise.
15079 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15081 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15082 -mspeculate-indirect-jumps.
15083 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15084 for -mno-speculate-indirect-jumps.
15085 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15086 (*call_value_indirect_elfv2<mode>): Disable for
15087 -mno-speculate-indirect-jumps.
15088 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15089 (indirect_jump): Emit different RTL for
15090 -mno-speculate-indirect-jumps.
15091 (*indirect_jump<mode>): Disable for
15092 -mno-speculate-indirect-jumps.
15093 (*indirect_jump<mode>_nospec): New define_insn.
15094 (tablejump): Emit different RTL for
15095 -mno-speculate-indirect-jumps.
15096 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15097 (tablejumpsi_nospec): New define_expand.
15098 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15099 (tablejumpdi_nospec): New define_expand.
15100 (*tablejump<mode>_internal1): Disable for
15101 -mno-speculate-indirect-jumps.
15102 (*tablejump<mode>_internal1_nospec): New define_insn.
15103 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15106 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15108 * caller-save.c (insert_save): Drop unnecessary parameter. All
15111 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15112 Richard Biener <rguenth@suse.de>
15115 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15116 return early, inline manually is_gimple_sizepos. Make sure if we
15117 call gimplify_expr we don't end up with a gimple constant.
15118 * tree.c (variably_modified_type_p): Don't return true for
15119 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15120 * gimplify.h (is_gimple_sizepos): Remove.
15122 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15124 PR tree-optimization/83857
15125 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15126 vectorizable_live_operation for pure SLP statements.
15127 (vectorizable_live_operation): Handle PHIs.
15129 2018-01-16 Richard Biener <rguenther@suse.de>
15131 PR tree-optimization/83867
15132 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15133 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15135 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15138 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15139 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15140 If off is not INTEGER_CST, issue a may not be aligned warning
15141 rather than isn't aligned. Use isn%'t rather than isn't.
15142 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15144 <case MULT_EXPR>: Improve the case when bottom and one of the
15145 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15146 operand, in that case check if the other operand is multiple of
15147 bottom divided by the INTEGER_CST operand.
15149 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15152 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15153 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15154 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15155 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15156 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15157 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15158 (pa_function_arg_size): New function.
15160 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15162 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15163 in a separate statement.
15165 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15167 PR tree-optimization/83847
15168 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15169 group gathers and scatters.
15171 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15173 PR rtl-optimization/86620
15174 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15176 PR rtl-optimization/83213
15177 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15178 to last if both are JUMP_INSNs.
15180 PR tree-optimization/83843
15181 * gimple-ssa-store-merging.c
15182 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15183 store_immediate_info for bswap/nop orig_stores.
15185 2018-01-15 Andrew Waterman <andrew@sifive.com>
15187 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15189 <UDIV>: Increase cost if !TARGET_DIV.
15191 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15193 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15194 (define_attr "cr_logical_3op"): New.
15195 (cceq_ior_compare): Adjust.
15196 (cceq_ior_compare_complement): Adjust.
15197 (*cceq_rev_compare): Adjust.
15198 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15199 (is_cracked_insn): Adjust.
15200 (insn_must_be_first_in_group): Adjust.
15201 * config/rs6000/40x.md: Adjust.
15202 * config/rs6000/440.md: Adjust.
15203 * config/rs6000/476.md: Adjust.
15204 * config/rs6000/601.md: Adjust.
15205 * config/rs6000/603.md: Adjust.
15206 * config/rs6000/6xx.md: Adjust.
15207 * config/rs6000/7450.md: Adjust.
15208 * config/rs6000/7xx.md: Adjust.
15209 * config/rs6000/8540.md: Adjust.
15210 * config/rs6000/cell.md: Adjust.
15211 * config/rs6000/e300c2c3.md: Adjust.
15212 * config/rs6000/e500mc.md: Adjust.
15213 * config/rs6000/e500mc64.md: Adjust.
15214 * config/rs6000/e5500.md: Adjust.
15215 * config/rs6000/e6500.md: Adjust.
15216 * config/rs6000/mpc.md: Adjust.
15217 * config/rs6000/power4.md: Adjust.
15218 * config/rs6000/power5.md: Adjust.
15219 * config/rs6000/power6.md: Adjust.
15220 * config/rs6000/power7.md: Adjust.
15221 * config/rs6000/power8.md: Adjust.
15222 * config/rs6000/power9.md: Adjust.
15223 * config/rs6000/rs64.md: Adjust.
15224 * config/rs6000/titan.md: Adjust.
15226 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15228 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15229 ix86_indirect_branch_register logic.
15231 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15233 * config/i386/constraints.md (Bs): Update
15234 ix86_indirect_branch_register check. Don't check
15235 ix86_indirect_branch_register with GOT_memory_operand.
15237 * config/i386/predicates.md (GOT_memory_operand): Don't check
15238 ix86_indirect_branch_register here.
15239 (GOT32_symbol_operand): Likewise.
15241 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15243 * config/i386/predicates.md (constant_call_address_operand):
15244 Rewrite ix86_indirect_branch_register logic.
15245 (sibcall_insn_operand): Likewise.
15247 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15249 * config/i386/constraints.md (Bs): Replace
15250 ix86_indirect_branch_thunk_register with
15251 ix86_indirect_branch_register.
15253 * config/i386/i386.md (indirect_jump): Likewise.
15254 (tablejump): Likewise.
15255 (*sibcall_memory): Likewise.
15256 (*sibcall_value_memory): Likewise.
15257 Peepholes of indirect call and jump via memory: Likewise.
15258 * config/i386/i386.opt: Likewise.
15259 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15260 (GOT_memory_operand): Likewise.
15261 (call_insn_operand): Likewise.
15262 (sibcall_insn_operand): Likewise.
15263 (GOT32_symbol_operand): Likewise.
15265 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15267 PR middle-end/83837
15268 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15269 type rather than type addr's type points to.
15270 (expand_omp_atomic_mutex): Likewise.
15271 (expand_omp_atomic): Likewise.
15273 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15276 * config/i386/i386.c (output_indirect_thunk_function): Use
15277 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15278 for __x86_return_thunk.
15280 2018-01-15 Richard Biener <rguenther@suse.de>
15282 PR middle-end/83850
15283 * expmed.c (extract_bit_field_1): Fix typo.
15285 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15288 * config/arm/iterators.md (VF): New mode iterator.
15289 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15290 Remove integer-related logic from pattern.
15291 (neon_vabd<mode>_3): Likewise.
15293 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15295 PR middle-end/82694
15296 * common.opt (fstrict-overflow): No longer an alias.
15297 (fwrapv-pointer): New option.
15298 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15299 also for pointer types based on flag_wrapv_pointer.
15300 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15301 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15302 opts->x_flag_wrapv got set.
15303 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15304 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15305 POINTER_TYPE_OVERFLOW_UNDEFINED.
15306 * match.pd: Likewise in address comparison pattern.
15307 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15309 2018-01-15 Richard Biener <rguenther@suse.de>
15312 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15313 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15314 Reset type names to their identifier if their TYPE_DECL doesn't
15315 have linkage (and thus is used for ODR and devirt).
15316 (save_debug_info_for_decl): Remove.
15317 (save_debug_info_for_type): Likewise.
15318 (add_tree_to_fld_list): Adjust.
15319 * tree-pretty-print.c (dump_generic_node): Make dumping of
15320 type names more robust.
15322 2018-01-15 Richard Biener <rguenther@suse.de>
15324 * BASE-VER: Bump to 8.0.1.
15326 2018-01-14 Martin Sebor <msebor@redhat.com>
15329 * builtins.c (check_access): Avoid warning when the no-warning bit
15332 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15334 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15335 * ira-color (allocno_hard_regs_compare): Likewise.
15337 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15340 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15341 Use .pushsection/.popsection.
15343 2018-01-14 Martin Sebor <msebor@redhat.com>
15346 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15348 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15350 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15351 entry from extra_headers.
15352 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15353 extra_headers, make the list bitwise identical to the i?86-*-* one.
15355 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15357 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15358 -mcmodel=large with -mindirect-branch=thunk,
15359 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15360 -mfunction-return=thunk-extern.
15361 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15362 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15363 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15365 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15367 * config/i386/i386.c (print_reg): Print the name of the full
15368 integer register without '%'.
15369 (ix86_print_operand): Handle 'V'.
15370 * doc/extend.texi: Document 'V' modifier.
15372 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15374 * config/i386/constraints.md (Bs): Disallow memory operand for
15375 -mindirect-branch-register.
15377 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15378 (GOT_memory_operand): Likewise.
15379 (call_insn_operand): Likewise.
15380 (sibcall_insn_operand): Likewise.
15381 (GOT32_symbol_operand): Likewise.
15382 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15383 for -mindirect-branch-register.
15384 (tablejump): Likewise.
15385 (*sibcall_memory): Likewise.
15386 (*sibcall_value_memory): Likewise.
15387 Disallow peepholes of indirect call and jump via memory for
15388 -mindirect-branch-register.
15389 (*call_pop): Replace m with Bw.
15390 (*call_value_pop): Likewise.
15391 (*sibcall_pop_memory): Replace m with Bs.
15392 * config/i386/i386.opt (mindirect-branch-register): New option.
15393 * doc/invoke.texi: Document -mindirect-branch-register option.
15395 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15397 * config/i386/i386-protos.h (ix86_output_function_return): New.
15398 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15399 set function_return_type.
15400 (indirect_thunk_name): Add ret_p to indicate thunk for function
15402 (output_indirect_thunk_function): Pass false to
15403 indirect_thunk_name.
15404 (ix86_output_indirect_branch_via_reg): Likewise.
15405 (ix86_output_indirect_branch_via_push): Likewise.
15406 (output_indirect_thunk_function): Create alias for function
15407 return thunk if regno < 0.
15408 (ix86_output_function_return): New function.
15409 (ix86_handle_fndecl_attribute): Handle function_return.
15410 (ix86_attribute_table): Add function_return.
15411 * config/i386/i386.h (machine_function): Add
15412 function_return_type.
15413 * config/i386/i386.md (simple_return_internal): Use
15414 ix86_output_function_return.
15415 (simple_return_internal_long): Likewise.
15416 * config/i386/i386.opt (mfunction-return=): New option.
15417 (indirect_branch): Mention -mfunction-return=.
15418 * doc/extend.texi: Document function_return function attribute.
15419 * doc/invoke.texi: Document -mfunction-return= option.
15421 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15423 * config/i386/i386-opts.h (indirect_branch): New.
15424 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15425 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15426 with local indirect jump when converting indirect call and jump.
15427 (ix86_set_indirect_branch_type): New.
15428 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15429 (indirectlabelno): New.
15430 (indirect_thunk_needed): Likewise.
15431 (indirect_thunk_bnd_needed): Likewise.
15432 (indirect_thunks_used): Likewise.
15433 (indirect_thunks_bnd_used): Likewise.
15434 (INDIRECT_LABEL): Likewise.
15435 (indirect_thunk_name): Likewise.
15436 (output_indirect_thunk): Likewise.
15437 (output_indirect_thunk_function): Likewise.
15438 (ix86_output_indirect_branch_via_reg): Likewise.
15439 (ix86_output_indirect_branch_via_push): Likewise.
15440 (ix86_output_indirect_branch): Likewise.
15441 (ix86_output_indirect_jmp): Likewise.
15442 (ix86_code_end): Call output_indirect_thunk_function if needed.
15443 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15445 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15446 (ix86_attribute_table): Add indirect_branch.
15447 * config/i386/i386.h (machine_function): Add indirect_branch_type
15448 and has_local_indirect_jump.
15449 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15451 (tablejump): Likewise.
15452 (*indirect_jump): Use ix86_output_indirect_jmp.
15453 (*tablejump_1): Likewise.
15454 (simple_return_indirect_internal): Likewise.
15455 * config/i386/i386.opt (mindirect-branch=): New option.
15456 (indirect_branch): New.
15459 (thunk-inline): Likewise.
15460 (thunk-extern): Likewise.
15461 * doc/extend.texi: Document indirect_branch function attribute.
15462 * doc/invoke.texi: Document -mindirect-branch= option.
15464 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15467 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15469 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15471 * ipa-inline.c (want_inline_small_function_p): Return false if
15472 inlining has already failed with CIF_FINAL_ERROR.
15473 (update_caller_keys): Call want_inline_small_function_p before
15475 (update_callee_keys): Likewise.
15477 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15479 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15481 (rs6000_quadword_masked_address_p): Likewise.
15482 (quad_aligned_load_p): Likewise.
15483 (quad_aligned_store_p): Likewise.
15484 (const_load_sequence_p): Add comment to describe the outer-most loop.
15485 (mimic_memory_attributes_and_flags): New function.
15486 (rs6000_gen_stvx): Likewise.
15487 (replace_swapped_aligned_store): Likewise.
15488 (rs6000_gen_lvx): Likewise.
15489 (replace_swapped_aligned_load): Likewise.
15490 (replace_swapped_load_constant): Capitalize argument name in
15491 comment describing this function.
15492 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15493 and stores that access quad-word aligned addresses and replace
15494 with stvx or lvx instructions when appropriate.
15495 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15496 New function prototype.
15497 (rs6000_quadword_masked_address_p): Likewise.
15498 (rs6000_gen_lvx): Likewise.
15499 (rs6000_gen_stvx): Likewise.
15500 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15501 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15502 when memory address is aligned.
15503 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15504 this split to select lvx instruction when memory address is aligned.
15505 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15506 instruction when memory address is aligned.
15507 (*vsx_le_perm_load_v16qi): Likewise.
15508 (four unnamed splitters): Modify to select the stvx instruction
15509 when memory is aligned.
15511 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15513 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15514 which appears in the queue multiple times.
15516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15517 Alan Hayward <alan.hayward@arm.com>
15518 David Sherwood <david.sherwood@arm.com>
15520 * tree-vectorizer.h (vec_lower_bound): New structure.
15521 (_loop_vec_info): Add check_nonzero and lower_bounds.
15522 (LOOP_VINFO_CHECK_NONZERO): New macro.
15523 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15524 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15525 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15526 fields. Make seg_len the distance travelled, not including the
15528 (dr_direction_indicator): Declare.
15529 (dr_zero_step_indicator): Likewise.
15530 (dr_known_forward_stride_p): Likewise.
15531 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15533 (runtime_alias_check_p): Allow runtime alias checks with
15535 (operator ==): Compare access_size and align.
15536 (prune_runtime_alias_test_list): Rework for new distinction between
15537 the access_size and seg_len.
15538 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15540 (get_segment_min_max): New function.
15541 (create_intersect_range_checks): Use it.
15542 (dr_step_indicator): New function.
15543 (dr_direction_indicator): Likewise.
15544 (dr_zero_step_indicator): Likewise.
15545 (dr_known_forward_stride_p): Likewise.
15546 * tree-loop-distribution.c (data_ref_segment_size): Return
15547 DR_STEP * (niters - 1).
15548 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15550 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15551 (vect_preserves_scalar_order_p): New function, split out from...
15552 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15553 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15554 (vect_vfa_access_size): New function.
15555 (vect_vfa_align): Likewise.
15556 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15557 (dump_lower_bound): New function.
15558 (vect_check_lower_bound): Likewise.
15559 (vect_small_gap_p): Likewise.
15560 (vectorizable_with_step_bound_p): Likewise.
15561 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15562 depencies if the vectorization factor is 1. Convert the checks
15563 for nonzero steps into checks on the bounds of DR_STEP. Try using
15564 a bunds check for variable steps if the minimum required step is
15565 relatively small. Update calls to the dr_with_seg_len
15566 constructor and to vect_compile_time_alias.
15567 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15569 (vect_loop_versioning): Call it.
15570 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15572 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15575 Alan Hayward <alan.hayward@arm.com>
15576 David Sherwood <david.sherwood@arm.com>
15578 * doc/sourcebuild.texi (vect_scatter_store): Document.
15579 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15581 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15583 * genopinit.c (main): Add supports_vec_scatter_store and
15584 supports_vec_scatter_store_cached to target_optabs.
15585 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15586 IFN_MASK_SCATTER_STORE.
15587 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15589 * internal-fn.h (internal_store_fn_p): Declare.
15590 (internal_fn_stored_value_index): Likewise.
15591 * internal-fn.c (scatter_store_direct): New macro.
15592 (expand_scatter_store_optab_fn): New function.
15593 (direct_scatter_store_optab_supported_p): New macro.
15594 (internal_store_fn_p): New function.
15595 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15596 IFN_MASK_SCATTER_STORE.
15597 (internal_fn_mask_index): Likewise.
15598 (internal_fn_stored_value_index): New function.
15599 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15600 for scatter stores.
15601 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15602 * optabs-query.c (supports_vec_scatter_store_p): New function.
15603 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15604 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15605 true for scatter stores.
15606 (vect_gather_scatter_fn_p): Handle scatter stores too.
15607 (vect_check_gather_scatter): Consider using scatter stores if
15608 supports_vec_scatter_store_p.
15609 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15610 scatter stores too.
15611 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15612 internal_fn_stored_value_index.
15613 (check_load_store_masking): Handle scatter stores too.
15614 (vect_get_store_rhs): Make public.
15615 (vectorizable_call): Use internal_store_fn_p.
15616 (vectorizable_store): Handle scatter store internal functions.
15617 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15618 when deciding whether the end of the group has been reached.
15619 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15620 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15621 (mask_scatter_store<mode>): New insns.
15623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15624 Alan Hayward <alan.hayward@arm.com>
15625 David Sherwood <david.sherwood@arm.com>
15627 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15628 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15629 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15631 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15632 Use vect_truncate_gather_scatter_offset if we can't treat the
15633 operation as a normal gather load or scatter store.
15634 (get_group_load_store_type): Take the gather_scatter_info
15635 as argument. Try using a gather load or scatter store for
15636 single-element groups.
15637 (get_load_store_type): Update calls to get_group_load_store_type
15638 and vect_use_strided_gather_scatters_p.
15640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15641 Alan Hayward <alan.hayward@arm.com>
15642 David Sherwood <david.sherwood@arm.com>
15644 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15645 optional tree argument.
15646 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15648 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15649 but continue to use the current value as a fallback.
15650 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15651 to compare the updates.
15652 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15653 (get_load_store_type): Use it when handling a strided access.
15654 (vect_get_strided_load_store_ops): New function.
15655 (vect_get_data_ptr_increment): Likewise.
15656 (vectorizable_load): Handle strided gather loads. Always pass
15657 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15660 Alan Hayward <alan.hayward@arm.com>
15661 David Sherwood <david.sherwood@arm.com>
15663 * doc/md.texi (gather_load@var{m}): Document.
15664 (mask_gather_load@var{m}): Likewise.
15665 * genopinit.c (main): Add supports_vec_gather_load and
15666 supports_vec_gather_load_cached to target_optabs.
15667 * optabs-tree.c (init_tree_optimization_optabs): Use
15668 ggc_cleared_alloc to allocate target_optabs.
15669 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15670 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15672 * internal-fn.h (internal_load_fn_p): Declare.
15673 (internal_gather_scatter_fn_p): Likewise.
15674 (internal_fn_mask_index): Likewise.
15675 (internal_gather_scatter_fn_supported_p): Likewise.
15676 * internal-fn.c (gather_load_direct): New macro.
15677 (expand_gather_load_optab_fn): New function.
15678 (direct_gather_load_optab_supported_p): New macro.
15679 (direct_internal_fn_optab): New function.
15680 (internal_load_fn_p): Likewise.
15681 (internal_gather_scatter_fn_p): Likewise.
15682 (internal_fn_mask_index): Likewise.
15683 (internal_gather_scatter_fn_supported_p): Likewise.
15684 * optabs-query.c (supports_at_least_one_mode_p): New function.
15685 (supports_vec_gather_load_p): Likewise.
15686 * optabs-query.h (supports_vec_gather_load_p): Declare.
15687 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15688 and memory_type field.
15689 (NUM_PATTERNS): Bump to 15.
15690 * tree-vect-data-refs.c: Include internal-fn.h.
15691 (vect_gather_scatter_fn_p): New function.
15692 (vect_describe_gather_scatter_call): Likewise.
15693 (vect_check_gather_scatter): Try using internal functions for
15694 gather loads. Recognize existing calls to a gather load function.
15695 (vect_analyze_data_refs): Consider using gather loads if
15696 supports_vec_gather_load_p.
15697 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15698 (vect_get_gather_scatter_offset_type): Likewise.
15699 (vect_convert_mask_for_vectype): Likewise.
15700 (vect_add_conversion_to_patterm): Likewise.
15701 (vect_try_gather_scatter_pattern): Likewise.
15702 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15703 (vect_vect_recog_func_ptrs): Add it.
15704 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15705 internal_fn_mask_index and internal_gather_scatter_fn_p.
15706 (check_load_store_masking): Take the gather_scatter_info as an
15707 argument and handle gather loads.
15708 (vect_get_gather_scatter_ops): New function.
15709 (vectorizable_call): Check internal_load_fn_p.
15710 (vectorizable_load): Likewise. Handle gather load internal
15712 (vectorizable_store): Update call to check_load_store_masking.
15713 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15714 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15715 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15716 (aarch64_gather_scale_operand_d): New predicates.
15717 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15718 (mask_gather_load<mode>): New insns.
15720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15721 Alan Hayward <alan.hayward@arm.com>
15722 David Sherwood <david.sherwood@arm.com>
15724 * optabs.def (fold_left_plus_optab): New optab.
15725 * doc/md.texi (fold_left_plus_@var{m}): Document.
15726 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15727 * internal-fn.c (fold_left_direct): Define.
15728 (expand_fold_left_optab_fn): Likewise.
15729 (direct_fold_left_optab_supported_p): Likewise.
15730 * fold-const-call.c (fold_const_fold_left): New function.
15731 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15732 * tree-parloops.c (valid_reduction_p): New function.
15733 (gather_scalar_reductions): Use it.
15734 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15735 (vect_finish_replace_stmt): Declare.
15736 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15737 (needs_fold_left_reduction_p): New function, split out from...
15738 (vect_is_simple_reduction): ...here. Accept reductions that
15739 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15740 (vect_force_simple_reduction): Also store the reduction type in
15741 the assignment's STMT_VINFO_REDUC_TYPE.
15742 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15743 (merge_with_identity): New function.
15744 (vect_expand_fold_left): Likewise.
15745 (vectorize_fold_left_reduction): Likewise.
15746 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15747 scalar phi in place for it. Check for target support and reject
15748 cases that would reassociate the operation. Defer the transform
15749 phase to vectorize_fold_left_reduction.
15750 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15751 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15752 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15756 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15757 call to ifc_temp_var.
15759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15760 Alan Hayward <alan.hayward@arm.com>
15761 David Sherwood <david.sherwood@arm.com>
15763 * target.def (legitimize_address_displacement): Take the original
15764 offset as a poly_int.
15765 * targhooks.h (default_legitimize_address_displacement): Update
15767 * targhooks.c (default_legitimize_address_displacement): Likewise.
15768 * doc/tm.texi: Regenerate.
15769 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15770 as an argument, moving assert of ad->disp == ad->disp_term to...
15771 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15772 Try calling targetm.legitimize_address_displacement before expanding
15773 the address rather than afterwards, and adjust for the new interface.
15774 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15775 Match the new hook interface. Handle SVE addresses.
15776 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15777 new hook interface.
15779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15781 * Makefile.in (OBJS): Add early-remat.o.
15782 * target.def (select_early_remat_modes): New hook.
15783 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15784 * doc/tm.texi: Regenerate.
15785 * targhooks.h (default_select_early_remat_modes): Declare.
15786 * targhooks.c (default_select_early_remat_modes): New function.
15787 * timevar.def (TV_EARLY_REMAT): New timevar.
15788 * passes.def (pass_early_remat): New pass.
15789 * tree-pass.h (make_pass_early_remat): Declare.
15790 * early-remat.c: New file.
15791 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15793 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15796 Alan Hayward <alan.hayward@arm.com>
15797 David Sherwood <david.sherwood@arm.com>
15799 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15800 vfm1 with a bound_epilog parameter.
15801 (vect_do_peeling): Update calls accordingly, and move the prologue
15802 call earlier in the function. Treat the base bound_epilog as 0 for
15803 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15804 this base when peeling for gaps.
15805 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15806 with fully-masked loops.
15807 (vect_estimate_min_profitable_iters): Handle the single peeled
15808 iteration in that case.
15810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15811 Alan Hayward <alan.hayward@arm.com>
15812 David Sherwood <david.sherwood@arm.com>
15814 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15815 single-element interleaving even if the size is not a power of 2.
15816 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15817 accesses for single-element interleaving if the group size is
15820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15821 Alan Hayward <alan.hayward@arm.com>
15822 David Sherwood <david.sherwood@arm.com>
15824 * doc/md.texi (fold_extract_last_@var{m}): Document.
15825 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15826 * optabs.def (fold_extract_last_optab): New optab.
15827 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15828 * internal-fn.c (fold_extract_direct): New macro.
15829 (expand_fold_extract_optab_fn): Likewise.
15830 (direct_fold_extract_optab_supported_p): Likewise.
15831 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15832 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15833 EXTRACT_LAST_REDUCTION.
15834 (get_initial_def_for_reduction): Do not create an initial vector
15835 for EXTRACT_LAST_REDUCTION reductions.
15836 (vectorizable_reduction): Leave the scalar phi in place for
15837 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15838 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15839 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15840 transform phase to vectorizable_condition.
15841 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15843 (vect_finish_stmt_generation): ...here.
15844 (vect_finish_replace_stmt): New function.
15845 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15846 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15848 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15851 Alan Hayward <alan.hayward@arm.com>
15852 David Sherwood <david.sherwood@arm.com>
15854 * doc/md.texi (extract_last_@var{m}): Document.
15855 * optabs.def (extract_last_optab): New optab.
15856 * internal-fn.def (EXTRACT_LAST): New internal function.
15857 * internal-fn.c (cond_unary_direct): New macro.
15858 (expand_cond_unary_optab_fn): Likewise.
15859 (direct_cond_unary_optab_supported_p): Likewise.
15860 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15861 loops using EXTRACT_LAST.
15862 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15863 (extract_last_<mode>): ...this optab.
15864 (vec_extract<mode><Vel>): Update accordingly.
15866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15867 Alan Hayward <alan.hayward@arm.com>
15868 David Sherwood <david.sherwood@arm.com>
15870 * target.def (empty_mask_is_expensive): New hook.
15871 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15872 * doc/tm.texi: Regenerate.
15873 * targhooks.h (default_empty_mask_is_expensive): Declare.
15874 * targhooks.c (default_empty_mask_is_expensive): New function.
15875 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15876 if the target says that empty masks are expensive.
15877 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15879 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15882 Alan Hayward <alan.hayward@arm.com>
15883 David Sherwood <david.sherwood@arm.com>
15885 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15886 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15887 (vect_use_loop_mask_for_alignment_p): New function.
15888 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15889 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15890 niters_skip argument. Make sure that the first niters_skip elements
15891 of the first iteration are inactive.
15892 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15893 Update call to vect_set_loop_masks_directly.
15894 (get_misalign_in_elems): New function, split out from...
15895 (vect_gen_prolog_loop_niters): ...here.
15896 (vect_update_init_of_dr): Take a code argument that specifies whether
15897 the adjustment should be added or subtracted.
15898 (vect_update_init_of_drs): Likewise.
15899 (vect_prepare_for_masked_peels): New function.
15900 (vect_do_peeling): Skip prologue peeling if we're using a mask
15901 instead. Update call to vect_update_inits_of_drs.
15902 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15904 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15905 alignment. Do not include the number of peeled iterations in
15906 the minimum threshold in that case.
15907 (vectorizable_induction): Adjust the start value down by
15908 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15909 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15910 Take the number of skipped iterations into account when calculating
15912 * tree-vect-stmts.c (vect_gen_while_not): New function.
15914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15915 Alan Hayward <alan.hayward@arm.com>
15916 David Sherwood <david.sherwood@arm.com>
15918 * doc/sourcebuild.texi (vect_fully_masked): Document.
15919 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15920 default value to 0.
15921 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15923 (vect_analyze_loop_2): ...here. Don't check the vectorization
15924 factor against the number of loop iterations if the loop is
15927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15928 Alan Hayward <alan.hayward@arm.com>
15929 David Sherwood <david.sherwood@arm.com>
15931 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15932 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15933 (dump_groups): Update accordingly.
15934 (iv_use::mem_type): New member variable.
15935 (address_p): New function.
15936 (record_use): Add a mem_type argument and initialize the new
15938 (record_group_use): Add a mem_type argument. Use address_p.
15939 Remove obsolete null checks of base_object. Update call to record_use.
15940 (find_interesting_uses_op): Update call to record_group_use.
15941 (find_interesting_uses_cond): Likewise.
15942 (find_interesting_uses_address): Likewise.
15943 (get_mem_type_for_internal_fn): New function.
15944 (find_address_like_use): Likewise.
15945 (find_interesting_uses_stmt): Try find_address_like_use before
15946 calling find_interesting_uses_op.
15947 (addr_offset_valid_p): Use the iv mem_type field as the type
15948 of the addressed memory.
15949 (add_autoinc_candidates): Likewise.
15950 (get_address_cost): Likewise.
15951 (split_small_address_groups_p): Use address_p.
15952 (split_address_groups): Likewise.
15953 (add_iv_candidate_for_use): Likewise.
15954 (autoinc_possible_for_pair): Likewise.
15955 (rewrite_groups): Likewise.
15956 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15957 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15958 (get_alias_ptr_type_for_ptr_address): New function.
15959 (rewrite_use_address): Rewrite address uses in calls that were
15960 identified by find_address_like_use.
15962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15963 Alan Hayward <alan.hayward@arm.com>
15964 David Sherwood <david.sherwood@arm.com>
15966 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15968 * gimple-expr.h (is_gimple_addressable: Likewise.
15969 * gimple-expr.c (is_gimple_address): Likewise.
15970 * internal-fn.c (expand_call_mem_ref): New function.
15971 (expand_mask_load_optab_fn): Use it.
15972 (expand_mask_store_optab_fn): Likewise.
15974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15975 Alan Hayward <alan.hayward@arm.com>
15976 David Sherwood <david.sherwood@arm.com>
15978 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15979 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15980 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15981 (cond_umax@var{mode}): Document.
15982 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15983 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15984 (cond_umin_optab, cond_umax_optab): New optabs.
15985 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15986 (COND_IOR, COND_XOR): New internal functions.
15987 * internal-fn.h (get_conditional_internal_fn): Declare.
15988 * internal-fn.c (cond_binary_direct): New macro.
15989 (expand_cond_binary_optab_fn): Likewise.
15990 (direct_cond_binary_optab_supported_p): Likewise.
15991 (get_conditional_internal_fn): New function.
15992 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15993 Cope with reduction statements that are vectorized as calls rather
15995 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15996 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15997 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15998 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15999 (UNSPEC_COND_EOR): New unspecs.
16000 (optab): Add mappings for them.
16001 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16002 (sve_int_op, sve_fp_op): New int attributes.
16004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16005 Alan Hayward <alan.hayward@arm.com>
16006 David Sherwood <david.sherwood@arm.com>
16008 * optabs.def (while_ult_optab): New optab.
16009 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16010 * internal-fn.def (WHILE_ULT): New internal function.
16011 * internal-fn.h (direct_internal_fn_supported_p): New override
16012 that takes two types as argument.
16013 * internal-fn.c (while_direct): New macro.
16014 (expand_while_optab_fn): New function.
16015 (convert_optab_supported_p): Likewise.
16016 (direct_while_optab_supported_p): New macro.
16017 * wide-int.h (wi::udiv_ceil): New function.
16018 * tree-vectorizer.h (rgroup_masks): New structure.
16019 (vec_loop_masks): New typedef.
16020 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16021 and fully_masked_p.
16022 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16023 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16024 (vect_max_vf): New function.
16025 (slpeel_make_loop_iterate_ntimes): Delete.
16026 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16027 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16028 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16029 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16030 internal-fn.h, stor-layout.h and optabs-query.h.
16031 (vect_set_loop_mask): New function.
16032 (add_preheader_seq): Likewise.
16033 (add_header_seq): Likewise.
16034 (interleave_supported_p): Likewise.
16035 (vect_maybe_permute_loop_masks): Likewise.
16036 (vect_set_loop_masks_directly): Likewise.
16037 (vect_set_loop_condition_masked): Likewise.
16038 (vect_set_loop_condition_unmasked): New function, split out from
16039 slpeel_make_loop_iterate_ntimes.
16040 (slpeel_make_loop_iterate_ntimes): Rename to..
16041 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16042 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16043 (vect_do_peeling): Update call accordingly.
16044 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16046 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16047 mask_compare_type, can_fully_mask_p and fully_masked_p.
16048 (release_vec_loop_masks): New function.
16049 (_loop_vec_info): Use it to free the loop masks.
16050 (can_produce_all_loop_masks_p): New function.
16051 (vect_get_max_nscalars_per_iter): Likewise.
16052 (vect_verify_full_masking): Likewise.
16053 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16054 retries, and free the mask rgroups before retrying. Check loop-wide
16055 reasons for disallowing fully-masked loops. Make the final decision
16056 about whether use a fully-masked loop or not.
16057 (vect_estimate_min_profitable_iters): Do not assume that peeling
16058 for the number of iterations will be needed for fully-masked loops.
16059 (vectorizable_reduction): Disable fully-masked loops.
16060 (vectorizable_live_operation): Likewise.
16061 (vect_halve_mask_nunits): New function.
16062 (vect_double_mask_nunits): Likewise.
16063 (vect_record_loop_mask): Likewise.
16064 (vect_get_loop_mask): Likewise.
16065 (vect_transform_loop): Handle the case in which the final loop
16066 iteration might handle a partial vector. Call vect_set_loop_condition
16067 instead of slpeel_make_loop_iterate_ntimes.
16068 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16069 (check_load_store_masking): New function.
16070 (prepare_load_store_mask): Likewise.
16071 (vectorizable_store): Handle fully-masked loops.
16072 (vectorizable_load): Likewise.
16073 (supportable_widening_operation): Use vect_halve_mask_nunits for
16075 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16076 (vect_gen_while): New function.
16077 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16078 (aarch64_uqdec<mode>): New insn.
16080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16081 Alan Hayward <alan.hayward@arm.com>
16082 David Sherwood <david.sherwood@arm.com>
16084 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16085 (reduc_xor_scal_optab): New optabs.
16086 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16087 (reduc_xor_scal_@var{m}): Document.
16088 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16089 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16090 internal functions.
16091 * fold-const-call.c (fold_const_call): Handle them.
16092 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16093 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16094 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16095 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16096 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16097 (UNSPEC_XORV): New unspecs.
16098 (optab): Add entries for them.
16099 (BITWISEV): New int iterator.
16100 (bit_reduc_op): New int attributes.
16102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16103 Alan Hayward <alan.hayward@arm.com>
16104 David Sherwood <david.sherwood@arm.com>
16106 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16107 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16108 * optabs.def (vec_shl_insert_optab): New optab.
16109 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16110 (duplicate_and_interleave): Likewise.
16111 * tree-vect-loop.c: Include internal-fn.h.
16112 (neutral_op_for_slp_reduction): New function, split out from
16113 get_initial_defs_for_reduction.
16114 (get_initial_def_for_reduction): Handle option 2 for variable-length
16115 vectors by loading the neutral value into a vector and then shifting
16116 the initial value into element 0.
16117 (get_initial_defs_for_reduction): Replace the code argument with
16118 the neutral value calculated by neutral_op_for_slp_reduction.
16119 Use gimple_build_vector for constant-length vectors.
16120 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16121 but the first group_size elements have a neutral value.
16122 Use duplicate_and_interleave otherwise.
16123 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16124 Update call to get_initial_defs_for_reduction. Handle SLP
16125 reductions for variable-length vectors by creating one vector
16126 result for each scalar result, with the elements associated
16127 with other scalar results stubbed out with the neutral value.
16128 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16129 Require IFN_VEC_SHL_INSERT for double reductions on
16130 variable-length vectors, or SLP reductions that have
16131 a neutral value. Require can_duplicate_and_interleave_p
16132 support for variable-length unchained SLP reductions if there
16133 is no neutral value, such as for MIN/MAX reductions. Also require
16134 the number of vector elements to be a multiple of the number of
16135 SLP statements when doing variable-length unchained SLP reductions.
16136 Update call to vect_create_epilog_for_reduction.
16137 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16138 and remove initial values.
16139 (duplicate_and_interleave): Make public.
16140 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16141 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16144 Alan Hayward <alan.hayward@arm.com>
16145 David Sherwood <david.sherwood@arm.com>
16147 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16148 (can_duplicate_and_interleave_p): New function.
16149 (vect_get_and_check_slp_defs): Take the vector of statements
16150 rather than just the current one. Remove excess parentheses.
16151 Restriction rejectinon of vect_constant_def and vect_external_def
16152 for variable-length vectors to boolean types, or types for which
16153 can_duplicate_and_interleave_p is false.
16154 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16155 (duplicate_and_interleave): New function.
16156 (vect_get_constant_vectors): Use gimple_build_vector for
16157 constant-length vectors and suitable variable-length constant
16158 vectors. Use duplicate_and_interleave for other variable-length
16159 vectors. Don't defer the update when inserting new statements.
16161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16162 Alan Hayward <alan.hayward@arm.com>
16163 David Sherwood <david.sherwood@arm.com>
16165 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16166 min_profitable_iters doesn't go negative.
16168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16169 Alan Hayward <alan.hayward@arm.com>
16170 David Sherwood <david.sherwood@arm.com>
16172 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16173 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16174 * optabs.def (vec_mask_load_lanes_optab): New optab.
16175 (vec_mask_store_lanes_optab): Likewise.
16176 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16177 (MASK_STORE_LANES): Likewise.
16178 * internal-fn.c (mask_load_lanes_direct): New macro.
16179 (mask_store_lanes_direct): Likewise.
16180 (expand_mask_load_optab_fn): Handle masked operations.
16181 (expand_mask_load_lanes_optab_fn): New macro.
16182 (expand_mask_store_optab_fn): Handle masked operations.
16183 (expand_mask_store_lanes_optab_fn): New macro.
16184 (direct_mask_load_lanes_optab_supported_p): Likewise.
16185 (direct_mask_store_lanes_optab_supported_p): Likewise.
16186 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16188 (vect_load_lanes_supported): Likewise.
16189 * tree-vect-data-refs.c (strip_conversion): New function.
16190 (can_group_stmts_p): Likewise.
16191 (vect_analyze_data_ref_accesses): Use it instead of checking
16192 for a pair of assignments.
16193 (vect_store_lanes_supported): Take a masked_p parameter.
16194 (vect_load_lanes_supported): Likewise.
16195 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16196 vect_store_lanes_supported and vect_load_lanes_supported.
16197 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16198 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16199 parameter. Don't allow gaps for masked accesses.
16200 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16201 and vect_load_lanes_supported.
16202 (get_load_store_type): Take a masked_p parameter and update
16203 call to get_group_load_store_type.
16204 (vectorizable_store): Update call to get_load_store_type.
16205 Handle IFN_MASK_STORE_LANES.
16206 (vectorizable_load): Update call to get_load_store_type.
16207 Handle IFN_MASK_LOAD_LANES.
16209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16210 Alan Hayward <alan.hayward@arm.com>
16211 David Sherwood <david.sherwood@arm.com>
16213 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16215 * config/aarch64/aarch64-protos.h
16216 (aarch64_sve_struct_memory_operand_p): Declare.
16217 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16218 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16219 (VPRED, vpred): Handle SVE structure modes.
16220 * config/aarch64/constraints.md (Utx): New constraint.
16221 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16222 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16223 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16224 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16225 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16226 structure modes. Split into pieces after RA.
16227 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16228 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16230 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16231 SVE structure modes.
16232 (aarch64_classify_address): Likewise.
16233 (sizetochar): Move earlier in file.
16234 (aarch64_print_operand): Handle SVE register lists.
16235 (aarch64_array_mode): New function.
16236 (aarch64_sve_struct_memory_operand_p): Likewise.
16237 (TARGET_ARRAY_MODE): Redefine.
16239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16240 Alan Hayward <alan.hayward@arm.com>
16241 David Sherwood <david.sherwood@arm.com>
16243 * target.def (array_mode): New target hook.
16244 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16245 * doc/tm.texi: Regenerate.
16246 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16247 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16248 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16249 targetm.array_mode.
16250 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16254 Alan Hayward <alan.hayward@arm.com>
16255 David Sherwood <david.sherwood@arm.com>
16257 * fold-const.c (fold_binary_loc): Check the argument types
16258 rather than the result type when testing for a vector operation.
16260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16262 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16263 * doc/tm.texi: Regenerate.
16265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16266 Alan Hayward <alan.hayward@arm.com>
16267 David Sherwood <david.sherwood@arm.com>
16269 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16270 (sve): Document new AArch64 extension.
16271 * doc/md.texi (w): Extend the description of the AArch64
16272 constraint to include SVE vectors.
16273 (Upl, Upa): Document new AArch64 predicate constraints.
16274 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16276 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16277 (msve-vector-bits=): New option.
16278 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16279 SVE when these are disabled.
16280 (sve): New extension.
16281 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16282 modes. Adjust their number of units based on aarch64_sve_vg.
16283 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16284 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16285 aarch64_addr_query_type.
16286 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16287 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16288 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16289 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16290 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16291 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16292 (aarch64_simd_imm_zero_p): Delete.
16293 (aarch64_check_zero_based_sve_index_immediate): Declare.
16294 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16295 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16296 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16297 (aarch64_sve_float_mul_immediate_p): Likewise.
16298 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16299 rather than an rtx.
16300 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16301 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16302 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16303 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16304 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16305 (aarch64_regmode_natural_size): Likewise.
16306 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16307 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16309 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16310 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16311 for VG and the SVE predicate registers.
16312 (V_ALIASES): Add a "z"-prefixed alias.
16313 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16314 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16315 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16316 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16317 (REG_CLASS_NAMES): Add entries for them.
16318 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16319 and the predicate registers.
16320 (aarch64_sve_vg): Declare.
16321 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16322 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16323 (REGMODE_NATURAL_SIZE): Define.
16324 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16326 * config/aarch64/aarch64.c: Include cfgrtl.h.
16327 (simd_immediate_info): Add a constructor for series vectors,
16328 and an associated step field.
16329 (aarch64_sve_vg): New variable.
16330 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16331 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16332 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16333 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16334 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16335 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16336 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16337 (aarch64_get_mask_mode): New functions.
16338 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16339 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16340 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16341 predicate modes and predicate registers. Explicitly restrict
16342 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16343 to store a vector mode if it is recognized by
16344 aarch64_classify_vector_mode.
16345 (aarch64_regmode_natural_size): New function.
16346 (aarch64_hard_regno_caller_save_mode): Return the original mode
16348 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16349 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16350 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16351 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16353 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16354 does not overlap dest if the function is frame-related. Handle
16356 (aarch64_split_add_offset): New function.
16357 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16358 them aarch64_add_offset.
16359 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16360 and update call to aarch64_sub_sp.
16361 (aarch64_add_cfa_expression): New function.
16362 (aarch64_expand_prologue): Pass extra temporary registers to the
16363 functions above. Handle the case in which we need to emit new
16364 DW_CFA_expressions for registers that were originally saved
16365 relative to the stack pointer, but now have to be expressed
16366 relative to the frame pointer.
16367 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16369 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16370 IP0 and IP1 values for SVE frames.
16371 (aarch64_expand_vec_series): New function.
16372 (aarch64_expand_sve_widened_duplicate): Likewise.
16373 (aarch64_expand_sve_const_vector): Likewise.
16374 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16375 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16376 into the register, rather than emitting a SET directly.
16377 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16378 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16379 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16380 (offset_9bit_signed_scaled_p): New functions.
16381 (aarch64_replicate_bitmask_imm): New function.
16382 (aarch64_bitmask_imm): Use it.
16383 (aarch64_cannot_force_const_mem): Reject expressions involving
16384 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16385 (aarch64_classify_index): Handle SVE indices, by requiring
16386 a plain register index with a scale that matches the element size.
16387 (aarch64_classify_address): Handle SVE addresses. Assert that
16388 the mode of the address is VOIDmode or an integer mode.
16389 Update call to aarch64_classify_symbol.
16390 (aarch64_classify_symbolic_expression): Update call to
16391 aarch64_classify_symbol.
16392 (aarch64_const_vec_all_in_range_p): New function.
16393 (aarch64_print_vector_float_operand): Likewise.
16394 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16395 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16396 and the FP immediates 1.0 and 0.5.
16397 (aarch64_print_address_internal): Handle SVE addresses.
16398 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16399 (aarch64_regno_regclass): Handle predicate registers.
16400 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16402 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16403 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16404 (aarch64_convert_sve_vector_bits): New function.
16405 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16406 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16407 rather than an rtx.
16408 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16409 Handle SVE vector and predicate modes. Accept VL-based constants
16410 that need only one temporary register, and VL offsets that require
16411 no temporary registers.
16412 (aarch64_conditional_register_usage): Mark the predicate registers
16413 as fixed if SVE isn't available.
16414 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16415 Return true for SVE vector and predicate modes.
16416 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16417 rather than an unsigned int. Handle SVE modes.
16418 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16420 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16422 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16423 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16424 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16425 (aarch64_sve_float_mul_immediate_p): New functions.
16426 (aarch64_sve_valid_immediate): New function.
16427 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16428 Explicitly reject structure modes. Check for INDEX constants.
16429 Handle PTRUE and PFALSE constants.
16430 (aarch64_check_zero_based_sve_index_immediate): New function.
16431 (aarch64_simd_imm_zero_p): Delete.
16432 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16433 vector modes. Accept constants in the range of CNT[BHWD].
16434 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16435 ask for an Advanced SIMD mode.
16436 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16437 (aarch64_simd_vector_alignment): Handle SVE predicates.
16438 (aarch64_vectorize_preferred_vector_alignment): New function.
16439 (aarch64_simd_vector_alignment_reachable): Use it instead of
16441 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16442 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16444 (MAX_VECT_LEN): Delete.
16445 (expand_vec_perm_d): Add a vec_flags field.
16446 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16447 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16448 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16450 (aarch64_evpc_rev): Rename to...
16451 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16452 (aarch64_evpc_rev_global): New function.
16453 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16454 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16456 (aarch64_evpc_sve_tbl): New function.
16457 (aarch64_expand_vec_perm_const_1): Update after rename of
16458 aarch64_evpc_rev. Handle SVE permutes too, trying
16459 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16460 than aarch64_evpc_tbl.
16461 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16462 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16463 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16464 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16465 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16466 (aarch64_expand_sve_vcond): New functions.
16467 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16468 of aarch64_vector_mode_p.
16469 (aarch64_dwarf_poly_indeterminate_value): New function.
16470 (aarch64_compute_pressure_classes): Likewise.
16471 (aarch64_can_change_mode_class): Likewise.
16472 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16473 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16474 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16475 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16476 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16477 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16478 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16479 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16481 (Dn, Dl, Dr): Accept const as well as const_vector.
16482 (Dz): Likewise. Compare against CONST0_RTX.
16483 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16484 of "vector" where appropriate.
16485 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16486 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16487 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16488 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16489 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16490 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16491 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16492 (v_int_equiv): Extend to SVE modes.
16493 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16495 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16496 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16497 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16498 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16499 (SVE_COND_FP_CMP): New int iterators.
16500 (perm_hilo): Handle the new unpack unspecs.
16501 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16503 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16504 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16505 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16506 (aarch64_equality_operator, aarch64_constant_vector_operand)
16507 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16508 (aarch64_sve_nonimmediate_operand): Likewise.
16509 (aarch64_sve_general_operand): Likewise.
16510 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16511 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16512 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16513 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16514 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16515 (aarch64_sve_float_arith_immediate): Likewise.
16516 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16517 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16518 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16519 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16520 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16521 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16522 (aarch64_sve_float_arith_operand): Likewise.
16523 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16524 (aarch64_sve_float_mul_operand): Likewise.
16525 (aarch64_sve_vec_perm_operand): Likewise.
16526 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16527 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16528 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16529 as well as const_vector.
16530 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16531 in file. Use CONST0_RTX and CONSTM1_RTX.
16532 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16533 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16534 Use aarch64_simd_imm_zero.
16535 * config/aarch64/aarch64-sve.md: New file.
16536 * config/aarch64/aarch64.md: Include it.
16537 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16538 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16539 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16540 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16541 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16542 (sve): New attribute.
16543 (enabled): Disable instructions with the sve attribute unless
16545 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16546 aarch64_expand_mov_immediate.
16547 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16548 CNT[BHSD] immediates.
16549 (movti): Split CONST_POLY_INT moves into two halves.
16550 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16551 Split additions that need a temporary here if the destination
16552 is the stack pointer.
16553 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16554 (*add<mode>3_poly_1): New instruction.
16555 (set_clobber_cc): New expander.
16557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16559 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16560 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16561 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16562 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16563 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16564 Change innermode from fixed_mode_size to machine_mode.
16565 (simplify_subreg): Update call accordingly. Handle a constant-sized
16566 subreg of a variable-length CONST_VECTOR.
16568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16569 Alan Hayward <alan.hayward@arm.com>
16570 David Sherwood <david.sherwood@arm.com>
16572 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16573 (add_offset_to_base): New function, split out from...
16574 (create_mem_ref): ...here. When handling a scale other than 1,
16575 check first whether the address is valid without the offset.
16576 Add it into the base if so, leaving the index and scale as-is.
16578 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16581 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16582 fold_for_warn before checking if arg2 is INTEGER_CST.
16584 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16586 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16587 (store_multiple_operation): Delete.
16588 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16589 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16590 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16591 guarded by TARGET_STRING.
16592 (rs6000_output_load_multiple): Delete.
16593 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16594 OPTION_MASK_STRING / TARGET_STRING handling.
16595 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16596 (const rs6000_opt_masks) <"string">: Change mask to 0.
16597 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16598 (MASK_STRING): Delete.
16599 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16601 (load_multiple): Delete.
16608 (store_multiple): Delete.
16615 (movmemsi_8reg): Delete.
16616 (corresponding unnamed define_insn): Delete.
16617 (movmemsi_6reg): Delete.
16618 (corresponding unnamed define_insn): Delete.
16619 (movmemsi_4reg): Delete.
16620 (corresponding unnamed define_insn): Delete.
16621 (movmemsi_2reg): Delete.
16622 (corresponding unnamed define_insn): Delete.
16623 (movmemsi_1reg): Delete.
16624 (corresponding unnamed define_insn): Delete.
16625 * config/rs6000/rs6000.opt (mno-string): New.
16626 (mstring): Replace by deprecation warning stub.
16627 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16629 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16631 * regrename.c (regrename_do_replace): If replacing the same
16632 reg multiple times, try to reuse last created gen_raw_REG.
16635 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16636 main to workaround a bug in GDB.
16638 2018-01-12 Tom de Vries <tom@codesourcery.com>
16641 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16643 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16645 PR rtl-optimization/80481
16646 * ira-color.c (get_cap_member): New function.
16647 (allocnos_conflict_by_live_ranges_p): Use it.
16648 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16649 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16651 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16654 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16655 (*saddl_se_1): Ditto.
16656 (*ssubsi_1): Ditto.
16657 (*ssubl_se_1): Ditto.
16659 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16661 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16662 rather than wi::to_widest for DR_INITs.
16663 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16664 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16665 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16667 (vect_analyze_group_access_1): Note that here.
16669 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16671 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16672 polynomial type sizes.
16674 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16676 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16677 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16678 (gimple_add_tmp_var): Likewise.
16680 2018-01-12 Martin Liska <mliska@suse.cz>
16682 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16683 (gimple_alloc_sizes): Likewise.
16684 (dump_gimple_statistics): Use PRIu64 in printf format.
16685 * gimple.h: Change uint64_t to int.
16687 2018-01-12 Martin Liska <mliska@suse.cz>
16689 * tree-core.h: Use uint64_t instead of int.
16690 * tree.c (tree_node_counts): Likewise.
16691 (tree_node_sizes): Likewise.
16692 (dump_tree_statistics): Use PRIu64 in printf format.
16694 2018-01-12 Martin Liska <mliska@suse.cz>
16696 * Makefile.in: As qsort_chk is implemented in vec.c, add
16697 vec.o to linkage of gencfn-macros.
16698 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16699 passing the info to record_node_allocation_statistics.
16700 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16702 * ggc-common.c (struct ggc_usage): Add operator== and use
16703 it in operator< and compare function.
16704 * mem-stats.h (struct mem_usage): Likewise.
16705 * vec.c (struct vec_usage): Remove operator< and compare
16706 function. Can be simply inherited.
16708 2018-01-12 Martin Jambor <mjambor@suse.cz>
16711 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16712 * tree-ssa-math-opts.c: Include domwalk.h.
16713 (convert_mult_to_fma_1): New function.
16714 (fma_transformation_info): New type.
16715 (fma_deferring_state): Likewise.
16716 (cancel_fma_deferring): New function.
16717 (result_of_phi): Likewise.
16718 (last_fma_candidate_feeds_initial_phi): Likewise.
16719 (convert_mult_to_fma): Added deferring logic, split actual
16720 transformation to convert_mult_to_fma_1.
16721 (math_opts_dom_walker): New type.
16722 (math_opts_dom_walker::after_dom_children): New method, body moved
16723 here from pass_optimize_widening_mul::execute, added deferring logic
16725 (pass_optimize_widening_mul::execute): Moved most of code to
16726 math_opts_dom_walker::after_dom_children.
16727 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16728 * config/i386/i386.c (ix86_option_override_internal): Added
16729 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16731 2018-01-12 Richard Biener <rguenther@suse.de>
16734 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16735 inline instance vars.
16737 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16740 * config/rx/rx.c (rx_is_restricted_memory_address):
16741 Handle SUBREG case.
16743 2018-01-12 Richard Biener <rguenther@suse.de>
16745 PR tree-optimization/80846
16746 * target.def (split_reduction): New target hook.
16747 * targhooks.c (default_split_reduction): New function.
16748 * targhooks.h (default_split_reduction): Declare.
16749 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16750 target requests first reduce vectors by combining low and high
16752 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16753 (get_vectype_for_scalar_type_and_size): Export.
16754 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16755 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16756 * doc/tm.texi: Regenerate.
16757 * config/i386/i386.c (ix86_split_reduction): Implement
16758 TARGET_VECTORIZE_SPLIT_REDUCTION.
16760 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16763 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16764 in PIC mode except for TARGET_VXWORKS_RTP.
16765 * config/sparc/sparc.c: Include cfgrtl.h.
16766 (TARGET_INIT_PIC_REG): Define.
16767 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16768 (sparc_pic_register_p): New predicate.
16769 (sparc_legitimate_address_p): Use it.
16770 (sparc_legitimize_pic_address): Likewise.
16771 (sparc_delegitimize_address): Likewise.
16772 (sparc_mode_dependent_address_p): Likewise.
16773 (gen_load_pcrel_sym): Remove 4th parameter.
16774 (load_got_register): Adjust call to above. Remove obsolete stuff.
16775 (sparc_expand_prologue): Do not call load_got_register here.
16776 (sparc_flat_expand_prologue): Likewise.
16777 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16778 (sparc_use_pseudo_pic_reg): New function.
16779 (sparc_init_pic_reg): Likewise.
16780 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16781 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16783 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16785 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16786 Add item for branch_cost.
16788 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16790 PR rtl-optimization/83565
16791 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16792 not extend the result to a larger mode for rotate operations.
16793 (num_sign_bit_copies1): Likewise.
16795 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16798 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16800 Use values-Xc.o for -pedantic.
16801 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16803 2018-01-12 Martin Liska <mliska@suse.cz>
16806 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16808 (possible_polymorphic_call_targets): Use it.
16809 (ipa_devirt): Likewise.
16811 2018-01-12 Martin Liska <mliska@suse.cz>
16813 * profile-count.h (enum profile_quality): Use 0 as invalid
16814 enum value of profile_quality.
16816 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16818 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16819 -mext-string options.
16821 2018-01-12 Richard Biener <rguenther@suse.de>
16823 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16824 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16825 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16827 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16829 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16831 * configure.ac (--with-long-double-format): Add support for the
16832 configuration option to change the default long double format on
16834 * config.gcc (powerpc*-linux*-*): Likewise.
16835 * configure: Regenerate.
16836 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16837 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16838 used without modification.
16840 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16842 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16843 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16844 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16845 MISC_BUILTIN_SPEC_BARRIER.
16846 (rs6000_init_builtins): Likewise.
16847 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16849 (speculation_barrier): New define_insn.
16850 * doc/extend.texi: Document __builtin_speculation_barrier.
16852 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16855 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16856 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16857 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16859 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16860 integral modes instead of "ss" and "sd".
16861 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16862 vectors with 32-bit and 64-bit elements.
16863 (vecdupssescalarmodesuffix): New mode attribute.
16864 (vec_dup<mode>): Use it.
16866 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16869 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16870 frame if argument is passed on stack.
16872 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16875 * ree.c (combine_reaching_defs): Optimize also
16876 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16877 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16879 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16881 PR middle-end/83189
16882 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16884 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16886 PR middle-end/83718
16887 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16888 after they are computed.
16890 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16892 PR tree-optimization/83695
16893 * gimple-loop-linterchange.cc
16894 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16895 reset cached scev information after interchange.
16896 (pass_linterchange::execute): Remove call to scev_reset_htab.
16898 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16900 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16901 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16902 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16903 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16904 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16905 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16906 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16907 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16908 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16909 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16910 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16911 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16912 (V_lane_reg): Likewise.
16913 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16915 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16916 (vfmal_lane_low<mode>_intrinsic,
16917 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16918 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16919 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16920 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16921 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16922 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16924 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16926 * config/arm/arm-cpus.in (fp16fml): New feature.
16927 (ALL_SIMD): Add fp16fml.
16928 (armv8.2-a): Add fp16fml as an option.
16929 (armv8.3-a): Likewise.
16930 (armv8.4-a): Add fp16fml as part of fp16.
16931 * config/arm/arm.h (TARGET_FP16FML): Define.
16932 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16934 * config/arm/arm-modes.def (V2HF): Define.
16935 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16936 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16937 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16938 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16939 vfmsl_low, vfmsl_high): New set of builtins.
16940 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16941 (vfml_op): New code attribute.
16942 (VFMLHALVES): New int iterator.
16943 (VFML, VFMLSEL): New mode attributes.
16944 (V_reg): Define mapping for V2HF.
16945 (V_hi, V_lo): New mode attributes.
16946 (VF_constraint): Likewise.
16947 (vfml_half, vfml_half_selector): New int attributes.
16948 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16950 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16951 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16953 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16954 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16955 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16956 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16958 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16959 Document new effective target and option set.
16961 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16963 * config/arm/arm-cpus.in (armv8_4): New feature.
16964 (ARMv8_4a): New fgroup.
16965 (armv8.4-a): New arch.
16966 * config/arm/arm-tables.opt: Regenerate.
16967 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16968 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16969 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16970 Add matching rules for -march=armv8.4-a and extensions.
16971 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16973 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16976 * config/rx/rx.md (BW): New mode attribute.
16977 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16979 2018-01-11 Richard Biener <rguenther@suse.de>
16981 PR tree-optimization/83435
16982 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16983 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16984 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16986 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16987 Alan Hayward <alan.hayward@arm.com>
16988 David Sherwood <david.sherwood@arm.com>
16990 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16992 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16993 (aarch64_print_address_internal): Use it to check for a zero offset.
16995 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16996 Alan Hayward <alan.hayward@arm.com>
16997 David Sherwood <david.sherwood@arm.com>
16999 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17000 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17001 Return a poly_int64 rather than a HOST_WIDE_INT.
17002 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17003 rather than a HOST_WIDE_INT.
17004 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17005 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17006 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17007 final_offset from HOST_WIDE_INT to poly_int64.
17008 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17009 to_constant when getting the number of units in an Advanced SIMD
17011 (aarch64_builtin_vectorized_function): Check for a constant number
17013 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17015 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17016 attribute instead of GET_MODE_NUNITS.
17017 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17018 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17019 GET_MODE_SIZE for fixed-size registers.
17020 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17021 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17022 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17023 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17024 (aarch64_print_operand, aarch64_print_address_internal)
17025 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17026 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17027 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17028 Handle polynomial GET_MODE_SIZE.
17029 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17030 wider than SImode without modification.
17031 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17032 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17033 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17034 passing and returning SVE modes.
17035 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17036 rather than GEN_INT.
17037 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17038 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17039 (aarch64_allocate_and_probe_stack_space): Likewise.
17040 (aarch64_layout_frame): Cope with polynomial offsets.
17041 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17042 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17043 polynomial offsets.
17044 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17045 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17046 poly_int64 rather than a HOST_WIDE_INT.
17047 (aarch64_get_separate_components, aarch64_process_components)
17048 (aarch64_expand_prologue, aarch64_expand_epilogue)
17049 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17050 (aarch64_anchor_offset): New function, split out from...
17051 (aarch64_legitimize_address): ...here.
17052 (aarch64_builtin_vectorization_cost): Handle polynomial
17053 TYPE_VECTOR_SUBPARTS.
17054 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17056 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17057 number of elements from the PARALLEL rather than the mode.
17058 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17059 rather than GET_MODE_BITSIZE.
17060 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17061 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17062 (aarch64_expand_vec_perm_const_1): Handle polynomial
17063 d->perm.length () and d->perm elements.
17064 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17065 Apply to_constant to d->perm elements.
17066 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17067 polynomial CONST_VECTOR_NUNITS.
17068 (aarch64_move_pointer): Take amount as a poly_int64 rather
17070 (aarch64_progress_pointer): Avoid temporary variable.
17071 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17072 the mode attribute instead of GET_MODE.
17074 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17075 Alan Hayward <alan.hayward@arm.com>
17076 David Sherwood <david.sherwood@arm.com>
17078 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17079 x exists before using it.
17080 (aarch64_add_constant_internal): Rename to...
17081 (aarch64_add_offset_1): ...this. Replace regnum with separate
17082 src and dest rtxes. Handle the case in which they're different,
17083 including when the offset is zero. Replace scratchreg with an rtx.
17084 Use 2 additions if there is no spare register into which we can
17085 move a 16-bit constant.
17086 (aarch64_add_constant): Delete.
17087 (aarch64_add_offset): Replace reg with separate src and dest
17088 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17089 Use aarch64_add_offset_1.
17090 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17091 an rtx rather than an int. Take the delta as a poly_int64
17092 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17093 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17094 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17095 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17096 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17097 and aarch64_add_sp.
17098 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17099 aarch64_add_constant.
17101 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17103 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17104 Use scalar_float_mode.
17106 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17108 * config/aarch64/aarch64-simd.md
17109 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17110 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17111 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17112 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17113 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17114 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17115 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17116 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17117 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17118 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17120 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17123 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17124 targ_options->x_arm_arch_string is non NULL.
17126 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17128 * config/aarch64/aarch64.h
17129 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17131 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17134 * expmed.c (emit_store_flag_force): Swap if const op0
17135 and change VOIDmode to mode of op0.
17137 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17139 PR rtl-optimization/83761
17140 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17141 than bytes to mode_for_size.
17143 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17145 PR middle-end/83189
17146 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17147 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17150 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17152 PR middle-end/83575
17153 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17154 when in layout mode.
17155 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17156 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17159 2018-01-10 Michael Collison <michael.collison@arm.com>
17161 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17162 * config/aarch64/aarch64-option-extension.def: Add
17163 AARCH64_OPT_EXTENSION of 'fp16fml'.
17164 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17165 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17166 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17167 * config/aarch64/constraints.md (Ui7): New constraint.
17168 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17169 (VFMLA_SEL_W): Ditto.
17172 (VFMLA16_LOW): New int iterator.
17173 (VFMLA16_HIGH): Ditto.
17174 (UNSPEC_FMLAL): New unspec.
17175 (UNSPEC_FMLSL): Ditto.
17176 (UNSPEC_FMLAL2): Ditto.
17177 (UNSPEC_FMLSL2): Ditto.
17178 (f16mac): New code attribute.
17179 * config/aarch64/aarch64-simd-builtins.def
17180 (aarch64_fmlal_lowv2sf): Ditto.
17181 (aarch64_fmlsl_lowv2sf): Ditto.
17182 (aarch64_fmlalq_lowv4sf): Ditto.
17183 (aarch64_fmlslq_lowv4sf): Ditto.
17184 (aarch64_fmlal_highv2sf): Ditto.
17185 (aarch64_fmlsl_highv2sf): Ditto.
17186 (aarch64_fmlalq_highv4sf): Ditto.
17187 (aarch64_fmlslq_highv4sf): Ditto.
17188 (aarch64_fmlal_lane_lowv2sf): Ditto.
17189 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17190 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17191 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17192 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17193 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17194 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17195 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17196 (aarch64_fmlal_lane_highv2sf): Ditto.
17197 (aarch64_fmlsl_lane_highv2sf): Ditto.
17198 (aarch64_fmlal_laneq_highv2sf): Ditto.
17199 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17200 (aarch64_fmlalq_lane_highv4sf): Ditto.
17201 (aarch64_fmlsl_lane_highv4sf): Ditto.
17202 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17203 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17204 * config/aarch64/aarch64-simd.md:
17205 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17206 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17207 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17208 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17209 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17210 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17211 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17212 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17213 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17214 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17215 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17216 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17217 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17218 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17219 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17220 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17221 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17222 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17223 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17224 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17225 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17226 (vfmlsl_low_u32): Ditto.
17227 (vfmlalq_low_u32): Ditto.
17228 (vfmlslq_low_u32): Ditto.
17229 (vfmlal_high_u32): Ditto.
17230 (vfmlsl_high_u32): Ditto.
17231 (vfmlalq_high_u32): Ditto.
17232 (vfmlslq_high_u32): Ditto.
17233 (vfmlal_lane_low_u32): Ditto.
17234 (vfmlsl_lane_low_u32): Ditto.
17235 (vfmlal_laneq_low_u32): Ditto.
17236 (vfmlsl_laneq_low_u32): Ditto.
17237 (vfmlalq_lane_low_u32): Ditto.
17238 (vfmlslq_lane_low_u32): Ditto.
17239 (vfmlalq_laneq_low_u32): Ditto.
17240 (vfmlslq_laneq_low_u32): Ditto.
17241 (vfmlal_lane_high_u32): Ditto.
17242 (vfmlsl_lane_high_u32): Ditto.
17243 (vfmlal_laneq_high_u32): Ditto.
17244 (vfmlsl_laneq_high_u32): Ditto.
17245 (vfmlalq_lane_high_u32): Ditto.
17246 (vfmlslq_lane_high_u32): Ditto.
17247 (vfmlalq_laneq_high_u32): Ditto.
17248 (vfmlslq_laneq_high_u32): Ditto.
17249 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17250 (AARCH64_FL_FOR_ARCH8_4): New.
17251 (AARCH64_ISA_F16FML): New ISA flag.
17252 (TARGET_F16FML): New feature flag for fp16fml.
17253 (doc/invoke.texi): Document new fp16fml option.
17255 2018-01-10 Michael Collison <michael.collison@arm.com>
17257 * config/aarch64/aarch64-builtins.c:
17258 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17259 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17260 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17261 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17262 (AARCH64_ISA_SHA3): New ISA flag.
17263 (TARGET_SHA3): New feature flag for sha3.
17264 * config/aarch64/iterators.md (sha512_op): New int attribute.
17265 (CRYPTO_SHA512): New int iterator.
17266 (UNSPEC_SHA512H): New unspec.
17267 (UNSPEC_SHA512H2): Ditto.
17268 (UNSPEC_SHA512SU0): Ditto.
17269 (UNSPEC_SHA512SU1): Ditto.
17270 * config/aarch64/aarch64-simd-builtins.def
17271 (aarch64_crypto_sha512hqv2di): New builtin.
17272 (aarch64_crypto_sha512h2qv2di): Ditto.
17273 (aarch64_crypto_sha512su0qv2di): Ditto.
17274 (aarch64_crypto_sha512su1qv2di): Ditto.
17275 (aarch64_eor3qv8hi): Ditto.
17276 (aarch64_rax1qv2di): Ditto.
17277 (aarch64_xarqv2di): Ditto.
17278 (aarch64_bcaxqv8hi): Ditto.
17279 * config/aarch64/aarch64-simd.md:
17280 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17281 (aarch64_crypto_sha512su0qv2di): Ditto.
17282 (aarch64_crypto_sha512su1qv2di): Ditto.
17283 (aarch64_eor3qv8hi): Ditto.
17284 (aarch64_rax1qv2di): Ditto.
17285 (aarch64_xarqv2di): Ditto.
17286 (aarch64_bcaxqv8hi): Ditto.
17287 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17288 (vsha512h2q_u64): Ditto.
17289 (vsha512su0q_u64): Ditto.
17290 (vsha512su1q_u64): Ditto.
17291 (veor3q_u16): Ditto.
17292 (vrax1q_u64): Ditto.
17293 (vxarq_u64): Ditto.
17294 (vbcaxq_u16): Ditto.
17295 * config/arm/types.md (crypto_sha512): New type attribute.
17296 (crypto_sha3): Ditto.
17297 (doc/invoke.texi): Document new sha3 option.
17299 2018-01-10 Michael Collison <michael.collison@arm.com>
17301 * config/aarch64/aarch64-builtins.c:
17302 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17303 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17304 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17305 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17306 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17307 (AARCH64_ISA_SM4): New ISA flag.
17308 (TARGET_SM4): New feature flag for sm4.
17309 * config/aarch64/aarch64-simd-builtins.def
17310 (aarch64_sm3ss1qv4si): Ditto.
17311 (aarch64_sm3tt1aq4si): Ditto.
17312 (aarch64_sm3tt1bq4si): Ditto.
17313 (aarch64_sm3tt2aq4si): Ditto.
17314 (aarch64_sm3tt2bq4si): Ditto.
17315 (aarch64_sm3partw1qv4si): Ditto.
17316 (aarch64_sm3partw2qv4si): Ditto.
17317 (aarch64_sm4eqv4si): Ditto.
17318 (aarch64_sm4ekeyqv4si): Ditto.
17319 * config/aarch64/aarch64-simd.md:
17320 (aarch64_sm3ss1qv4si): Ditto.
17321 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17322 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17323 (aarch64_sm4eqv4si): Ditto.
17324 (aarch64_sm4ekeyqv4si): Ditto.
17325 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17326 (sm3part_op): Ditto.
17327 (CRYPTO_SM3TT): Ditto.
17328 (CRYPTO_SM3PART): Ditto.
17329 (UNSPEC_SM3SS1): New unspec.
17330 (UNSPEC_SM3TT1A): Ditto.
17331 (UNSPEC_SM3TT1B): Ditto.
17332 (UNSPEC_SM3TT2A): Ditto.
17333 (UNSPEC_SM3TT2B): Ditto.
17334 (UNSPEC_SM3PARTW1): Ditto.
17335 (UNSPEC_SM3PARTW2): Ditto.
17336 (UNSPEC_SM4E): Ditto.
17337 (UNSPEC_SM4EKEY): Ditto.
17338 * config/aarch64/constraints.md (Ui2): New constraint.
17339 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17340 * config/arm/types.md (crypto_sm3): New type attribute.
17341 (crypto_sm4): Ditto.
17342 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17343 (vsm3tt1aq_u32): Ditto.
17344 (vsm3tt1bq_u32): Ditto.
17345 (vsm3tt2aq_u32): Ditto.
17346 (vsm3tt2bq_u32): Ditto.
17347 (vsm3partw1q_u32): Ditto.
17348 (vsm3partw2q_u32): Ditto.
17349 (vsm4eq_u32): Ditto.
17350 (vsm4ekeyq_u32): Ditto.
17351 (doc/invoke.texi): Document new sm4 option.
17353 2018-01-10 Michael Collison <michael.collison@arm.com>
17355 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17356 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17357 (AARCH64_FL_FOR_ARCH8_4): New.
17358 (AARCH64_FL_V8_4): New flag.
17359 (doc/invoke.texi): Document new armv8.4-a option.
17361 2018-01-10 Michael Collison <michael.collison@arm.com>
17363 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17364 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17365 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17366 * config/aarch64/aarch64-option-extension.def: Add
17367 AARCH64_OPT_EXTENSION of 'sha2'.
17368 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17369 (crypto): Disable sha2 and aes if crypto disabled.
17370 (crypto): Enable aes and sha2 if enabled.
17371 (simd): Disable sha2 and aes if simd disabled.
17372 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17374 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17375 (TARGET_SHA2): New feature flag for sha2.
17376 (TARGET_AES): New feature flag for aes.
17377 * config/aarch64/aarch64-simd.md:
17378 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17379 conditional on TARGET_AES.
17380 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17381 (aarch64_crypto_sha1hsi): Make pattern conditional
17383 (aarch64_crypto_sha1hv4si): Ditto.
17384 (aarch64_be_crypto_sha1hv4si): Ditto.
17385 (aarch64_crypto_sha1su1v4si): Ditto.
17386 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17387 (aarch64_crypto_sha1su0v4si): Ditto.
17388 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17389 (aarch64_crypto_sha256su0v4si): Ditto.
17390 (aarch64_crypto_sha256su1v4si): Ditto.
17391 (doc/invoke.texi): Document new aes and sha2 options.
17393 2018-01-10 Martin Sebor <msebor@redhat.com>
17395 PR tree-optimization/83781
17396 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17399 2018-01-11 Martin Sebor <msebor@gmail.com>
17400 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17402 PR tree-optimization/83501
17403 PR tree-optimization/81703
17405 * tree-ssa-strlen.c (get_string_cst): Rename...
17406 (get_string_len): ...to this. Handle global constants.
17407 (handle_char_store): Adjust.
17409 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17410 Jim Wilson <jimw@sifive.com>
17412 * config/riscv/riscv-protos.h (riscv_output_return): New.
17413 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17414 (riscv_attribute_table, riscv_output_return),
17415 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17416 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17417 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17419 (riscv_expand_prologue): Add early return for naked function.
17420 (riscv_expand_epilogue): Likewise.
17421 (riscv_function_ok_for_sibcall): Return false for naked function.
17422 (riscv_set_current_function): New.
17423 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17424 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17425 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17426 * doc/extend.texi (RISC-V Function Attributes): New.
17428 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17430 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17431 check for 128-bit long double before checking TCmode.
17432 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17433 128-bit long doubles before checking TFmode or TCmode.
17434 (FLOAT128_IBM_P): Likewise.
17436 2018-01-10 Martin Sebor <msebor@redhat.com>
17438 PR tree-optimization/83671
17439 * builtins.c (c_strlen): Unconditionally return zero for the empty
17441 Use -Warray-bounds for warnings.
17442 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17443 for non-constant array indices with COMPONENT_REF, arrays of
17444 arrays, and pointers to arrays.
17445 (gimple_fold_builtin_strlen): Determine and set length range for
17446 non-constant character arrays.
17448 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17450 PR middle-end/81897
17451 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17454 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17456 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17458 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17461 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17462 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17463 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17464 indexed_or_indirect_operand predicate.
17465 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17466 (*vsx_le_perm_load_v8hi): Likewise.
17467 (*vsx_le_perm_load_v16qi): Likewise.
17468 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17469 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17470 (*vsx_le_perm_store_v8hi): Likewise.
17471 (*vsx_le_perm_store_v16qi): Likewise.
17472 (eight unnamed splitters): Likewise.
17474 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17476 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17477 * config/rs6000/emmintrin.h: Likewise.
17478 * config/rs6000/mmintrin.h: Likewise.
17479 * config/rs6000/xmmintrin.h: Likewise.
17481 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17484 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17486 * tree.c (tree_nop_conversion): Return true for location wrapper
17488 (maybe_wrap_with_location): New function.
17489 (selftest::check_strip_nops): New function.
17490 (selftest::test_location_wrappers): New function.
17491 (selftest::tree_c_tests): Call it.
17492 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17493 (maybe_wrap_with_location): New decl.
17494 (EXPR_LOCATION_WRAPPER_P): New macro.
17495 (location_wrapper_p): New inline function.
17496 (tree_strip_any_location_wrapper): New inline function.
17498 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17501 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17502 stack_realign_offset for the largest alignment of stack slot
17504 (ix86_find_max_used_stack_alignment): New function.
17505 (ix86_finalize_stack_frame_flags): Use it. Set
17506 max_used_stack_alignment if we don't realign stack.
17507 * config/i386/i386.h (machine_function): Add
17508 max_used_stack_alignment.
17510 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17512 * config/arm/arm.opt (-mbranch-cost): New option.
17513 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17516 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17519 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17520 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17522 2018-01-10 Richard Biener <rguenther@suse.de>
17525 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17526 early out so it also covers the case where we have a non-NULL
17529 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17531 PR tree-optimization/83753
17532 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17533 for non-strided grouped accesses if the number of elements is 1.
17535 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17538 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17539 * i386.h (TARGET_USE_GATHER): Define.
17540 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17542 2018-01-10 Martin Liska <mliska@suse.cz>
17545 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17546 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17548 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17549 CLEANUP_NO_PARTITIONING is not set.
17551 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17553 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17554 for vectors, as a partial revert of r254296.
17555 * rtl.h (const_vec_p): Delete.
17556 (const_vec_duplicate_p): Don't test for vector CONSTs.
17557 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17558 * expmed.c (make_tree): Likewise.
17561 * common.md (E, F): Use CONSTANT_P instead of checking for
17563 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17564 checking for CONST_VECTOR.
17566 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17568 PR middle-end/83575
17569 * predict.c (force_edge_cold): Handle in more sane way edges
17570 with no prediction.
17572 2018-01-09 Carl Love <cel@us.ibm.com>
17574 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17576 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17577 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17578 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17579 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17580 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17581 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17582 * config/rs6000/rs6000-protos.h: Add extern defition for
17583 rs6000_generate_float2_double_code.
17584 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17586 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17587 (float2_v2df): Add define_expand.
17589 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17592 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17593 op_mode in the force_to_mode call.
17595 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17597 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17598 instead of checking each element individually.
17599 (aarch64_evpc_uzp): Likewise.
17600 (aarch64_evpc_zip): Likewise.
17601 (aarch64_evpc_ext): Likewise.
17602 (aarch64_evpc_rev): Likewise.
17603 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17604 instead of checking each element individually. Return true without
17606 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17607 whether all selected elements come from the same input, instead of
17608 checking each element individually. Remove calls to gen_rtx_REG,
17609 start_sequence and end_sequence and instead assert that no rtl is
17612 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17614 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17615 order of HIGH and CONST checks.
17617 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17619 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17620 if the destination isn't an SSA_NAME.
17622 2018-01-09 Richard Biener <rguenther@suse.de>
17624 PR tree-optimization/83668
17625 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17627 (canonicalize_loop_form): ... here, renamed from ...
17628 (canonicalize_loop_closed_ssa_form): ... this and amended to
17629 swap successor edges for loop exit blocks to make us use
17630 the RPO order we need for initial schedule generation.
17632 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17634 PR tree-optimization/64811
17635 * match.pd: When optimizing comparisons with Inf, avoid
17636 introducing or losing exceptions from comparisons with NaN.
17638 2018-01-09 Martin Liska <mliska@suse.cz>
17641 * asan.c (shadow_mem_size): Add gcc_assert.
17643 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17645 Don't save registers in main().
17648 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17649 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17650 * config/avr/avr.c (avr_set_current_function): Don't error if
17651 naked, OS_task or OS_main are specified at the same time.
17652 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17654 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17656 * common/config/avr/avr-common.c (avr_option_optimization_table):
17657 Switch on -mmain-is-OS_task for optimizing compilations.
17659 2018-01-09 Richard Biener <rguenther@suse.de>
17661 PR tree-optimization/83572
17662 * graphite.c: Include cfganal.h.
17663 (graphite_transform_loops): Connect infinite loops to exit
17664 and remove fake edges at the end.
17666 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17668 * ipa-inline.c (edge_badness): Revert accidental checkin.
17670 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17673 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17674 symbols; not inline clones.
17676 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17679 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17680 hard registers. Formatting fixes.
17682 PR preprocessor/83722
17683 * gcc.c (try_generate_repro): Pass
17684 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17685 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17688 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17689 Kito Cheng <kito.cheng@gmail.com>
17691 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17692 (riscv_leaf_function_p): Delete.
17693 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17695 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17697 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17699 (do_ifelse): New function.
17700 (do_isel): New function.
17701 (do_sub3): New function.
17702 (do_add3): New function.
17703 (do_load_mask_compare): New function.
17704 (do_overlap_load_compare): New function.
17705 (expand_compare_loop): New function.
17706 (expand_block_compare): Call expand_compare_loop() when appropriate.
17707 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17708 option description.
17709 (-mblock-compare-inline-loop-limit): New option.
17711 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17714 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17715 Reverse order of second and third operands in first alternative.
17716 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17717 of first and second elements in UNSPEC_VPERMR vector.
17718 (altivec_expand_vec_perm_le): Likewise.
17720 2018-01-08 Jeff Law <law@redhat.com>
17722 PR rtl-optimizatin/81308
17723 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17724 (process_switch): If group_case_labels makes a change, then set
17726 (pass_convert_switch::execute): If a switch is converted, then
17727 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17729 PR rtl-optimization/81308
17730 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17733 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17735 PR target/83663 - Revert r255946
17736 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17737 generation for cases where splatting a value is not useful.
17738 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17739 across a vec_duplicate and a paradoxical subreg forming a vector
17740 mode to a vec_concat.
17742 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17744 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17745 -march=armv8.3-a variants.
17746 * config/arm/t-multilib: Likewise.
17747 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17749 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17751 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17753 (cceq_ior_compare_complement): Give it a name so I can use it, and
17754 change boolean_or_operator predicate to boolean_operator so it can
17755 be used to generate a crand.
17756 (eqne): New code iterator.
17757 (bd/bd_neg): New code_attrs.
17758 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17759 a single define_insn.
17760 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17761 decrement (bdnzt/bdnzf/bdzt/bdzf).
17762 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17763 with the new names of the branch decrement patterns, and added the
17764 names of the branch decrement conditional patterns.
17766 2018-01-08 Richard Biener <rguenther@suse.de>
17768 PR tree-optimization/83563
17769 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17772 2018-01-08 Richard Biener <rguenther@suse.de>
17774 PR middle-end/83713
17775 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17777 2018-01-08 Richard Biener <rguenther@suse.de>
17779 PR tree-optimization/83685
17780 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17781 references to abnormals.
17783 2018-01-08 Richard Biener <rguenther@suse.de>
17786 * dwarf2out.c (output_indirect_strings): Handle empty
17787 skeleton_debug_str_hash.
17788 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17790 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17792 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17793 (emit_store_direct): Likewise.
17794 (arc_trampoline_adjust_address): Likewise.
17795 (arc_asm_trampoline_template): New function.
17796 (arc_initialize_trampoline): Use asm_trampoline_template.
17797 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17798 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17799 * config/arc/arc.md (flush_icache): Delete pattern.
17801 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17803 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17804 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17807 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17810 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17811 by not USED_FOR_TARGET.
17812 (make_pass_resolve_sw_modes): Likewise.
17814 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17816 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17819 2018-01-08 Richard Biener <rguenther@suse.de>
17821 PR middle-end/83580
17822 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17824 2018-01-08 Richard Biener <rguenther@suse.de>
17826 PR middle-end/83517
17827 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17829 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17831 PR middle-end/81897
17832 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17833 basic blocks with a small number of successors.
17834 (convert_control_dep_chain_into_preds): Improve handling of
17836 (dump_predicates): Split apart into...
17837 (dump_pred_chain): ...here...
17838 (dump_pred_info): ...and here.
17839 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17840 (can_chain_union_be_invalidated_p): Improve check for invalidation
17842 (uninit_uses_cannot_happen): Avoid unnecessary if
17843 convert_control_dep_chain_into_preds yielded nothing.
17845 2018-01-06 Martin Sebor <msebor@redhat.com>
17847 PR tree-optimization/83640
17848 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17849 subtracting negative offset from size.
17850 (builtin_access::overlap): Adjust offset bounds of the access to fall
17851 within the size of the object if possible.
17853 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17855 PR rtl-optimization/83699
17856 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17857 extract_bit_field_as_subreg to cases in which the extracted
17858 value is also a vector.
17860 * lra-constraints.c (process_alt_operands): Test for the equivalence
17861 substitutions when detecting a possible reload cycle.
17863 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17866 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17867 by default if flag_selective_schedling{,2}. Formatting fixes.
17869 PR rtl-optimization/83682
17870 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17871 if it has non-VECTOR_MODE element mode.
17872 (vec_duplicate_p): Likewise.
17874 PR middle-end/83694
17875 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17876 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17878 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17881 * config/i386/i386-builtin.def
17882 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17883 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17884 Require also OPTION_MASK_ISA_AVX512F in addition to
17885 OPTION_MASK_ISA_GFNI.
17886 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17887 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17888 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17889 to OPTION_MASK_ISA_GFNI.
17890 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17891 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17892 OPTION_MASK_ISA_AVX512BW.
17893 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17894 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17895 addition to OPTION_MASK_ISA_GFNI.
17896 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17897 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17898 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17899 to OPTION_MASK_ISA_GFNI.
17900 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17901 a requirement for all ISAs rather than any of them with a few
17903 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17905 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17906 bitmasks to be enabled with 3 exceptions, instead of requiring any
17907 enabled ISA with lots of exceptions.
17908 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17909 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17910 Change avx512bw in isa attribute to avx512f.
17911 * config/i386/sgxintrin.h: Add license boilerplate.
17912 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17913 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17914 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17915 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17917 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17918 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17919 temporarily sse2 rather than sse if not enabled already.
17922 * config/i386/sse.md (VI248_VLBW): Rename to ...
17923 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17924 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17925 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17926 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17927 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17928 mode iterator instead of VI248_VLBW.
17930 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17932 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17933 (record_modified): Skip clobbers; add debug output.
17934 (param_change_prob): Use sreal frequencies.
17936 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17938 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17939 punt for user-aligned variables.
17941 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17943 * tree-chrec.c (chrec_contains_symbols): Return true for
17946 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17949 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17950 of (x|y) == x for BICS pattern.
17952 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17954 PR tree-optimization/83605
17955 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17956 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17959 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17961 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17962 * config/epiphany/rtems.h: New file.
17964 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17965 Uros Bizjak <ubizjak@gmail.com>
17968 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17969 QIreg_operand instead of register_operand predicate.
17970 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17971 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17972 comments instead of -fmitigate[-_]rop.
17974 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17977 * cgraphunit.c (symbol_table::compile): Switch to text_section
17978 before calling assembly_start debug hook.
17979 * run-rtl-passes.c (run_rtl_passes): Likewise.
17982 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17984 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17985 range_int_cst_p rather than !symbolic_range_p before calling
17986 extract_range_from_multiplicative_op_1.
17988 2018-01-04 Jeff Law <law@redhat.com>
17990 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17991 redundant test in assertion.
17993 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17995 * doc/rtl.texi: Document machine_mode wrapper classes.
17997 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17999 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18000 using tree_to_uhwi.
18002 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18004 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18005 the VEC_PERM_EXPR fold to fail.
18007 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18010 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18011 to switched_sections.
18013 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18016 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18017 test for d.testing.
18019 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18022 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18023 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18025 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18028 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18029 is BLKmode and bitpos not zero or mode change is needed.
18031 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18034 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18037 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18040 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18041 instead of MULT rtx. Update all corresponding splitters.
18042 (*saddl_se): Ditto.
18043 (*ssub<modesuffix>): Ditto.
18044 (*ssubl_se): Ditto.
18045 (*cmp_sadd_di): Update split patterns.
18046 (*cmp_sadd_si): Ditto.
18047 (*cmp_sadd_sidi): Ditto.
18048 (*cmp_ssub_di): Ditto.
18049 (*cmp_ssub_si): Ditto.
18050 (*cmp_ssub_sidi): Ditto.
18051 * config/alpha/predicates.md (const23_operand): New predicate.
18052 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18053 Look for ASHIFT, not MULT inner operand.
18054 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18056 2018-01-04 Martin Liska <mliska@suse.cz>
18058 PR gcov-profile/83669
18059 * gcov.c (output_intermediate_file): Add version to intermediate
18061 * doc/gcov.texi: Document new field 'version' in intermediate
18062 file format. Fix location of '-k' option of gcov command.
18064 2018-01-04 Martin Liska <mliska@suse.cz>
18067 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18069 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18071 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18073 2018-01-03 Martin Sebor <msebor@redhat.com>
18075 PR tree-optimization/83655
18076 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18077 checking calls with invalid arguments.
18079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18081 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18082 (vectorizable_mask_load_store): Delete.
18083 (vectorizable_call): Return false for masked loads and stores.
18084 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18085 instead of gimple_assign_rhs1.
18086 (vectorizable_load): Handle IFN_MASK_LOAD.
18087 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18091 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18093 (vectorizable_mask_load_store): ...here.
18094 (vectorizable_load): ...and here.
18096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18098 * tree-vect-stmts.c (vect_build_all_ones_mask)
18099 (vect_build_zero_merge_argument): New functions, split out from...
18100 (vectorizable_load): ...here.
18102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18104 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18106 (vectorizable_mask_load_store): ...here.
18107 (vectorizable_store): ...and here.
18109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18111 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18113 (vectorizable_mask_load_store): ...here.
18115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18117 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18118 (vect_model_store_cost): Take a vec_load_store_type instead of a
18120 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18121 (vect_model_store_cost): Take a vec_load_store_type instead of a
18123 (vectorizable_mask_load_store): Update accordingly.
18124 (vectorizable_store): Likewise.
18125 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18129 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18130 IFN_MASK_LOAD calls here rather than...
18131 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18134 Alan Hayward <alan.hayward@arm.com>
18135 David Sherwood <david.sherwood@arm.com>
18137 * expmed.c (extract_bit_field_1): For vector extracts,
18138 fall back to extract_bit_field_as_subreg if vec_extract
18141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18142 Alan Hayward <alan.hayward@arm.com>
18143 David Sherwood <david.sherwood@arm.com>
18145 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18146 they are variable or constant sized.
18147 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18148 slots for constant-sized data.
18150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18151 Alan Hayward <alan.hayward@arm.com>
18152 David Sherwood <david.sherwood@arm.com>
18154 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18155 handling COND_EXPRs with boolean comparisons, try to find a better
18156 basis for the mask type than the boolean itself.
18158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18160 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18161 is calculated and how it can be overridden.
18162 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18163 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18165 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18169 Alan Hayward <alan.hayward@arm.com>
18170 David Sherwood <david.sherwood@arm.com>
18172 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18173 Remove the mode argument.
18174 (aarch64_simd_valid_immediate): Remove the mode and inverse
18176 * config/aarch64/iterators.md (bitsize): New iterator.
18177 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18178 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18179 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18180 aarch64_simd_valid_immediate.
18181 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18182 (aarch64_reg_or_bic_imm): Likewise.
18183 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18184 with an insn_type enum and msl with a modifier_type enum.
18185 Replace element_width with a scalar_mode. Change the shift
18186 to unsigned int. Add constructors for scalar_float_mode and
18187 scalar_int_mode elements.
18188 (aarch64_vect_float_const_representable_p): Delete.
18189 (aarch64_can_const_movi_rtx_p)
18190 (aarch64_simd_scalar_immediate_valid_for_move)
18191 (aarch64_simd_make_constant): Update call to
18192 aarch64_simd_valid_immediate.
18193 (aarch64_advsimd_valid_immediate_hs): New function.
18194 (aarch64_advsimd_valid_immediate): Likewise.
18195 (aarch64_simd_valid_immediate): Remove mode and inverse
18196 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18197 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18198 and aarch64_float_const_representable_p on the result.
18199 (aarch64_output_simd_mov_immediate): Remove mode argument.
18200 Update call to aarch64_simd_valid_immediate and use of
18201 simd_immediate_info.
18202 (aarch64_output_scalar_simd_mov_immediate): Update call
18205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18206 Alan Hayward <alan.hayward@arm.com>
18207 David Sherwood <david.sherwood@arm.com>
18209 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18210 (mode_nunits): Likewise CONST_MODE_NUNITS.
18211 * machmode.def (ADJUST_NUNITS): Document.
18212 * genmodes.c (mode_data::need_nunits_adj): New field.
18213 (blank_mode): Update accordingly.
18214 (adj_nunits): New variable.
18215 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18217 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18218 listed in adj_nunits.
18219 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18220 listed in adj_nunits. Don't emit case statements for such modes.
18221 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18222 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18223 nothing if adj_nunits is nonnull.
18224 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18225 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18226 (emit_mode_fbit): Update use of print_maybe_const_decl.
18227 (emit_move_size): Likewise. Treat the array as non-const
18229 (emit_mode_adjustments): Handle adj_nunits.
18231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18233 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18234 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18235 (VECTOR_MODES): Use it.
18236 (make_vector_modes): Take the prefix as an argument.
18238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18239 Alan Hayward <alan.hayward@arm.com>
18240 David Sherwood <david.sherwood@arm.com>
18242 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18243 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18244 for MODE_VECTOR_BOOL.
18245 * machmode.def (VECTOR_BOOL_MODE): Document.
18246 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18247 (make_vector_bool_mode): New function.
18248 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18250 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18251 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18253 * stor-layout.c (int_mode_for_mode): Likewise.
18254 * tree.c (build_vector_type_for_mode): Likewise.
18255 * varasm.c (output_constant_pool_2): Likewise.
18256 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18257 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18258 for MODE_VECTOR_BOOL.
18259 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18260 of mode class checks.
18261 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18262 instead of a list of mode class checks.
18263 (expand_vector_scalar_condition): Likewise.
18264 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18267 Alan Hayward <alan.hayward@arm.com>
18268 David Sherwood <david.sherwood@arm.com>
18270 * machmode.h (mode_size): Change from unsigned short to
18272 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18273 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18274 or if measurement_type is not polynomial.
18275 (fixed_size_mode::includes_p): Check for constant-sized modes.
18276 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18277 return a poly_uint16 rather than an unsigned short.
18278 (emit_mode_size): Change the type of mode_size from unsigned short
18279 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18280 (emit_mode_adjustments): Cope with polynomial vector sizes.
18281 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18283 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18285 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18286 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18287 * caller-save.c (setup_save_areas): Likewise.
18288 (replace_reg_with_saved_mem): Likewise.
18289 * calls.c (emit_library_call_value_1): Likewise.
18290 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18291 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18292 (gen_lowpart_for_combine): Likewise.
18293 * convert.c (convert_to_integer_1): Likewise.
18294 * cse.c (equiv_constant, cse_insn): Likewise.
18295 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18296 (cselib_subst_to_values): Likewise.
18297 * dce.c (word_dce_process_block): Likewise.
18298 * df-problems.c (df_word_lr_mark_ref): Likewise.
18299 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18300 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18301 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18302 (rtl_for_decl_location): Likewise.
18303 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18304 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18305 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18306 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18307 (expand_expr_real_1): Likewise.
18308 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18309 (pad_below): Likewise.
18310 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18311 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18312 * ira.c (get_subreg_tracking_sizes): Likewise.
18313 * ira-build.c (ira_create_allocno_objects): Likewise.
18314 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18315 (ira_sort_regnos_for_alter_reg): Likewise.
18316 * ira-costs.c (record_operand_costs): Likewise.
18317 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18318 (resolve_simple_move): Likewise.
18319 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18320 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18321 (lra_constraints): Likewise.
18322 (CONST_POOL_OK_P): Reject variable-sized modes.
18323 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18324 (add_pseudo_to_slot, lra_spill): Likewise.
18325 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18326 * optabs-query.c (get_best_extraction_insn): Likewise.
18327 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18328 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18329 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18330 * recog.c (offsettable_address_addr_space_p): Likewise.
18331 * regcprop.c (maybe_mode_change): Likewise.
18332 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18333 * regrename.c (build_def_use): Likewise.
18334 * regstat.c (dump_reg_info): Likewise.
18335 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18336 (find_reloads, find_reloads_subreg_address): Likewise.
18337 * reload1.c (eliminate_regs_1): Likewise.
18338 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18339 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18340 (simplify_binary_operation_1, simplify_subreg): Likewise.
18341 * targhooks.c (default_function_arg_padding): Likewise.
18342 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18343 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18344 (verify_gimple_assign_ternary): Likewise.
18345 * tree-inline.c (estimate_move_cost): Likewise.
18346 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18347 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18348 (get_address_cost_ainc): Likewise.
18349 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18350 (vect_supportable_dr_alignment): Likewise.
18351 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18352 (vectorizable_reduction): Likewise.
18353 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18354 (vectorizable_operation, vectorizable_load): Likewise.
18355 * tree.c (build_same_sized_truth_vector_type): Likewise.
18356 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18357 * var-tracking.c (emit_note_insn_var_location): Likewise.
18358 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18359 (ADDR_VEC_ALIGN): Likewise.
18361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18362 Alan Hayward <alan.hayward@arm.com>
18363 David Sherwood <david.sherwood@arm.com>
18365 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18367 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18368 or if measurement_type is polynomial.
18369 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18370 * combine.c (make_extraction): Likewise.
18371 * dse.c (find_shift_sequence): Likewise.
18372 * dwarf2out.c (mem_loc_descriptor): Likewise.
18373 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18374 (extract_bit_field, extract_low_bits): Likewise.
18375 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18376 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18377 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18378 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18379 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18380 * reload.c (find_reloads): Likewise.
18381 * reload1.c (alter_reg): Likewise.
18382 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18383 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18384 * tree-if-conv.c (predicate_mem_writes): Likewise.
18385 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18386 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18387 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18388 * valtrack.c (dead_debug_insert_temp): Likewise.
18389 * varasm.c (mergeable_constant_section): Likewise.
18390 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18393 Alan Hayward <alan.hayward@arm.com>
18394 David Sherwood <david.sherwood@arm.com>
18396 * expr.c (expand_assignment): Cope with polynomial mode sizes
18397 when assigning to a CONCAT.
18399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18400 Alan Hayward <alan.hayward@arm.com>
18401 David Sherwood <david.sherwood@arm.com>
18403 * machmode.h (mode_precision): Change from unsigned short to
18405 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18407 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18408 or if measurement_type is not polynomial.
18409 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18410 in which the mode is already known to be a scalar_int_mode.
18411 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18412 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18414 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18415 for GET_MODE_PRECISION.
18416 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18417 for GET_MODE_PRECISION.
18418 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18420 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18421 (expand_field_assignment, make_extraction): Likewise.
18422 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18423 (get_last_value): Likewise.
18424 * convert.c (convert_to_integer_1): Likewise.
18425 * cse.c (cse_insn): Likewise.
18426 * expr.c (expand_expr_real_1): Likewise.
18427 * lra-constraints.c (simplify_operand_subreg): Likewise.
18428 * optabs-query.c (can_atomic_load_p): Likewise.
18429 * optabs.c (expand_atomic_load): Likewise.
18430 (expand_atomic_store): Likewise.
18431 * ree.c (combine_reaching_defs): Likewise.
18432 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18433 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18434 * tree.h (type_has_mode_precision_p): Likewise.
18435 * ubsan.c (instrument_si_overflow): Likewise.
18437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18438 Alan Hayward <alan.hayward@arm.com>
18439 David Sherwood <david.sherwood@arm.com>
18441 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18442 polynomial numbers of units.
18443 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18444 (valid_vector_subparts_p): New function.
18445 (build_vector_type): Remove temporary shim and take the number
18446 of units as a poly_uint64 rather than an int.
18447 (build_opaque_vector_type): Take the number of units as a
18448 poly_uint64 rather than an int.
18449 * tree.c (build_vector_from_ctor): Handle polynomial
18450 TYPE_VECTOR_SUBPARTS.
18451 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18452 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18453 (build_vector_from_val): If the number of units is variable,
18454 use build_vec_duplicate_cst for constant operands and
18455 VEC_DUPLICATE_EXPR otherwise.
18456 (make_vector_type): Remove temporary is_constant ().
18457 (build_vector_type, build_opaque_vector_type): Take the number of
18458 units as a poly_uint64 rather than an int.
18459 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18461 * cfgexpand.c (expand_debug_expr): Likewise.
18462 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18463 (store_constructor, expand_expr_real_1): Likewise.
18464 (const_scalar_mask_from_tree): Likewise.
18465 * fold-const-call.c (fold_const_reduction): Likewise.
18466 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18467 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18468 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18469 (fold_relational_const): Likewise.
18470 (native_interpret_vector): Likewise. Change the size from an
18471 int to an unsigned int.
18472 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18473 TYPE_VECTOR_SUBPARTS.
18474 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18475 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18476 duplicating a non-constant operand into a variable-length vector.
18477 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18478 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18479 * ipa-icf.c (sem_variable::equals): Likewise.
18480 * match.pd: Likewise.
18481 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18482 * print-tree.c (print_node): Likewise.
18483 * stor-layout.c (layout_type): Likewise.
18484 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18485 * tree-cfg.c (verify_gimple_comparison): Likewise.
18486 (verify_gimple_assign_binary): Likewise.
18487 (verify_gimple_assign_ternary): Likewise.
18488 (verify_gimple_assign_single): Likewise.
18489 * tree-pretty-print.c (dump_generic_node): Likewise.
18490 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18491 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18492 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18493 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18494 (vect_shift_permute_load_chain): Likewise.
18495 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18496 (expand_vector_condition, optimize_vector_constructor): Likewise.
18497 (lower_vec_perm, get_compute_type): Likewise.
18498 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18499 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18500 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18501 (vect_recog_mask_conversion_pattern): Likewise.
18502 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18503 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18504 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18505 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18506 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18507 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18508 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18509 (supportable_widening_operation): Likewise.
18510 (supportable_narrowing_operation): Likewise.
18511 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18513 * varasm.c (output_constant): Likewise.
18515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18516 Alan Hayward <alan.hayward@arm.com>
18517 David Sherwood <david.sherwood@arm.com>
18519 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18520 so that both the length == 3 and length != 3 cases set up their
18521 own permute vectors. Add comments explaining why we know the
18522 number of elements is constant.
18523 (vect_permute_load_chain): Likewise.
18525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18526 Alan Hayward <alan.hayward@arm.com>
18527 David Sherwood <david.sherwood@arm.com>
18529 * machmode.h (mode_nunits): Change from unsigned char to
18531 (ONLY_FIXED_SIZE_MODES): New macro.
18532 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18533 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18534 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18536 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18537 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18538 or if measurement_type is not polynomial.
18539 * genmodes.c (ZERO_COEFFS): New macro.
18540 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18542 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18543 Use ZERO_COEFFS when emitting initializers.
18544 * data-streamer.h (bp_pack_poly_value): New function.
18545 (bp_unpack_poly_value): Likewise.
18546 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18547 for GET_MODE_NUNITS.
18548 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18549 for GET_MODE_NUNITS.
18550 * tree.c (make_vector_type): Remove temporary shim and make
18551 the real function take the number of units as a poly_uint64
18552 rather than an int.
18553 (build_vector_type_for_mode): Handle polynomial nunits.
18554 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18555 * emit-rtl.c (const_vec_series_p_1): Likewise.
18556 (gen_rtx_CONST_VECTOR): Likewise.
18557 * fold-const.c (test_vec_duplicate_folding): Likewise.
18558 * genrecog.c (validate_pattern): Likewise.
18559 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18560 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18561 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18562 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18563 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18564 * rtlanal.c (subreg_get_info): Likewise.
18565 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18566 (vect_grouped_load_supported): Likewise.
18567 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18568 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18569 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18570 (simplify_const_unary_operation, simplify_binary_operation_1)
18571 (simplify_const_binary_operation, simplify_ternary_operation)
18572 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18573 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18574 instead of CONST_VECTOR_NUNITS.
18575 * varasm.c (output_constant_pool_2): Likewise.
18576 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18577 explicit-encoded elements in the XVEC for variable-length vectors.
18579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18581 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18584 Alan Hayward <alan.hayward@arm.com>
18585 David Sherwood <david.sherwood@arm.com>
18587 * coretypes.h (fixed_size_mode): Declare.
18588 (fixed_size_mode_pod): New typedef.
18589 * builtins.h (target_builtins::x_apply_args_mode)
18590 (target_builtins::x_apply_result_mode): Change type to
18591 fixed_size_mode_pod.
18592 * builtins.c (apply_args_size, apply_result_size, result_vector)
18593 (expand_builtin_apply_args_1, expand_builtin_apply)
18594 (expand_builtin_return): Update accordingly.
18596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18598 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18599 * cselib.c (cselib_hash_rtx): Likewise.
18600 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18601 CONST_VECTOR encoding.
18603 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18604 Jeff Law <law@redhat.com>
18607 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18608 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18609 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18610 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18613 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18614 explicitly probe *sp in a noreturn function if there were any callee
18615 register saves or frame pointer is needed.
18617 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18620 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18621 BLKmode for ternary, binary or unary expressions.
18624 * var-tracking.c (delete_vta_debug_insn): New inline function.
18625 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18626 insns from get_insns () to NULL instead of each bb separately.
18627 Use delete_vta_debug_insn. No longer static.
18628 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18629 delete_vta_debug_insns callers.
18630 * rtl.h (delete_vta_debug_insns): Declare.
18631 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18632 instead of variable_tracking_main.
18634 2018-01-03 Martin Sebor <msebor@redhat.com>
18636 PR tree-optimization/83603
18637 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18638 arguments past the endof the argument list in functions declared
18639 without a prototype.
18640 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18641 Avoid checking when arguments are null.
18643 2018-01-03 Martin Sebor <msebor@redhat.com>
18646 * doc/extend.texi (attribute const): Fix a typo.
18647 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18648 issuing -Wsuggest-attribute for void functions.
18650 2018-01-03 Martin Sebor <msebor@redhat.com>
18652 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18653 offset_int::from instead of wide_int::to_shwi.
18654 (maybe_diag_overlap): Remove assertion.
18655 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18656 * gimple-ssa-sprintf.c (format_directive): Same.
18657 (parse_directive): Same.
18658 (sprintf_dom_walker::compute_format_length): Same.
18659 (try_substitute_return_value): Same.
18661 2018-01-03 Jeff Law <law@redhat.com>
18663 PR middle-end/83654
18664 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18665 non-constant residual for zero at runtime and avoid probing in
18666 that case. Reorganize code for trailing problem to mirror handling
18669 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18671 PR tree-optimization/83501
18672 * tree-ssa-strlen.c (get_string_cst): New.
18673 (handle_char_store): Call get_string_cst.
18675 2018-01-03 Martin Liska <mliska@suse.cz>
18677 PR tree-optimization/83593
18678 * tree-ssa-strlen.c: Include tree-cfg.h.
18679 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18680 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18681 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18683 (strlen_dom_walker::before_dom_children): Call
18684 gimple_purge_dead_eh_edges. Dump tranformation with details
18686 (strlen_dom_walker::before_dom_children): Update call by adding
18687 new argument cleanup_eh.
18688 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18690 2018-01-03 Martin Liska <mliska@suse.cz>
18693 * cif-code.def (VARIADIC_THUNK): New enum value.
18694 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18697 2018-01-03 Jan Beulich <jbeulich@suse.com>
18699 * sse.md (mov<mode>_internal): Tighten condition for when to use
18700 vmovdqu<ssescalarsize> for TI and OI modes.
18702 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18704 Update copyright years.
18706 2018-01-03 Martin Liska <mliska@suse.cz>
18709 * ipa-visibility.c (function_and_variable_visibility): Skip
18710 functions with noipa attribure.
18712 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18714 * gcc.c (process_command): Update copyright notice dates.
18715 * gcov-dump.c (print_version): Ditto.
18716 * gcov.c (print_version): Ditto.
18717 * gcov-tool.c (print_version): Ditto.
18718 * gengtype.c (create_file): Ditto.
18719 * doc/cpp.texi: Bump @copying's copyright year.
18720 * doc/cppinternals.texi: Ditto.
18721 * doc/gcc.texi: Ditto.
18722 * doc/gccint.texi: Ditto.
18723 * doc/gcov.texi: Ditto.
18724 * doc/install.texi: Ditto.
18725 * doc/invoke.texi: Ditto.
18727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18729 * vector-builder.h (vector_builder::m_full_nelts): Change from
18730 unsigned int to poly_uint64.
18731 (vector_builder::full_nelts): Update prototype accordingly.
18732 (vector_builder::new_vector): Likewise.
18733 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18734 (vector_builder::operator ==): Likewise.
18735 (vector_builder::finalize): Likewise.
18736 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18737 Take the number of elements as a poly_uint64 rather than an
18739 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18740 from unsigned int to poly_uint64.
18741 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18742 (vec_perm_indices::new_vector): Likewise.
18743 (vec_perm_indices::length): Likewise.
18744 (vec_perm_indices::nelts_per_input): Likewise.
18745 (vec_perm_indices::input_nelts): Likewise.
18746 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18747 number of elements per input as a poly_uint64 rather than an
18748 unsigned int. Use the original encoding for variable-length
18749 vectors, rather than clamping each individual element.
18750 For the second and subsequent elements in each pattern,
18751 clamp the step and base before clamping their sum.
18752 (vec_perm_indices::series_p): Handle polynomial element counts.
18753 (vec_perm_indices::all_in_range_p): Likewise.
18754 (vec_perm_indices_to_tree): Likewise.
18755 (vec_perm_indices_to_rtx): Likewise.
18756 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18757 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18758 (tree_vector_builder::new_binary_operation): Handle polynomial
18759 element counts. Return false if we need to know the number
18760 of elements at compile time.
18761 * fold-const.c (fold_vec_perm): Punt if the number of elements
18762 isn't known at compile time.
18764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18766 * vec-perm-indices.h (vec_perm_builder): Change element type
18767 from HOST_WIDE_INT to poly_int64.
18768 (vec_perm_indices::element_type): Update accordingly.
18769 (vec_perm_indices::clamp): Handle polynomial element_types.
18770 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18771 (vec_perm_indices::all_in_range_p): Likewise.
18772 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18774 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18775 polynomial vec_perm_indices element types.
18776 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18777 * fold-const.c (fold_vec_perm): Likewise.
18778 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18779 * tree-vect-generic.c (lower_vec_perm): Likewise.
18780 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18781 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18782 element type to HOST_WIDE_INT.
18784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18785 Alan Hayward <alan.hayward@arm.com>
18786 David Sherwood <david.sherwood@arm.com>
18788 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18789 rather than an int. Use plus_constant.
18790 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18791 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18794 Alan Hayward <alan.hayward@arm.com>
18795 David Sherwood <david.sherwood@arm.com>
18797 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18798 a HOST_WIDE_INT to a poly_int64.
18800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18801 Alan Hayward <alan.hayward@arm.com>
18802 David Sherwood <david.sherwood@arm.com>
18804 * calls.c (load_register_parameters): Cope with polynomial
18805 mode sizes. Require a constant size for BLKmode parameters
18806 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18807 forces a parameter to be padded at the lsb end in order to
18808 fill a complete number of words, require the parameter size
18809 to be ordered wrt UNITS_PER_WORD.
18811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18812 Alan Hayward <alan.hayward@arm.com>
18813 David Sherwood <david.sherwood@arm.com>
18815 * reload1.c (spill_stack_slot_width): Change element type
18816 from unsigned int to poly_uint64_pod.
18817 (alter_reg): Treat mode sizes as polynomial.
18819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18820 Alan Hayward <alan.hayward@arm.com>
18821 David Sherwood <david.sherwood@arm.com>
18823 * reload.c (complex_word_subreg_p): New function.
18824 (reload_inner_reg_of_subreg, push_reload): Use it.
18826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18827 Alan Hayward <alan.hayward@arm.com>
18828 David Sherwood <david.sherwood@arm.com>
18830 * lra-constraints.c (process_alt_operands): Reject matched
18831 operands whose sizes aren't ordered.
18832 (match_reload): Refer to this check here.
18834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18835 Alan Hayward <alan.hayward@arm.com>
18836 David Sherwood <david.sherwood@arm.com>
18838 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18839 that the mode size is in the set {1, 2, 4, 8, 16}.
18841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18842 Alan Hayward <alan.hayward@arm.com>
18843 David Sherwood <david.sherwood@arm.com>
18845 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18846 Use plus_constant instead of gen_rtx_PLUS.
18848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18849 Alan Hayward <alan.hayward@arm.com>
18850 David Sherwood <david.sherwood@arm.com>
18852 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18853 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18854 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18855 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18856 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18857 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18858 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18859 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18860 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18861 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18863 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18864 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18865 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18866 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18867 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18868 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18869 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18870 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18871 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18872 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18874 * expr.c (emit_move_resolve_push): Treat the input and result
18875 of PUSH_ROUNDING as a poly_int64.
18876 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18877 (emit_push_insn): Likewise.
18878 * lra-eliminations.c (mark_not_eliminable): Likewise.
18879 * recog.c (push_operand): Likewise.
18880 * reload1.c (elimination_effects): Likewise.
18881 * rtlanal.c (nonzero_bits1): Likewise.
18882 * calls.c (store_one_arg): Likewise. Require the padding to be
18883 known at compile time.
18885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18886 Alan Hayward <alan.hayward@arm.com>
18887 David Sherwood <david.sherwood@arm.com>
18889 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18890 Use plus_constant instead of gen_rtx_PLUS.
18892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18893 Alan Hayward <alan.hayward@arm.com>
18894 David Sherwood <david.sherwood@arm.com>
18896 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18897 rather than an int.
18899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18900 Alan Hayward <alan.hayward@arm.com>
18901 David Sherwood <david.sherwood@arm.com>
18903 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18904 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18905 via stack temporaries. Treat the mode size as polynomial too.
18907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18908 Alan Hayward <alan.hayward@arm.com>
18909 David Sherwood <david.sherwood@arm.com>
18911 * expr.c (expand_expr_real_2): When handling conversions involving
18912 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18913 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18914 as a poly_uint64 too.
18916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18917 Alan Hayward <alan.hayward@arm.com>
18918 David Sherwood <david.sherwood@arm.com>
18920 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18923 Alan Hayward <alan.hayward@arm.com>
18924 David Sherwood <david.sherwood@arm.com>
18926 * combine.c (can_change_dest_mode): Handle polynomial
18927 REGMODE_NATURAL_SIZE.
18928 * expmed.c (store_bit_field_1): Likewise.
18929 * expr.c (store_constructor): Likewise.
18930 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18931 and polynomial REGMODE_NATURAL_SIZE.
18932 (gen_lowpart_common): Likewise.
18933 * reginfo.c (record_subregs_of_mode): Likewise.
18934 * rtlanal.c (read_modify_subreg_p): Likewise.
18936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18937 Alan Hayward <alan.hayward@arm.com>
18938 David Sherwood <david.sherwood@arm.com>
18940 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18941 numbers of elements.
18943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18944 Alan Hayward <alan.hayward@arm.com>
18945 David Sherwood <david.sherwood@arm.com>
18947 * match.pd: Cope with polynomial numbers of vector elements.
18949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18950 Alan Hayward <alan.hayward@arm.com>
18951 David Sherwood <david.sherwood@arm.com>
18953 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18954 in a POINTER_PLUS_EXPR.
18956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18957 Alan Hayward <alan.hayward@arm.com>
18958 David Sherwood <david.sherwood@arm.com>
18960 * omp-simd-clone.c (simd_clone_subparts): New function.
18961 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18962 (ipa_simd_modify_function_body): Likewise.
18964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18965 Alan Hayward <alan.hayward@arm.com>
18966 David Sherwood <david.sherwood@arm.com>
18968 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18969 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18970 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18971 (expand_vector_condition, vector_element): Likewise.
18972 (subparts_gt): New function.
18973 (get_compute_type): Use subparts_gt.
18974 (count_type_subparts): Delete.
18975 (expand_vector_operations_1): Use subparts_gt instead of
18976 count_type_subparts.
18978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18979 Alan Hayward <alan.hayward@arm.com>
18980 David Sherwood <david.sherwood@arm.com>
18982 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18983 (vect_compile_time_alias): ...this new function. Do the calculation
18984 on poly_ints rather than trees.
18985 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18993 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19000 constant and extern definitions for variable-length vectors.
19001 (vect_get_constant_vectors): Note that the number of units
19002 is known to be constant.
19004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19005 Alan Hayward <alan.hayward@arm.com>
19006 David Sherwood <david.sherwood@arm.com>
19008 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19009 of units as polynomial. Choose between WIDE and NARROW based
19012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19013 Alan Hayward <alan.hayward@arm.com>
19014 David Sherwood <david.sherwood@arm.com>
19016 * tree-vect-stmts.c (simd_clone_subparts): New function.
19017 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19020 Alan Hayward <alan.hayward@arm.com>
19021 David Sherwood <david.sherwood@arm.com>
19023 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19024 vectors as polynomial. Use build_index_vector for
19025 IFN_GOMP_SIMD_LANE.
19027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19028 Alan Hayward <alan.hayward@arm.com>
19029 David Sherwood <david.sherwood@arm.com>
19031 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19032 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19033 for variable-length vectors.
19034 (vectorizable_mask_load_store): Treat the number of units as
19035 polynomial, asserting that it is constant if the condition has
19036 already been enforced.
19037 (vectorizable_store, vectorizable_load): Likewise.
19039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19040 Alan Hayward <alan.hayward@arm.com>
19041 David Sherwood <david.sherwood@arm.com>
19043 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19044 of units as polynomial. Punt if we can't tell at compile time
19045 which vector contains the final result.
19047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19048 Alan Hayward <alan.hayward@arm.com>
19049 David Sherwood <david.sherwood@arm.com>
19051 * tree-vect-loop.c (vectorizable_induction): Treat the number
19052 of units as polynomial. Punt on SLP inductions. Use an integer
19053 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19054 cast of such a series for variable-length floating-point
19057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19058 Alan Hayward <alan.hayward@arm.com>
19059 David Sherwood <david.sherwood@arm.com>
19061 * tree.h (build_index_vector): Declare.
19062 * tree.c (build_index_vector): New function.
19063 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19064 of units as polynomial, forcibly converting it to a constant if
19065 vectorizable_reduction has already enforced the condition.
19066 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19067 to create a {1,2,3,...} vector.
19068 (vectorizable_reduction): Treat the number of units as polynomial.
19069 Choose vectype_in based on the largest scalar element size rather
19070 than the smallest number of units. Enforce the restrictions
19073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19074 Alan Hayward <alan.hayward@arm.com>
19075 David Sherwood <david.sherwood@arm.com>
19077 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19078 number of units as polynomial.
19080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19081 Alan Hayward <alan.hayward@arm.com>
19082 David Sherwood <david.sherwood@arm.com>
19084 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19085 * target.def (autovectorize_vector_sizes): Return the vector sizes
19086 by pointer, using vector_sizes rather than a bitmask.
19087 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19088 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19089 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19091 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19092 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19093 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19094 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19095 * omp-general.c (omp_max_vf): Likewise.
19096 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19097 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19098 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19099 * tree-vect-slp.c (vect_slp_bb): Likewise.
19100 * doc/tm.texi: Regenerate.
19101 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19103 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19104 the vector size as a poly_uint64 rather than an unsigned int.
19105 (current_vector_size): Change from an unsigned int to a poly_uint64.
19106 (get_vectype_for_scalar_type): Update accordingly.
19107 * tree.h (build_truth_vector_type): Take the size and number of
19108 units as a poly_uint64 rather than an unsigned int.
19109 (build_vector_type): Add a temporary overload that takes
19110 the number of units as a poly_uint64 rather than an unsigned int.
19111 * tree.c (make_vector_type): Likewise.
19112 (build_truth_vector_type): Take the number of units as a poly_uint64
19113 rather than an unsigned int.
19115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19116 Alan Hayward <alan.hayward@arm.com>
19117 David Sherwood <david.sherwood@arm.com>
19119 * target.def (get_mask_mode): Take the number of units and length
19120 as poly_uint64s rather than unsigned ints.
19121 * targhooks.h (default_get_mask_mode): Update accordingly.
19122 * targhooks.c (default_get_mask_mode): Likewise.
19123 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19124 * doc/tm.texi: Regenerate.
19126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19127 Alan Hayward <alan.hayward@arm.com>
19128 David Sherwood <david.sherwood@arm.com>
19130 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19131 * omp-general.c (omp_max_vf): Likewise.
19132 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19133 (expand_omp_simd): Handle polynomial safelen.
19134 * omp-low.c (omplow_simd_context): Add a default constructor.
19135 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19136 (lower_rec_simd_input_clauses): Update accordingly.
19137 (lower_rec_input_clauses): Likewise.
19139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19140 Alan Hayward <alan.hayward@arm.com>
19141 David Sherwood <david.sherwood@arm.com>
19143 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19144 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19145 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19146 (vect_analyze_slp_cost): Likewise.
19147 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19148 (vect_model_load_cost): Likewise.
19150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19151 Alan Hayward <alan.hayward@arm.com>
19152 David Sherwood <david.sherwood@arm.com>
19154 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19155 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19156 from an unsigned int * to a poly_uint64_pod *.
19157 (calculate_unrolling_factor): New function.
19158 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19161 Alan Hayward <alan.hayward@arm.com>
19162 David Sherwood <david.sherwood@arm.com>
19164 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19165 from an unsigned int to a poly_uint64.
19166 (_loop_vec_info::slp_unrolling_factor): Likewise.
19167 (_loop_vec_info::vectorization_factor): Change from an int
19169 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19170 (vect_get_num_vectors): New function.
19171 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19172 (vect_get_num_copies): Use vect_get_num_vectors.
19173 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19174 to an unsigned int *.
19175 (vect_analyze_data_refs): Change min_vf from an int * to a
19177 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19178 than an unsigned HOST_WIDE_INT.
19179 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19180 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19181 to an unsigned int *.
19182 (vect_analyze_data_ref_dependences): Likewise.
19183 (vect_compute_data_ref_alignment): Handle polynomial vf.
19184 (vect_enhance_data_refs_alignment): Likewise.
19185 (vect_prune_runtime_alias_test_list): Likewise.
19186 (vect_shift_permute_load_chain): Likewise.
19187 (vect_supportable_dr_alignment): Likewise.
19188 (dependence_distance_ge_vf): Take the vectorization factor as a
19189 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19190 (vect_analyze_data_refs): Change min_vf from an int * to a
19192 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19193 vfm1 as a poly_uint64 rather than an int. Make the same change
19194 for the returned bound_scalar.
19195 (vect_gen_vector_loop_niters): Handle polynomial vf.
19196 (vect_do_peeling): Likewise. Update call to
19197 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19198 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19200 * tree-vect-loop.c (vect_determine_vectorization_factor)
19201 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19202 (vect_get_known_peeling_cost): Likewise.
19203 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19204 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19205 (vect_transform_loop): Likewise. Use the lowest possible VF when
19206 updating the upper bounds of the loop.
19207 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19208 rather than an int.
19209 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19210 polynomial unroll factors.
19211 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19212 (vect_make_slp_decision): Likewise.
19213 (vect_supported_load_permutation_p): Likewise, and polynomial
19215 (vect_analyze_slp_cost): Handle polynomial vf.
19216 (vect_slp_analyze_node_operations): Likewise.
19217 (vect_slp_analyze_bb_1): Likewise.
19218 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19219 than an unsigned HOST_WIDE_INT.
19220 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19221 (vectorizable_load): Handle polynomial vf.
19222 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19224 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19227 Alan Hayward <alan.hayward@arm.com>
19228 David Sherwood <david.sherwood@arm.com>
19230 * match.pd: Handle bit operations involving three constants
19231 and try to fold one pair.
19233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19235 * tree-vect-loop-manip.c: Include gimple-fold.h.
19236 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19237 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19238 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19239 Add a path that uses a step of VF instead of 1, but disable it
19241 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19242 and niters_no_overflow parameters. Update calls to
19243 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19244 Create a new SSA name if the latter choses to use a ste other
19245 than zero, and return it via niters_vector_mult_vf_var.
19246 * tree-vect-loop.c (vect_transform_loop): Update calls to
19247 vect_do_peeling, vect_gen_vector_loop_niters and
19248 slpeel_make_loop_iterate_ntimes.
19249 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19250 (vect_gen_vector_loop_niters): Update declarations after above changes.
19252 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19254 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19255 128-bit round to integer instructions.
19256 (ceil<mode>2): Likewise.
19257 (btrunc<mode>2): Likewise.
19258 (round<mode>2): Likewise.
19260 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19262 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19263 unaligned VSX load/store on P8/P9.
19264 (expand_block_clear): Allow the use of unaligned VSX
19265 load/store on P8/P9.
19267 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19269 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19271 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19272 swap associated with both a load and a store.
19274 2018-01-02 Andrew Waterman <andrew@sifive.com>
19276 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19277 * config/riscv/riscv.md (clear_cache): Use it.
19279 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19281 * web.c: Remove out-of-date comment.
19283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19285 * expr.c (fixup_args_size_notes): Check that any existing
19286 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19287 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19288 (emit_single_push_insn): ...here.
19290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19292 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19293 (const_vector_encoded_nelts): New function.
19294 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19295 (const_vector_int_elt, const_vector_elt): Declare.
19296 * emit-rtl.c (const_vector_int_elt_1): New function.
19297 (const_vector_elt): Likewise.
19298 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19299 of CONST_VECTOR_ELT.
19301 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19303 * expr.c: Include rtx-vector-builder.h.
19304 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19305 directly on the tree encoding.
19306 (const_vector_from_tree): Likewise.
19307 * optabs.c: Include rtx-vector-builder.h.
19308 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19309 sequence of "u" values.
19310 * vec-perm-indices.c: Include rtx-vector-builder.h.
19311 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19312 directly on the vec_perm_indices encoding.
19314 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19316 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19317 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19318 * rtx-vector-builder.h: New file.
19319 * rtx-vector-builder.c: Likewise.
19320 * rtl.h (rtx_def::u2): Add a const_vector field.
19321 (CONST_VECTOR_NPATTERNS): New macro.
19322 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19323 (CONST_VECTOR_DUPLICATE_P): Likewise.
19324 (CONST_VECTOR_STEPPED_P): Likewise.
19325 (CONST_VECTOR_ENCODED_ELT): Likewise.
19326 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19327 (unwrap_const_vec_duplicate): Likewise.
19328 (const_vec_series_p): Check for a non-duplicated vector encoding.
19329 Say that the function only returns true for integer vectors.
19330 * emit-rtl.c: Include rtx-vector-builder.h.
19331 (gen_const_vec_duplicate_1): Delete.
19332 (gen_const_vector): Call gen_const_vec_duplicate instead of
19333 gen_const_vec_duplicate_1.
19334 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19335 (gen_const_vec_duplicate): Use rtx_vector_builder.
19336 (gen_const_vec_series): Likewise.
19337 (gen_rtx_CONST_VECTOR): Likewise.
19338 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19339 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19340 Build a new vector rather than modifying a CONST_VECTOR in-place.
19341 (handle_special_swappables): Update call accordingly.
19342 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19343 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19344 Build a new vector rather than modifying a CONST_VECTOR in-place.
19345 (handle_special_swappables): Update call accordingly.
19347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19349 * simplify-rtx.c (simplify_const_binary_operation): Use
19350 CONST_VECTOR_ELT instead of XVECEXP.
19352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19354 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19355 the selector elements to be different from the data elements
19356 if the selector is a VECTOR_CST.
19357 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19358 ssizetype for the selector.
19360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19362 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19363 before testing each element individually.
19364 * tree-vect-generic.c (lower_vec_perm): Likewise.
19366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19368 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19369 * selftest-run-tests.c (selftest::run_tests): Call it.
19370 * vector-builder.h (vector_builder::operator ==): New function.
19371 (vector_builder::operator !=): Likewise.
19372 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19373 (vec_perm_indices::all_from_input_p): New function.
19374 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19375 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19376 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19377 instead of reading the VECTOR_CST directly. Detect whether both
19378 vector inputs are the same before constructing the vec_perm_indices,
19379 and update the number of inputs argument accordingly. Use the
19380 utility functions added above. Only construct sel2 if we need to.
19382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19384 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19385 the broadcast of the low byte.
19386 (expand_mult_highpart): Use an explicit encoding for the permutes.
19387 * optabs-query.c (can_mult_highpart_p): Likewise.
19388 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19389 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19390 (vectorizable_bswap): Likewise.
19391 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19392 explicit encoding for the power-of-2 permutes.
19393 (vect_permute_store_chain): Likewise.
19394 (vect_grouped_load_supported): Likewise.
19395 (vect_permute_load_chain): Likewise.
19397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19399 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19400 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19401 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19402 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19403 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19404 (vect_gen_perm_mask_any): Likewise.
19406 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19408 * int-vector-builder.h: New file.
19409 * vec-perm-indices.h: Include int-vector-builder.h.
19410 (vec_perm_indices): Redefine as an int_vector_builder.
19411 (auto_vec_perm_indices): Delete.
19412 (vec_perm_builder): Redefine as a stand-alone class.
19413 (vec_perm_indices::vec_perm_indices): New function.
19414 (vec_perm_indices::clamp): Likewise.
19415 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19416 (vec_perm_indices::new_vector): New function.
19417 (vec_perm_indices::new_expanded_vector): Update for new
19418 vec_perm_indices class.
19419 (vec_perm_indices::rotate_inputs): New function.
19420 (vec_perm_indices::all_in_range_p): Operate directly on the
19421 encoded form, without computing elided elements.
19422 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19423 encoding. Update for new vec_perm_indices class.
19424 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19425 the given vec_perm_builder.
19426 (expand_vec_perm_var): Update vec_perm_builder constructor.
19427 (expand_mult_highpart): Use vec_perm_builder instead of
19428 auto_vec_perm_indices.
19429 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19430 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19431 or double series encoding as appropriate.
19432 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19433 vec_perm_indices instead of auto_vec_perm_indices.
19434 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19435 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19436 (vect_permute_store_chain): Likewise.
19437 (vect_grouped_load_supported): Likewise.
19438 (vect_permute_load_chain): Likewise.
19439 (vect_shift_permute_load_chain): Likewise.
19440 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19441 (vect_transform_slp_perm_load): Likewise.
19442 (vect_schedule_slp_instance): Likewise.
19443 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19444 (vectorizable_mask_load_store): Likewise.
19445 (vectorizable_bswap): Likewise.
19446 (vectorizable_store): Likewise.
19447 (vectorizable_load): Likewise.
19448 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19449 vec_perm_indices instead of auto_vec_perm_indices. Use
19450 tree_to_vec_perm_builder to read the vector from a tree.
19451 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19452 vec_perm_builder instead of a vec_perm_indices.
19453 (have_whole_vector_shift): Use vec_perm_builder and
19454 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19455 truncation to calc_vec_perm_mask_for_shift.
19456 (vect_create_epilog_for_reduction): Likewise.
19457 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19458 from auto_vec_perm_indices to vec_perm_indices.
19459 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19460 instead of changing individual elements.
19461 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19462 the vector in d.perm.
19463 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19464 from auto_vec_perm_indices to vec_perm_indices.
19465 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19466 instead of changing individual elements.
19467 (arm_vectorize_vec_perm_const): Use new_vector to install
19468 the vector in d.perm.
19469 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19470 Update vec_perm_builder constructor.
19471 (rs6000_expand_interleave): Likewise.
19472 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19473 (rs6000_expand_interleave): Likewise.
19475 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19477 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19478 to qimode could truncate the indices.
19479 * optabs.c (expand_vec_perm_var): Likewise.
19481 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19483 * Makefile.in (OBJS): Add vec-perm-indices.o.
19484 * vec-perm-indices.h: New file.
19485 * vec-perm-indices.c: Likewise.
19486 * target.h (vec_perm_indices): Replace with a forward class
19488 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19489 * optabs.h: Include vec-perm-indices.h.
19490 (expand_vec_perm): Delete.
19491 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19492 (expand_vec_perm_const): Declare.
19493 * target.def (vec_perm_const_ok): Replace with...
19494 (vec_perm_const): ...this new hook.
19495 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19496 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19497 * doc/tm.texi: Regenerate.
19498 * optabs.def (vec_perm_const): Delete.
19499 * doc/md.texi (vec_perm_const): Likewise.
19500 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19501 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19502 expand_vec_perm for constant permutation vectors. Assert that
19503 the mode of variable permutation vectors is the integer equivalent
19504 of the mode that is being permuted.
19505 * optabs-query.h (selector_fits_mode_p): Declare.
19506 * optabs-query.c: Include vec-perm-indices.h.
19507 (selector_fits_mode_p): New function.
19508 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19509 is defined, instead of checking whether the vec_perm_const_optab
19510 exists. Use targetm.vectorize.vec_perm_const instead of
19511 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19512 fit in the vector mode before using a variable permute.
19513 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19514 vec_perm_indices instead of an rtx.
19515 (expand_vec_perm): Replace with...
19516 (expand_vec_perm_const): ...this new function. Take the selector
19517 as a vec_perm_indices rather than an rtx. Also take the mode of
19518 the selector. Update call to shift_amt_for_vec_perm_mask.
19519 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19520 Use vec_perm_indices::new_expanded_vector to expand the original
19521 selector into bytes. Check whether the indices fit in the vector
19522 mode before using a variable permute.
19523 (expand_vec_perm_var): Make global.
19524 (expand_mult_highpart): Use expand_vec_perm_const.
19525 * fold-const.c: Includes vec-perm-indices.h.
19526 * tree-ssa-forwprop.c: Likewise.
19527 * tree-vect-data-refs.c: Likewise.
19528 * tree-vect-generic.c: Likewise.
19529 * tree-vect-loop.c: Likewise.
19530 * tree-vect-slp.c: Likewise.
19531 * tree-vect-stmts.c: Likewise.
19532 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19534 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19535 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19536 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19537 (aarch64_vectorize_vec_perm_const): ...this new function.
19538 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19539 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19540 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19541 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19542 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19543 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19544 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19546 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19547 check for NEON modes.
19548 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19549 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19550 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19551 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19553 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19554 the old VEC_PERM_CONST conditions.
19555 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19556 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19557 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19558 (ia64_vectorize_vec_perm_const_ok): Merge into...
19559 (ia64_vectorize_vec_perm_const): ...this new function.
19560 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19561 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19562 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19563 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19564 * config/mips/mips.c (mips_expand_vec_perm_const)
19565 (mips_vectorize_vec_perm_const_ok): Merge into...
19566 (mips_vectorize_vec_perm_const): ...this new function.
19567 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19568 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19569 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19570 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19571 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19572 (rs6000_expand_vec_perm_const): Delete.
19573 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19575 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19576 (altivec_expand_vec_perm_const_le): Take each operand individually.
19577 Operate on constant selectors rather than rtxes.
19578 (altivec_expand_vec_perm_const): Likewise. Update call to
19579 altivec_expand_vec_perm_const_le.
19580 (rs6000_expand_vec_perm_const): Delete.
19581 (rs6000_vectorize_vec_perm_const_ok): Delete.
19582 (rs6000_vectorize_vec_perm_const): New function.
19583 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19584 an element count and rtx array.
19585 (rs6000_expand_extract_even): Update call accordingly.
19586 (rs6000_expand_interleave): Likewise.
19587 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19588 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19589 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19590 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19591 (rs6000_expand_vec_perm_const): Delete.
19592 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19593 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19594 (altivec_expand_vec_perm_const_le): Take each operand individually.
19595 Operate on constant selectors rather than rtxes.
19596 (altivec_expand_vec_perm_const): Likewise. Update call to
19597 altivec_expand_vec_perm_const_le.
19598 (rs6000_expand_vec_perm_const): Delete.
19599 (rs6000_vectorize_vec_perm_const_ok): Delete.
19600 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19601 reference to the SPE evmerge intructions.
19602 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19603 an element count and rtx array.
19604 (rs6000_expand_extract_even): Update call accordingly.
19605 (rs6000_expand_interleave): Likewise.
19606 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19607 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19609 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19611 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19613 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19614 vector mode and that that mode matches the mode of the data
19616 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19617 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19618 directly using expand_vec_perm_1 when forcing selectors into
19620 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19622 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19624 * optabs-query.h (can_vec_perm_p): Delete.
19625 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19626 * optabs-query.c (can_vec_perm_p): Split into...
19627 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19628 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19629 particular selector is valid.
19630 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19631 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19632 (vect_grouped_load_supported): Likewise.
19633 (vect_shift_permute_load_chain): Likewise.
19634 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19635 (vect_transform_slp_perm_load): Likewise.
19636 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19637 (vectorizable_bswap): Likewise.
19638 (vect_gen_perm_mask_checked): Likewise.
19639 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19640 implementations of variable permutation vectors into account
19641 when deciding which selector to use.
19642 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19643 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19644 with a false third argument.
19645 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19646 to test whether the constant selector is valid and can_vec_perm_var_p
19647 to test whether a variable selector is valid.
19649 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19651 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19652 * optabs-query.c (can_vec_perm_p): Likewise.
19653 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19654 instead of vec_perm_indices.
19655 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19656 (vect_gen_perm_mask_checked): Likewise,
19657 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19658 (vect_gen_perm_mask_checked): Likewise,
19660 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19662 * optabs-query.h (qimode_for_vec_perm): Declare.
19663 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19664 (qimode_for_vec_perm): ...this new function.
19665 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19667 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19669 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19670 does not have a conditional at the top.
19672 2018-01-02 Richard Biener <rguenther@suse.de>
19674 * ipa-inline.c (big_speedup_p): Fix expression.
19676 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19679 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19682 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19686 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19687 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19688 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19689 cond_taken_branch_cost 3->4.
19691 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19693 PR tree-optimization/83581
19694 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19695 TODO_cleanup_cfg if any changes have been made.
19697 PR middle-end/83608
19698 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19699 convert_modes if target mode has the right side, but different mode
19702 PR middle-end/83609
19703 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19704 last argument when extracting from CONCAT. If either from_real or
19705 from_imag is NULL, use expansion through memory. If result is not
19706 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19707 the parts directly to inner mode, if even that fails, use expansion
19710 PR middle-end/83623
19711 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19712 check for bswap in mode rather than HImode and use that in expand_unop
19715 Copyright (C) 2018 Free Software Foundation, Inc.
19717 Copying and distribution of this file, with or without modification,
19718 are permitted in any medium without royalty provided the copyright
19719 notice and this notice are preserved.