1 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
4 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5 (riscv_set_current_function): Extract getting interrupt type to new
7 (riscv_get_interrupt_type): New function.
8 (riscv_merge_decl_attributes): New function, checking interrupt type is
10 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12 2018-07-12 Paul Koning <ni1d@arrl.net>
14 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
17 2018-07-12 Paul Koning <ni1d@arrl.net>
19 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
20 zero reference, add doloop_end instead.
21 * doc/md.texi (decrement_and_branch_until_zero): Remove.
22 (Looping patterns): Remove decrement_and_branch_until_zero. Add
23 detail for doloop_end.
25 2018-07-12 Martin Sebor <msebor@redhat.com>
28 * attribs.c (decl_attributes): Reject conflicting attributes before
29 calling attribute handlers.
31 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
33 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
35 (gcc::dump_manager::get_dump_file_name): likewise.
36 (dump_begin): Likewise.
37 * dumpfile.h (dump_begin): Update prototype.
38 (gcc::dump_manager::get_dump_file_name,
39 gcc::dump_manager::get_dump_file_name): Update prototype.
41 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
43 * internal-fn.h (vectorizable_internal_fn_p): New function.
44 * tree-vect-slp.c (compatible_calls_p): Likewise.
45 (vect_build_slp_tree_1): Remove nops argument. Handle calls
46 to internal functions.
47 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
49 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
51 * fold-const.h (inverse_conditions_p): Declare.
52 * fold-const.c (inverse_conditions_p): New function.
53 * match.pd: Use inverse_conditions_p. Add folds of view_converts
54 that test the inverse condition of a conditional internal function.
55 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
56 * internal-fn.c (internal_fn_mask_index): Handle conditional
58 (vectorized_internal_fn_supported_p): New function.
59 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
60 (any_pred_load_store): Replace with...
61 (need_to_predicate): ...this new variable.
62 (redundant_ssa_names): New variable.
63 (ifcvt_can_use_mask_load_store): Move initial checks to...
64 (ifcvt_can_predicate): ...this new function. Handle tree codes
65 for which a conditional internal function exists.
66 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
67 instead of ifcvt_can_use_mask_load_store. Update after variable
69 (predicate_load_or_store): New function, split out from
71 (check_redundant_cond_expr): New function.
72 (value_available_p): Likewise.
73 (predicate_rhs_code): Likewise.
74 (predicate_mem_writes): Rename to...
75 (predicate_statements): ...this. Use predicate_load_or_store
76 and predicate_rhs_code.
77 (combine_blocks, tree_if_conversion): Update after above name changes.
78 (ifcvt_local_dce): Handle redundant_ssa_names.
79 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
80 general conditional functions.
81 * tree-vect-stmts.c (vectorizable_call): Likewise.
83 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
84 Alan Hayward <alan.hayward@arm.com>
85 David Sherwood <david.sherwood@arm.com>
87 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
88 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
89 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
90 plus and minus and convert them into IFN_COND_FMA-based sequences.
91 (convert_mult_to_fma): Handle conditional plus and minus.
93 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
95 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
96 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
97 (cond_fnms_optab): New optabs.
98 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
100 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
101 * internal-fn.h (get_conditional_internal_fn): Declare.
102 (get_unconditional_internal_fn): Likewise.
103 * internal-fn.c (cond_ternary_direct): New macro.
104 (expand_cond_ternary_optab_fn): Likewise.
105 (direct_cond_ternary_optab_supported_p): Likewise.
106 (FOR_EACH_COND_FN_PAIR): Likewise.
107 (get_conditional_internal_fn): New function.
108 (get_unconditional_internal_fn): Likewise.
109 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
110 (gimple_match_op::gimple_match_op): Add a new overload for 5
112 (gimple_match_op::set_op): Likewise.
113 (gimple_resimplify5): Declare.
114 * genmatch.c (decision_tree::gen): Generate simplifications for
116 * gimple-match-head.c (gimple_simplify): Define an overload for
117 5 operands. Handle calls with 5 arguments in the top-level overload.
118 (convert_conditional_op): Handle conversions from unconditional
119 internal functions to conditional ones.
120 (gimple_resimplify5): New function.
121 (build_call_internal): Pass a fifth operand.
122 (maybe_push_res_to_seq): Likewise.
123 (try_conditional_simplification): Try converting conditional
124 internal functions to unconditional internal functions.
125 Handle 3-operand unconditional forms.
126 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
127 Define ternary equivalents of the current rules for binary conditional
129 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
131 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
132 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
133 (optab): Handle them.
134 (SVE_COND_FP_TERNARY): New int iterator.
135 (sve_fmla_op, sve_fmad_op): New int attributes.
136 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
137 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
138 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
140 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
142 * target.def (preferred_else_value): New target hook.
143 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
144 * doc/tm.texi: Regenerate.
145 * targhooks.h (default_preferred_else_value): Declare.
146 * targhooks.c (default_preferred_else_value): New function.
147 * internal-fn.h (conditional_internal_fn_code): Declare.
148 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
149 (get_conditional_internal_fn): Use it.
150 (conditional_internal_fn_code): New function.
151 * gimple-match.h (gimple_match_cond): New struct.
152 (gimple_match_op): Add a cond member function.
153 (gimple_match_op::gimple_match_op): Update all forms to take a
155 * genmatch.c (expr::gen_transform): Use the same condition as res_op
156 for the suboperation, but don't specify a particular else_value.
157 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
158 (visit_nary_op, visit_reference_op_load): Pass
159 gimple_match_cond::UNCOND to the gimple_match_op constructor.
160 * gimple-match-head.c: Include tree-eh.h
161 (convert_conditional_op): New function.
162 (maybe_resimplify_conditional_op): Likewise.
163 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
164 (gimple_resimplify2): Likewise.
165 (gimple_resimplify3): Likewise.
166 (gimple_resimplify4): Likewise.
167 (maybe_push_res_to_seq): Return null for conditional operations.
168 (try_conditional_simplification): New function.
169 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
171 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
173 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
175 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
177 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
179 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
181 (hash_tree): Do not hash DECL_FCONTEXT
182 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
183 Do not stream DECL_FCONTEXT.
184 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
185 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
187 2018-07-12 Richard Biener <rguenther@suse.de>
190 * dwarf2out.c (gen_block_die): Only output blocks when they have
191 at least one !DECL_IGNORED_P variable.
193 2018-07-12 Richard Biener <rguenther@suse.de>
196 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
197 Remove -mieee-fp handling.
199 2018-07-12 Richard Biener <rguenther@suse.de>
201 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
202 left-over from last patch.
204 2018-07-12 Jakub Jelinek <jakub@redhat.com>
206 PR tree-optimization/86492
207 * gimple-ssa-store-merging.c
208 (imm_store_chain_info::coalesce_immediate_stores): Call
209 check_no_overlap even for the merge_overlapping case. Formatting fix.
211 2018-07-12 Richard Biener <rguenther@suse.de>
214 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
215 move possibly trapping operations into the conditional.
217 2018-07-12 Richard Biener <rguenther@suse.de>
219 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
220 (vn_lookup_simplify_result): Remove recursion limit applied
222 (vn_nary_build_or_lookup_1): Adjust.
223 (try_to_simplify): Likewise.
224 * gimple-match-head.c (gimple_resimplify1): Instead apply one
226 (gimple_resimplify2): Likewise.
227 (gimple_resimplify3): Likewise.
228 (gimple_resimplify4): Likewise.
230 2018-07-11 Jakub Jelinek <jakub@redhat.com>
232 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
233 Use __mmask64 type instead of __mmask8 for __M argument.
234 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
235 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
237 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
238 __mmask16 for __M argument.
239 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
240 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
241 to __mmask16 instead of __mmask8.
242 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
243 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
244 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
245 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
246 instead of __mmask16 for __U argument.
247 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
248 __mmask16 instead of __mmask8 for __U argument.
249 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
251 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
253 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
255 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
257 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
259 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
261 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
263 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
265 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
267 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
269 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
270 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
271 return type as well as __M argument type and all casts from __mmask8
273 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
274 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
275 return type as well as __M argument type and all casts from __mmask8
277 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
278 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
279 return type as well as __M argument type and all casts from __mmask8
281 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
282 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
283 return type as well as __M argument type and all casts from __mmask8
285 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
286 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
289 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
291 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
292 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
295 2018-07-11 Paul Koning <ni1d@arrl.net>
297 * doc/md.texi (define_subst): Document how multiple occurrences of
298 the same argument in the replacement pattern are handled.
300 2018-07-11 Paul Koning <ni1d@arrl.net>
302 * doc/extend.texi (Common Variable Attributes): Move "mode" into
304 (Common Type Attributes): Add "mode" attribute.
306 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
308 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
309 stream DECL_ORIGINAL_TYPE.
310 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
311 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
312 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
313 Do not walk original type.
314 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
316 (write_ts_decl_non_common_tree_pointers): Do not stream
318 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
319 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
321 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
323 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
324 thread twice from the same starting edge.
326 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
328 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
329 * gimple.c (gimple_call_nonnull_result_p): ...here...
330 (gimple_call_nonnull_arg): ...and here.
331 * gimple.h (gimple_call_nonnull_result_p): New.
332 (gimple_call_nonnull_arg): New.
334 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
336 * config/arm/arm-cpus.in: Move information from fpu field of each
337 cpu definition to the isa field.
338 * config/arm/parsecpu.awk (fpu): Delete match rule.
339 (gen_comm_data): Don't add bits from the CPU's FPU entry.
341 2018-07-11 Richard Biener <rguenther@suse.de>
344 * dwarf2out.c (init_sections_and_labels): Use
345 output_asm_line_debug_info consistently.
346 (dwarf2out_early_finish): Likewise.
347 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
350 2018-07-11 Richard Biener <rguenther@suse.de>
352 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
353 Rework father_bb setting in a way to avoid propagating constants
354 multiple times on a loop body.
356 2018-07-10 Mark Wielaard <mark@klomp.org>
359 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
361 2018-07-10 Richard Biener <rguenther@suse.de>
363 * hash-map.h (hash_map::iterator::operator*): Return
364 references to key and value.
366 2018-07-10 Jakub Jelinek <jakub@redhat.com>
369 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
370 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
371 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
372 (gimplify_omp_for): For composite loops, move outer
373 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
374 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
375 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
376 TREE_LIST for both the original class iterator and the "last" helper
377 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
378 loop, remember has_decl_expr from outer composite loops for the
379 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
381 2018-07-09 Martin Sebor <msebor@redhat.com>
385 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
386 accessing implicitly initialized array elements.
387 * expr.c (string_constant): Handle string initializers of
388 character arrays within aggregates.
389 * gimple-fold.c (fold_array_ctor_reference): Add argument.
390 Store element offset. As a special case, handle zero size.
391 (fold_nonarray_ctor_reference): Same.
392 (fold_ctor_reference): Add argument. Store subobject offset.
393 * gimple-fold.h (fold_ctor_reference): Add argument.
395 2018-07-09 Paul Koning <ni1d@arrl.net>
397 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
398 (pdp11_insn_cost): New function.
399 (pdp11_md_asm_adjust): New function.
400 (TARGET_INVALID_WITHIN_DOLOOP): Define.
401 (pdp11_rtx_costs): Update to match machine better.
402 (output_addr_const_pdp11): Correct format mismatch warnings.
403 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
404 * config/pdp11/pdp11.md: General change to add base_cost and/or
405 length attributes for use by new pdp11_insn_cost function.
406 (MIN_BRANCH): Correct definition.
408 (doloop_end): Use standard pattern name for looping pattern.
409 (doloop_end_nocc): New.
410 (movsf): Add another constraint alternative.
411 (zero_extendqihi2): Add constraint alternatives for not in place
413 (zero_extendhisi2): Remove.
414 (shift patterns): Add CC handling variants.
418 (define_peephole2): New peephole to recognize mov that sets CC for
421 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
423 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
424 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
427 2018-07-09 Jakub Jelinek <jakub@redhat.com>
430 * real.c (real_nextafter): Return true if result is denormal.
432 2018-07-09 Martin Liska <mliska@suse.cz>
434 * common.opt: Add back wrongly removed attribute.
436 2018-07-09 Richard Biener <rguenther@suse.de>
439 * dwarf2out.c (gen_block_die): For an early generated DIE
440 always output high/low PC attributes.
442 2018-07-09 Tom de Vries <tdevries@suse.de>
444 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
445 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
446 onto VAR_DECL with abstract origin.
448 2018-07-07 Jim Wilson <jimw@sifive.com>
450 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
452 2018-07-07 Tom de Vries <tdevries@suse.de>
454 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
456 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
458 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
460 (combine_bound): Use wide-int overflow calculation instead of
462 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
464 * fold-const.c (int_const_binop_2): Same.
465 (extract_muldiv_1): Same.
466 (fold_div_compare): Same.
467 (fold_abs_const): Same.
469 * poly-int.h (add): Same.
473 * predict.c (predict_iv_comparison): Same.
474 * profile-count.c (slow_safe_scale_64bit): Same.
475 * simplify-rtx.c (simplify_const_binary_operation): Same.
476 * tree-chrec.c (tree_fold_binomial): Same.
477 * tree-data-ref.c (split_constant_offset_1): Same.
478 * tree-if-conv.c (idx_within_array_bound): Same.
479 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
480 * tree-ssa-phiopt.c (minmax_replacement): Same.
481 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
482 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
483 * vr-values.c (vr_values::adjust_range_with_scev): Same.
484 * wide-int.cc (wi::add_large): Same.
485 (wi::mul_internal): Same.
486 (wi::sub_large): Same.
487 (wi::divmod_internal): Same.
488 * wide-int.h: Change overflow type to overflow_type for neg, add,
489 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
490 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
491 mul_internal, divmod_internal.
492 (overflow_type): New enum.
493 (accumulate_overflow): New.
495 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
497 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
498 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
500 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
502 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
503 argument is checked for zero before entering loop, avoid checking again.
505 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
507 * gimplify.h (generic_expr_could_trap_p): Set as global function.
508 * gimplify.h (generic_expr_could_trap_p): Likwise.
509 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
511 2018-07-06 Jakub Jelinek <jakub@redhat.com>
513 PR tree-optimization/86401
514 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
515 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
516 (fold_bit_and_mask): ... here. New helper function for match.pd.
517 * fold-const.h (fold_bit_and_mask): Declare.
518 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
520 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
523 * target.def (translate_mode_attribute): New hook.
524 * targhooks.h (default_translate_mode_attribute): Declare.
525 * targhooks.c (default_translate_mode_attribute): New function.
526 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
527 * doc/tm.texi: Regenerate.
528 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
529 (rs6000_translate_mode_attribute): New function.
531 2018-07-06 Paul Koning <ni1d@arrl.net>
533 * doc/md.texi (define_split): Document DONE and FAIL.
534 (define_peephole2): Ditto.
536 2018-07-05 Jeff Law <law@redhat.com>
538 PR tree-optimization/86010
539 * tree-ssa-dse.c (compute_trims): More aggressively trim at
540 both the head and tail of mem* and str* calls.
542 2018-07-05 Jim Wilson <jimw@sifive.com>
544 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
546 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
548 * config/aarch64/aarch64-simd.md: correct flags text for
549 MIN_EXPR replacement.
551 2018-07-05 James Clarke <jrtc27@jrtc27.com>
553 * configure: Regenerated.
555 2018-07-05 Carl Love <cel@us.ibm.com>
557 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
558 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
559 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
560 VSX_BUILTIN_DOUBLEL_V4SF.
562 2018-07-05 Martin Sebor <msebor@redhat.com>
565 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
566 than its domain to compute its the upper bound of a char array.
568 2018-07-05 Nathan Sidwell <nathan@acm.org>
570 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
571 * doc/cpp.texi: Update comment.
572 * doc/tm.texi: Rebuilt.
573 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
574 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
575 * doc/extend.texi (Backwards Compatibility): Clarify it is system
576 headers affected by extern "C".
577 * system.h: Poison NO_IMPLICIT_EXTERN_C.
578 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
579 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
580 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
581 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
582 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
583 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
584 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
585 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
586 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
587 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
588 config/sparc/sp64-elf.h, config/spu/spu.h,
589 config/stormy16/stormy16.h, config/v850/v850.h,
590 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
591 define NO_IMPLICIT_EXTERN_C.
592 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
594 2018-07-05 Tamar Christina <tamar.christina@arm.com>
597 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
598 instead of GET_MODE_SIZE when comparing Units.
600 2018-07-05 Tamar Christina <tamar.christina@arm.com>
603 * rtlanal.c (set_noop_p): Constrain on mode change,
604 include hard-reg-set.h
606 2018-07-05 Tamar Christina <tamar.christina@arm.com>
608 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
610 2018-07-05 Jakub Jelinek <jakub@redhat.com>
613 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
616 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
618 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
620 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
623 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
625 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
627 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
630 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
632 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
634 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
637 2018-07-04 Martin Liska <mliska@suse.cz>
639 * tree-switch-conversion.c: Define
640 max_ratio_for_speed and max_ratio_for_size constants.
642 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
643 Martin Liska <mliska@suse.cz>
648 * common.opt: Rename align options with 'str_' prefix.
649 * common/config/i386/i386-common.c (set_malign_value): New
651 (ix86_handle_option): Use it to set -falign-* options/
652 * config/aarch64/aarch64-protos.h (struct tune_params): Change
653 type from int to string.
654 * config/aarch64/aarch64.c: Update default values from int
656 * config/alpha/alpha.c (alpha_override_options_after_change):
658 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
659 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
660 max skip conditionally.
661 * config/i386/freebsd.h (SUBALIGN_LOG): New.
662 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
663 max skip conditionally.
664 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
665 max skip conditionally.
666 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
667 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
668 max skip conditionally.
669 * config/i386/i386.c (struct ptt): Change type from int to
671 (ix86_default_align): Set default values.
672 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
673 max skip conditionally.
674 * config/i386/iamcu.h (SUBALIGN_LOG): New.
675 (ASM_OUTPUT_MAX_SKIP_ALIGN):
676 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
677 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
678 max skip conditionally.
679 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
680 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
681 * config/i386/x86-64.h (SUBALIGN_LOG): New.
682 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
683 max skip conditionally.
684 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
685 * config/ia64/ia64.c (ia64_option_override): Set default values
686 for alignment options.
687 * config/m68k/m68k.c: Handle new str_align_* options.
688 * config/mips/mips.c (mips_set_compression_mode): Change
690 (mips_option_override): Set default values for options.
691 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
693 * config/rs6000/rs6000.c (rs6000_option_override_internal):
695 * config/rx/rx.c (rx_option_override): Likewise.
696 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
697 (LABEL_ALIGN): Use align_labels_log.
698 (LOOP_ALIGN): Use align_loops_align.
699 * config/s390/s390.c (s390_asm_output_function_label): Use new
701 * config/sh/sh.c (sh_override_options_after_change):
702 Change type of constants.
703 * config/spu/spu.c (spu_sched_init): Likewise.
704 * config/sparc/sparc.c (sparc_option_override): Set default
706 * config/visium/visium.c (visium_option_override): Likewise.
707 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
708 emit p2align format with last argument if it's not needed.
709 * doc/invoke.texi: Document extended format of -falign-*.
710 * final.c: Use align_labels alignment.
711 * flags.h (struct target_flag_state): Change type to use
713 (struct align_flags_tuple): New.
714 (struct align_flags): Likewise.
715 (align_loops_log): Redefine macro to use new types.
716 (align_loops_max_skip): Redefine macro to use new types.
717 (align_jumps_log): Redefine macro to use new types.
718 (align_jumps_max_skip): Redefine macro to use new types.
719 (align_labels_log): Redefine macro to use new types.
720 (align_labels_max_skip): Redefine macro to use new types.
721 (align_functions_log): Redefine macro to use new types.
722 (align_loops): Redefine macro to use new types.
723 (align_jumps): Redefine macro to use new types.
724 (align_labels): Redefine macro to use new types.
725 (align_functions): Redefine macro to use new types.
726 (align_functions_max_skip): Redefine macro to use new types.
727 (align_loops_value): New macro.
728 (align_jumps_value): New macro.
729 (align_labels_value): New macro.
730 (align_functions_value): New macro.
731 * function.c (invoke_set_current_function_hook): Propagate
732 alignment values from flags to global variables default in
734 * ipa-icf.c (sem_function::equals_wpa): Use
735 cl_optimization_option_eq instead of memcmp.
736 * lto-streamer.h (cl_optimization_stream_out): Support streaming
738 (cl_optimization_stream_in): Likewise.
739 * optc-save-gen.awk: Support strings in cl_optimization.
740 * opth-gen.awk: Likewise.
741 * opts.c (finish_options): Remove error checking of invalid
743 (MAX_CODE_ALIGN): Remove.
744 (MAX_CODE_ALIGN_VALUE): Likewise.
745 (parse_and_check_align_values): New function.
746 (check_alignment_argument): Likewise.
747 (common_handle_option): Use check_alignment_argument.
748 * opts.h (parse_and_check_align_values): Declare.
749 * toplev.c (init_alignments): Remove.
750 (read_log_maxskip): New.
751 (parse_N_M): Likewise.
752 (parse_alignment_opts): Likewise.
753 (backend_init_target): Remove usage of init_alignments.
754 * toplev.h (parse_alignment_opts): Declare.
755 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
757 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
758 * tree.c (cl_option_hasher::equal): New.
759 * varasm.c: Use new global macros.
761 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
763 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
764 Use a simpler align directive also if MAXSKIP = ALIGN-1.
765 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
766 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
767 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
768 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
769 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
770 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
771 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
772 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
773 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
774 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
775 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
777 2018-07-04 Martin Liska <mliska@suse.cz>
778 Jonathan Wakely <jwakely@redhat.com>
780 * coverage.c: Use correct type.
781 * doc/invoke.texi: Language correction.
783 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
786 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
787 ENDBRANCH for non-tail call which may return via indirect branch.
788 * doc/extend.texi: Document indirect_return attribute.
790 2018-07-03 Martin Sebor <msebor@redhat.com>
792 PR tree-optimization/86274
793 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
795 (format_floating): Correct handling of infinities and NaNs.
797 2018-07-03 Martin Sebor <msebor@redhat.com>
799 * print-tree.c (print_real_cst): New function.
800 (print_node_brief): Call it.
803 2018-07-03 Jeff Law <law@redhat.com>
805 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
806 into a single pattern.
808 * config/h8300/h8300.md (ors code_iterator): New.
809 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
810 a single pattern and single splitter.
811 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
812 (iorqi3_1, xorqi3_1): Likewise.
813 (iorqi3, xorqi3 expanders): Similarly.
815 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
816 (movmd_internal) into a single pattern using the P mode iterator.
817 (movmd splitters): Similarly.
818 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
819 (movsd splitters): Similarly.
821 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
822 ADDB, ADDW and ADDL into a single ADD attribute which selects the
823 right table based on the size of the operand.
824 * config/h8300/h8300.md (length_table): Corresponding changes. All
825 references to "addb", "addw" and "addl" changed to "add".
826 (btst patterns): Merge two variants into a single pattern.
827 (tstqi, tsthi): Likewise.
828 (addhi3_incdec, addsi3_incdec): Likewise.
829 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
830 (mulhi3, mulsi3): Likewise.
831 (udivhi3, udivsi3): Likewise.
832 (divhi3, divsi3): Likewise.
833 (andorqi3, andorhi3, andorsi3): Likewise.
835 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
838 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
839 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
841 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
843 PR tree-optimization/85694
844 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
845 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
847 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
848 (<u>avg<mode>3_ceil): New patterns.
850 2018-07-03 David Malcolm <dmalcolm@redhat.com>
852 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
853 scan-tree-dump directive.
854 * gcc.dg/vect/slp-perm-2.c: Likewise.
855 * gcc.dg/vect/slp-perm-3.c: Likewise.
856 * gcc.dg/vect/slp-perm-5.c: Likewise.
857 * gcc.dg/vect/slp-perm-6.c: Likewise.
858 * gcc.dg/vect/slp-perm-7.c: Likewise.
859 * gcc.dg/vect/slp-perm-8.c: Likewise.
861 2018-07-03 Marek Polacek <polacek@redhat.com>
864 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
866 2018-07-03 Richard Biener <rguenther@suse.de>
869 * tree-ssa-structalias.c (find_func_clobbers): Properly
870 handle indirect calls.
872 2018-07-03 Jeff Law <law@redhat.com>
874 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
875 (shifts): New code iterator.
876 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
877 expander. Fix HImode handling on H8/SX.
878 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
879 (subqi3, subhi3, subsi3 expanders): Likewise.
880 (andqi3, andhi3, andsi3 expanders): Likewise.
881 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
882 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
883 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
884 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
885 (zero_extendqihi2, zero_extendqisi2): Likewise.
886 (extendqihi2, extendqisi2): Likewise.
887 (rotlqi3, rotlhi3, rotlsi3): Likewise.
888 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
889 (rotlqi3_1, rotlhi3_1): Likewise.
890 (logicalhi3_sn, logicalsi3_sn): Likewise.
891 (logicalhi3, logicalsi3): Likewise.
893 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
895 * tree-vect-patterns.c (vect_recog_rotate_pattern)
896 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
897 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
898 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
899 type to append_pattern_def_seq instead of creating a stmt_vec_info
901 (build_mask_conversion): Likewise. Remove vinfo argument.
902 (vect_add_conversion_to_patterm): Likewise, renaming to...
903 (vect_add_conversion_to_pattern): ...this.
904 (vect_recog_mask_conversion_pattern): Update call to
905 build_mask_conversion. Pass the vector type to
906 append_pattern_def_seq here too.
907 (vect_recog_gather_scatter_pattern): Update call to
908 vect_add_conversion_to_pattern.
910 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
912 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
913 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
914 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
915 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
916 STMT_VINFO_PATTERN_DEF_SEQ to null here.
917 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
918 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
919 append_pattern_def_seq instead of new_pattern_def_seq.
920 (vect_recog_divmod_pattern): Do both of the above.
921 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
924 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
926 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
927 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
928 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
929 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
930 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
931 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
932 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
933 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
934 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
935 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
936 parameter with a single stmt_vec_info.
937 (vect_recog_func_ptr): Likewise.
938 (vect_recog_gather_scatter_pattern): Likewise, folding in...
939 (vect_try_gather_scatter_pattern): ...this.
940 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
941 the stmt_vec_info of the statement to be matched. Don't clear
942 STMT_VINFO_RELATED_STMT.
943 (vect_pattern_recog): Update call accordingly.
945 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
947 PR tree-optimization/85694
948 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
949 (uavgM3_ceil): Document new optabs.
950 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
951 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
953 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
954 (savg_ceil_optab): New optabs.
955 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
956 (vect_vect_recog_func_ptrs): Add it.
957 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
958 constant directly from the associated lhs.
960 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
962 * tree-vect-patterns.c (vect_split_statement): New function.
963 (vect_convert_input): Use it to try to split an existing cast.
965 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
967 * poly-int.h (print_hex): New function.
968 * dumpfile.h (dump_dec, dump_hex): Declare.
969 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
970 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
971 min_input_precision, operation_precision and operation_sign.
972 * tree-vect-patterns.c (vect_get_range_info): New function.
973 (vect_same_loop_or_bb_p, vect_single_imm_use)
974 (vect_operation_fits_smaller_type): Delete.
975 (vect_look_through_possible_promotion): Add an optional
976 single_use_p parameter.
977 (vect_recog_over_widening_pattern): Rewrite to use new
978 stmt_vec_info infomration. Handle one operation at a time.
979 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
980 (vect_truncatable_operation_p, vect_set_operation_type)
981 (vect_set_min_input_precision): New functions.
982 (vect_determine_min_output_precision_1): Likewise.
983 (vect_determine_min_output_precision): Likewise.
984 (vect_determine_precisions_from_range): Likewise.
985 (vect_determine_precisions_from_users): Likewise.
986 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
987 (vect_vect_recog_func_ptrs): Put over_widening first.
989 (vect_pattern_recog): Call vect_determine_precisions.
991 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
993 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
994 statements that have been replaced by further pattern statements.
995 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
997 2018-07-03 Richard Biener <rguenther@suse.de>
999 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1000 always set *dt. Dump vectype in vectype overload.
1001 * dumpfile.h (dump_gimple_expr): New function.
1002 (dump_gimple_expr_loc): Likewise.
1003 * dumpfile.c (dump_gimple_expr): New function.
1004 (dump_gimple_expr_loc): Likewise.
1006 2018-07-02 Jeff Law <law@redhat.com>
1008 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1009 the H8/300, H8/300H and H8/S variants into a single pattern.
1010 (movhi_h8300, movqi_h8300hs): Similarly.
1011 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1012 (QHI mode iterator): New.
1014 * config/h8300/h8300.md: Remove trailing whitespace.
1016 2018-07-02 Jim Wilson <jimw@sifive.com>
1018 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1019 instead of emit_insn for interrupt returns.
1020 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1021 (riscv_sret, riscv_uret): Likewise.
1023 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1025 * pretty-print.c (selftest::test_pp_format): Move save and restore
1026 of quotes to class auto_fix_quotes, and add an instance.
1027 * selftest.c: Include "intl.h".
1028 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1029 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1030 * selftest.h (selftest::auto_fix_quotes): New class.
1032 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1034 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1035 (aarch64_sve_prepare_conditional_op): Remove.
1036 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1037 Allow aarch64_simd_reg_or_zero as select operand; remove
1038 the aarch64_sve_prepare_conditional_op call.
1039 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1040 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1041 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1042 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1043 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1044 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1045 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1046 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1047 and a splitters to match all of the *_any patterns.
1048 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1050 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1051 (SVE_COND_FP_BINARY_REV): Remove.
1052 (sve_int_op_rev, sve_fp_op_rev): New.
1053 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1054 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1055 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1056 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1057 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1058 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1059 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1060 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1061 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1063 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1064 Remove match_dup 1 from the inner unspec.
1065 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1067 * config/aarch64/aarch64.md (movprfx): New attr.
1068 (length): Default movprfx to 8.
1069 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1070 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1071 (*<su>mul<SVE_I>3_highpart): Likewise.
1072 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1073 (*v<ASHIFT><SVE_I>3): Likewise.
1074 (*<su><MAXMIN><SVE_I>3): Likewise.
1075 (*<su><MAXMIN><SVE_F>3): Likewise.
1076 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1077 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1078 (*div<SVE_F>4): Likewise.
1080 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1082 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1085 2018-07-02 Richard Biener <rguenther@suse.de>
1087 PR tree-optimization/86363
1088 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1089 memset argument refers to a non-variable address.
1091 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1093 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1094 {PLUS,MINUS}_EXPR code to...
1095 (adjust_symbolic_bound): ...here,
1096 (combine_bound): ...here,
1097 (set_value_range_with_overflow): ...and here.
1099 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1101 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1103 (extract_range_from_abs_expr): ...here.
1105 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1107 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1108 -fno-omit-frame-pointer when not optimizing.
1110 2018-07-02 Martin Liska <mliska@suse.cz>
1113 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1114 (propagate_nothrow): Likewise.
1116 2018-07-02 Martin Liska <mliska@suse.cz>
1119 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1121 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1123 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1124 function in r262149, changing "loc" param from source_location to
1125 const dump_location_t &.
1126 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1127 declaration, as above.
1129 2018-07-01 Paul Koning <ni1d@arrl.net>
1131 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1132 -munit-asm, -mgnu-asm, -mdec-asm.
1133 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1134 (pdp11_output_labelref): New.
1135 (pdp11_output_def): New.
1136 (pdp11_output_addr_vec_elt): New.
1137 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1138 %# and %@ format codes.
1139 (pdp11_option_override): New.
1140 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1141 (pdp11_output_ident): New.
1142 (pdp11_asm_named_section): New.
1143 (pdp11_asm_init_sections): New.
1144 (pdp11_file_start): New.
1145 (pdp11_file_end): New.
1146 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1147 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1148 %o, like %c but octal.
1149 (pdp11_option_override): New.
1150 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1152 (DATA_SECTION_ASM_OP): Ditto.
1153 (READONLY_DATA_SECTION_ASM_OP): New.
1154 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1155 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1156 (ASM_OUTPUT_LABELREF): Ditto.
1157 (ASM_OUTPUT_DEF): Ditto.
1158 (ASM_OUTPUT_EXTERNAL): New.
1159 (ASM_OUTPUT_SOURCE_FILENAME): New.
1160 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1161 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1162 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1163 %# and %@ format codes.
1164 * config/pdp11/pdp11.opt (mgnu-asm): New.
1165 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1166 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1167 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1169 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1171 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1172 dereferencing path[] beyond its length.
1174 (debug_all_paths): New.
1175 (rewire_first_differing_edge): New.
1176 (adjust_paths_after_duplication): New.
1177 (duplicate_thread_path): Call adjust_paths_after_duplication.
1179 (thread_through_all_blocks): Add new argument to
1180 duplicate_thread_path.
1182 2018-06-30 Jim Wilson <jimw@sifive.com>
1184 * config/riscv/predicates.md (p2m1_shift_operand): New.
1185 (high_mask_shift_operand): New.
1186 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1187 pattern using p2m1_shift_operand.
1188 (lshsi3_zero_extend_3+2): New combiner pattern using
1189 high_mask_shift_operand.
1191 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1193 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1195 (vect_recog_rotate_pattern): ...here.
1196 (vect_convert_input): Try to insert casts of invariants in the
1198 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1199 preheader to be empty.
1201 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1203 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1204 vector type. If given, install it in the new statement's
1206 (vect_element_precision): New function.
1207 (vect_unpromoted_value): New struct.
1208 (vect_unpromoted_value::vect_unpromoted_value): New function.
1209 (vect_unpromoted_value::set_op): Likewise.
1210 (vect_look_through_possible_promotion): Likewise.
1211 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1212 (vect_widened_op_tree, vect_convert_input): Likewise.
1213 (vect_convert_inputs, vect_convert_output): Likewise.
1214 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1215 to handle the optional cast of the multiplication result and
1216 vect_widened_op_tree to detect the widened multiplication itself.
1217 Do not require the input and output of promotion casts to have
1218 the same sign, but base the signedness of the operation on the
1219 input rather than the result. If the pattern includes two
1220 promotions, check that those promotions have the same sign.
1221 Do not restrict the MULT_EXPR handling to a double-width result;
1222 handle quadruple-width results and wider. Use vect_convert_inputs
1223 to convert the inputs to the common type.
1224 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1225 to handle the optional cast of the ABS result. Also allow a sign
1226 change or a sign extension between the ABS and MINUS.
1227 Use vect_widened_op_tree to detect the widened subtraction and use
1228 vect_convert_inputs to convert the inputs to the common type.
1229 (vect_handle_widen_op_by_const): Delete.
1230 (vect_recog_widen_op_pattern): New function.
1231 (vect_recog_widen_mult_pattern): Use it.
1232 (vect_recog_widen_shift_pattern): Likewise.
1233 (vect_recog_widen_sum_pattern): Use
1234 vect_look_through_possible_promotion to handle the promoted
1237 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1239 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1240 the containing gimple_seq *.
1241 * gimple-iterator.h (gsi_for_stmt): Declare it.
1242 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1243 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1244 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1245 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1246 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1248 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1250 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1251 statement being replaced is part of an existing pattern
1252 definition sequence, inserting the new pattern statements before
1254 (vect_pattern_recog_1): Don't return a bool. If the statement
1255 is already part of a pattern, instead apply pattern matching
1256 to the pattern definition statements. Don't clear the
1257 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1258 (vect_pattern_recog): Don't break after the first match;
1259 continue processing the pattern definition statements instead.
1260 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1262 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1264 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1265 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1266 (vect_recog_widen_sum_pattern): Use it.
1268 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1270 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1271 phi is not a pattern statement and has not been replaced by
1272 a pattern statement.
1273 * tree-vect-patterns.c (type_conversion_p): Don't check
1274 STMT_VINFO_IN_PATTERN_P.
1275 (vect_recog_vector_vector_shift_pattern): Likewise.
1276 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1277 the pattern statement rather than the original statement; check
1278 directly for a WIDEN_MULT_EXPR here.
1279 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1280 vect_is_simple_use to return the pattern statement rather
1281 than the original statement; use is_pattern_stmt_p to check
1282 for such a pattern statement.
1283 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1284 to return the pattern statement rather than the original statement;
1285 don't do the same transformation here.
1286 (vect_is_simple_use): If the defining statement has been replaced
1287 by a pattern statement, return the pattern statement instead.
1288 Remove the corresponding (local) transformation from the vectype
1291 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1293 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1294 end and default to null.
1295 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1296 (vectorizable_reduction): Update calls accordingly, dropping the
1297 gimple ** argument if the passed-back statement isn't needed.
1298 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1299 (vect_recog_rotate_pattern): Likewise.
1300 (vect_recog_mask_conversion_pattern): Likewise.
1301 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1302 (vect_mask_constant_operand_p): Likewise.
1303 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1304 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1305 (get_group_load_store_type, get_load_store_type): Likewise.
1306 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1307 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1308 (vectorizable_conversion, vectorizable_assignment): Likewise.
1309 (vectorizable_shift, vectorizable_operation): Likewise.
1310 (vectorizable_store, vect_is_simple_cond): Likewise.
1311 (vectorizable_condition, vectorizable_comparison): Likewise.
1312 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1313 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1314 and move it to the end. Cope with null def_stmt_outs.
1316 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1318 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1320 2018-06-29 Jeff Law <law@redhat.com>
1322 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1323 displacements for TARGET_V850E2V3 and newer.
1324 (TARGET_LRA_P): Remove. Defaults to LRA now.
1325 * config/v850/v850.md (sign23byte_load): Remove.
1326 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1327 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1329 2018-06-29 Martin Liska <mliska@suse.cz>
1332 * coverage.c (coverage_init): Mangle full path name.
1333 * doc/invoke.texi: Document the change.
1334 * gcov-io.c (mangle_path): New.
1335 * gcov-io.h (mangle_path): Likewise.
1336 * gcov.c (mangle_name): Use mangle_path for path mangling.
1338 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1340 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1341 if starting source register is not even.
1343 2018-06-29 Martin Liska <mliska@suse.cz>
1345 PR tree-optimization/86263
1346 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1347 Make edge redirection.
1349 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1351 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1352 (dump_scope_depth): New variable.
1353 (get_dump_scope_depth): New function.
1354 (dump_begin_scope): New function.
1355 (dump_end_scope): New function.
1356 * dumpfile.h (get_dump_scope_depth): New declaration.
1357 (dump_begin_scope): New declaration.
1358 (dump_end_scope): New declaration.
1359 (class auto_dump_scope): New class.
1360 (AUTO_DUMP_SCOPE): New macro.
1361 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1364 2018-06-29 Richard Biener <rguenther@suse.de>
1366 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1367 compute_all_dependences succeeds.
1368 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1369 exceed --param loop-max-datarefs-for-datadeps.
1371 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1373 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1375 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1378 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1379 alternative 0 in preferred_for_speed attribute.
1381 2018-06-28 Paul Koning <ni1d@arrl.net>
1383 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1384 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1385 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1386 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1387 for shift insn patterns.
1389 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1391 * cgraph.c (cgraph_node::get_body): Replace assignments to
1392 "dump_file" with calls to set_dump_file.
1393 * dumpfile.c (alt_dump_file): Make static, and group with...
1394 (alt_flags): ...this definition.
1395 (dumps_are_enabled): New variable.
1396 (refresh_dumps_are_enabled): New function.
1397 (set_dump_file): New function.
1398 (set_alt_dump_file): New function.
1399 (gcc::dump_manager::dump_start): Replace assignments to
1400 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1402 (gcc::dump_manager::dump_finish): Likewise.
1403 * dumpfile.h (alt_dump_file): Delete decl.
1404 (dumps_are_enabled): New variable decl.
1405 (set_dump_file): New function decl.
1406 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1408 * tree-nested.c (lower_nested_functions): Replace assignments to
1409 "dump_file" with calls to set_dump_file.
1411 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1413 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1414 goto_locus of each outgoing edge of each basic block.
1416 2018-06-28 Richard Biener <rguenther@suse.de>
1418 * dwarf2out.c (decl_scope_table): Remove.
1419 (push_decl_scope): Likewise.
1420 (pop_decl_scope): Likewise.
1421 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1422 (gen_struct_or_union_type_die): Likewise.
1423 (gen_tagged_type_die): Likewise.
1424 (dwarf2out_init): Do not initialize decl_scope_table.
1425 (dwarf2out_c_finalize): Do not free it.
1427 2018-06-28 Richard Biener <rguenther@suse.de>
1429 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1430 deciding whether to not re-use a DIE.
1432 2018-06-28 Richard Biener <rguenther@suse.de>
1434 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1435 DW_AT_abstract_origin attribute.
1437 2018-06-28 Martin Liska <mliska@suse.cz>
1439 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1440 Use newly introduced constants.
1441 * tree-switch-conversion.h (struct jump_table_cluster):
1442 Define max_ratio_for_size and max_ratio_for_speed.
1444 2018-06-28 Martin Liska <mliska@suse.cz>
1446 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1447 Add new checking assert to catch invalid state.
1448 (jump_table_cluster::can_be_handled): Handle single case
1450 (jump_table_cluster::is_beneficial): Bail out for such case.
1451 (bit_test_cluster::find_bit_tests):
1452 Add new checking assert to catch invalid state.
1453 (bit_test_cluster::can_be_handled): Handle single case
1455 (bit_test_cluster::is_beneficial): Bail out for such case.
1456 (switch_decision_tree::analyze_switch_statement):
1459 2018-06-28 Martin Liska <mliska@suse.cz>
1461 * common.opt: Introduce -completion option.
1462 * gcc.c (driver_handle_option): Handle it.
1463 (driver::main): Print completions if completion
1465 * opt-suggestions.c (option_proposer::get_completions):
1467 (option_proposer::suggest_completion): Likewise.
1468 (option_proposer::find_param_completions): Likewise.
1469 (verify_autocompletions): Likewise.
1470 (test_completion_valid_options): Likewise.
1471 (test_completion_valid_params): Likewise.
1472 (in_completion_p): Likewise.
1473 (empty_completion_p): Likewise.
1474 (test_completion_partial_match): Likewise.
1475 (test_completion_garbage): Likewise.
1476 (opt_proposer_c_tests): Likewise.
1477 * opt-suggestions.h: Declare new functions.
1478 * opts.c (common_handle_option): Handle OPT__completion_.
1479 * selftest-run-tests.c (selftest::run_tests): Add
1480 opt_proposer_c_tests.
1481 * selftest.c (assert_str_startswith): New.
1482 * selftest.h (assert_str_startswith): Likewise.
1483 (opt_proposer_c_tests): New.
1484 (ASSERT_STR_STARTSWITH): Likewise.
1486 2018-06-28 Martin Liska <mliska@suse.cz>
1488 * Makefile.in: Add opt-suggestions.o.
1489 * gcc-main.c: Include opt-suggestions.h.
1490 * gcc.c (driver::driver): Likewise.
1491 (driver::~driver): Remove m_option_suggestions.
1492 (driver::build_option_suggestions): Moved to option_proposer.
1493 (driver::suggest_option): Likewise.
1494 (driver::handle_unrecognized_options): Use option_proposer.
1495 * gcc.h (class driver): Add new memver m_option_proposer.
1496 * opt-suggestions.c: New file.
1497 * opt-suggestions.h: New file.
1499 2018-06-28 Martin Liska <mliska@suse.cz>
1501 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1502 (auto_string_vec::~auto_string_vec): Likewise.
1504 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1506 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1507 prevent_decl_creation_for_types fields up and add reset_location field.
1508 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1509 statement if id->reset_location is true.
1510 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1511 id->reset_location is true.
1512 (copy_phis_for_bb): Force input_location on the arguments if
1513 id->reset_location is true.
1514 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1515 is set on the function to be inlined.
1517 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1519 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1521 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1523 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1524 registers for Pmode.
1525 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1526 hard registers for the clobbered pseudo.
1528 2018-06-27 Paul Koning <ni1d@arrl.net>
1530 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1531 mutually exclusive options.
1532 * config/pdp11/constraints.md (h): New constraint.
1533 (O): Update definition to match shift code generation.
1534 (D): New constraint.
1535 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1537 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1539 (output_jump): Change arguments.
1540 (pdp11_fixed_cc_regs): New function.
1541 (pdp11_cc_mode): Ditto.
1542 (pdp11_expand_shift): Ditto.
1543 (pdp11_assemble_shift): Ditto.
1544 (pdp11_small_shift): Ditto.
1545 (pdp11_branch_cost): Remove.
1546 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1548 (pdp11_register_move_cost): Update for CC registers.
1549 (pdp11_rtx_costs): Add case for LSHIFTRT.
1550 (pdp11_output_jump): Add CCNZ mode conditional branches.
1551 (notice_update_cc_on_set): Remove.
1552 (pdp11_cc_mode): New function.
1553 (simple_memory_operand): Correct pre/post decrement case.
1554 (no_side_effect_operand): New function.
1555 (pdp11_regno_reg_class): Add CC_REGS class.
1556 (pdp11_fixed_cc_regs): New function.
1557 (pdp11_small_shift): New function.
1558 (pdp11_expand_shift): New function to expand shift insns.
1559 (pdp11_assemble_shift): New function to output shifts.
1560 (pdp11_branch_cost): Remove.
1561 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1562 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1563 (WCHAR_TYPE): Ditto.
1564 (PTRDIFF_TYPE): Ditto.
1565 (ADJUST_INSN_LENGTH): New macro.
1566 (FIXED_REGISTERS): Add CC registers.
1567 (CALL_USED_REGISTERS): Ditto.
1569 (REG_CLASS_NAMES): Ditto.
1570 (REG_CLASS_CONTENTS): Ditto.
1571 (SELECT_CC_MODE): Use new function.
1572 (TARGET_FLAGS_REGNUM): New macro.
1573 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1574 (cc0_reg_rtx): Remove.
1575 (CC_STATUS_MDEP): Remove.
1576 (CC_STATUS_MDEFP_INIT): Remove.
1577 (CC_IN_FPU): Remove.
1578 (NOTICE_UPDATE_CC): Remove.
1579 (REGISTER_NAMES): Add CC registers.
1580 (BRANCH_COST): Change to constant 1.
1581 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1583 * config/pdp11/pdp11.opt (mbcopy): Remove.
1584 (mbcopy-builtin): Remove.
1585 (mbranch-cheap): Remove.
1586 (mbranch-expensive): Remove.
1587 * config/pdp11/predicates.md (expand_shift_operand): Update to
1588 match shift code generation.
1589 (ccnz_operator): New predicate.
1590 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1591 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1592 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1594 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1595 description of O constraint.
1597 2018-06-27 Jeff Law <law@redhat.com>
1598 Austin Law <austinklaw@gmail.com>
1600 * config/v850/v850.md (addsi3_set_flags): New pattern.
1601 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1602 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1603 (zero_extendhisi2_v850_set_flags): Likewise.
1604 (zero_extendqisi2_v850_set_flags): Likewise.
1605 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1606 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1607 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1609 * config/v850/v850-protos.h (notice_update_cc): Remove.
1610 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1611 (v850_print_operand): Handle 'D' and "d".
1612 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1613 Add handling of arithmetic/logical operations compared against zero.
1614 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1615 Do not look at v850_compare_op, instead get mode from last argument.
1616 (v850_gen_compare): Remove
1617 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1618 after reload for prologue insns.
1619 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1621 (construct_save_jarl): Likewise.
1622 (TARGET_FLAGS_REGNUM): Define.
1623 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1624 (NOTICE_UPDATE_CC): Remove.
1625 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1626 than cc0. Conditionalize on reload_completed.
1627 (cmpsi_insn, setfcc_insn): Likewise.
1628 (tst1 splitter): Turn into define_and_split which sets the flags
1630 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1631 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1632 (cstoresf4, cstoredf4): Clobber the flags.
1633 (cmpsi, cmpsf, cmpdf): Remove expanders.
1634 (setf_insn): Remove pattern.
1635 (addsi3): Turn into define_and_split which clobbers the flags after
1636 reload and a suitable pattern (addsi3_clobber_flags) for use after
1638 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1639 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1640 (ashrsi3, ashrsi3_v850e2): Likewise.
1641 (bins): Clobber the flags.
1642 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1643 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1644 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1645 (call_value_internal_short, call_value_internal_long): Likewise.
1646 (callt_save_interrupt, callt_return_interrupt): Likewise.
1647 (save_interrupt, return_interrupt): Likewise.
1648 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1649 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1650 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1651 (All FP comparisons): Only allow after reload has completed.
1653 (divh, divhu): Tweak output template.
1654 (branch_z_normal, branch_z_invert): Remove
1655 (branch_nz_normal, branch_nz_invert): Likewise.
1656 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1658 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1659 * config/v850/v850.c (notice_update_cc): Remove.
1660 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1661 (CC_NO_CARRY): Likewise.
1662 (NOTICE_UPDATE_CC): Define to nothing.
1663 * config/v850/v850.md: Remove block comment on cc0 handling
1664 Remove "cc" attribute from all patterns. Remove cc_status handling
1665 from all patterns. Minor formatting fixes.
1667 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1669 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1670 (cortex-a76.cortex-a55): Likewise.
1671 * config/aarch64/aarch64-tune.md: Regenerate.
1672 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1673 cortex-a76.cortex-a55.
1675 2018-06-27 Jeff Law <law@redhat.com>
1677 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1678 (MULTILIB_DIRNAMES): Similarly.
1680 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1682 * gimple.h (gimple_return_retbnd): Delete.
1683 (gimple_return_set_retbnd): Likewise.
1684 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1685 gimple_return_set_retbnd.
1686 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1687 gimple_return_retbnd and adjust.
1688 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1689 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1690 Explicitly return NULL in a couple more cases. Move assertion
1691 on debug statements and remove unreachable code.
1692 (reset_debug_binding): Do not test id->retbnd.
1693 (expand_call_inline): Do not set it.
1695 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1697 * configure.ac: Add --disable-gcov option.
1698 * configure: Regenerate.
1699 * Makefile.in: Honour @enable_gcov@.
1700 * doc/install.texi: Document --disable-gcov.
1702 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1704 * config/arm/arm-cpus.in (cortex-a76): New entry.
1705 (cortex-a76.cortex-a55): Likewise.
1706 * config/arm/arm-tables.opt: Regenerate.
1707 * config/arm/arm-tune.md: Likewise.
1708 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1709 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1710 cortex-a76.cortex-a55.
1712 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1715 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1717 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1719 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1721 (EPILOGUE_USES): Likewise.
1723 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1725 * tree-inline.c (remap_location): New function extracted from...
1726 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1727 (copy_phis_for_bb): ...here. Call remap_location.
1728 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1730 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1732 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1733 unaligned vsx for 16B memset.
1735 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1738 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1739 ieee128_float_type_node to long_double_type_node unless
1740 TARGET_LONG_DOUBLE_128 is set.
1742 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1744 * cfgloop.c (get_loop_location): Convert return type from
1745 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1746 by implicit construction from rtx_insn *, and using
1747 dump_user_location_t::from_function_decl for the fallback case.
1748 * cfgloop.h (get_loop_location): Convert return type from
1749 location_t to dump_user_location_t.
1750 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1751 dump_printf_loc to pass in a dump_location_t rather than a
1752 location_t, via the gimple stmt.
1753 * coverage.c (get_coverage_counts): Update calls to
1754 dump_printf_loc to pass in dump_location_t rather than a
1756 * doc/optinfo.texi (Dump types): Convert example of
1757 dump_printf_loc from taking "locus" to taking "insn". Update
1758 description of the "_loc" calls to cover dump_location_t.
1759 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1761 (dump_user_location_t::dump_user_location_t): New constructors,
1762 from gimple *stmt and rtx_insn *.
1763 (dump_user_location_t::from_function_decl): New function.
1764 (dump_loc): Make static.
1765 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1766 const dump_location_t &.
1767 (dump_generic_expr_loc): Delete.
1768 (dump_printf_loc): Convert param "loc" from location_t to
1769 const dump_location_t &.
1770 (selftest::test_impl_location): New function.
1771 (selftest::dumpfile_c_tests): New function.
1772 * dumpfile.h: Include "profile-count.h".
1773 (class dump_user_location_t): New class.
1774 (struct dump_impl_location_t): New struct.
1775 (class dump_location_t): New class.
1776 (dump_printf_loc): Convert 2nd param from source_location to
1777 const dump_location_t &.
1778 (dump_generic_expr_loc): Delete.
1779 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1780 const dump_location_t &.
1781 * gimple-fold.c (fold_gimple_assign): Update call to
1782 dump_printf_loc to pass in a dump_location_t rather than a
1783 location_t, via the gimple stmt.
1784 (gimple_fold_call): Likewise.
1785 * gimple-loop-interchange.cc
1786 (loop_cand::analyze_iloop_reduction_var): Update for change to
1787 check_reduction_path.
1788 (tree_loop_interchange::interchange): Update for change to
1790 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1791 change in return-type of find_loop_location.
1792 (graphite_regenerate_ast_isl): Likewise.
1793 * graphite-optimize-isl.c (optimize_isl): Likewise.
1794 * graphite.c (graphite_transform_loops): Likewise.
1795 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1796 pass in a dump_location_t rather than a location_t, via the
1798 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1799 * ipa.c (walk_polymorphic_call_targets): Likewise.
1800 * loop-unroll.c (report_unroll): Convert "locus" param from
1801 location_t to dump_location_t.
1802 (decide_unrolling): Update for change to get_loop_location's
1804 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1805 location_t to dump_user_location_t.
1806 (grid_find_single_omp_among_assignments_1): Updates calls to
1807 dump_printf_loc to pass in a dump_location_t rather than a
1808 location_t, via the gimple stmt.
1809 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1810 location_t to dump_location_t. Updates calls to dump_printf_loc
1811 to pass in a dump_location_t rather than a location_t, via the
1813 (grid_inner_loop_gridifiable_p): Likewise.
1814 (grid_dist_follows_simple_pattern): Likewise.
1815 (grid_gfor_follows_tiling_pattern): Likewise.
1816 (grid_target_follows_gridifiable_pattern): Likewise.
1817 (grid_attempt_target_gridification): Convert initialization
1818 of local "grid" from memset to zero-initialization; FIXME: does
1819 this require C++11? Update call to dump_printf_loc to pass in a
1820 optinfo_location rather than a location_t, via the gimple stmt.
1821 * profile.c (read_profile_edge_counts): Updates call to
1822 dump_printf_loc to pass in a dump_location_t rather than a
1824 (compute_branch_probabilities): Likewise.
1825 * selftest-run-tests.c (selftest::run_tests): Call
1827 * selftest.h (dumpfile_c_tests): New decl.
1828 * tree-loop-distribution.c (pass_loop_distribution::execute):
1829 Update for change in return type of find_loop_location.
1830 * tree-parloops.c (parallelize_loops): Likewise.
1831 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1832 "locus" from location_t to dump_user_location_t.
1833 (canonicalize_loop_induction_variables): Likewise.
1834 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1835 for change in return type of find_loop_location.
1836 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1837 to dump_printf_loc to pass in a dump_location_t rather than a
1838 location_t, via the stmt.
1839 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1841 * tree-vect-loop-manip.c (find_loop_location): Convert return
1842 type from source_location to dump_user_location_t.
1843 (vect_do_peeling): Update for above change.
1844 (vect_loop_versioning): Update for change in type of
1846 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1847 from location_t to dump_user_location_t.
1848 (vect_estimate_min_profitable_iters): Update for change in type
1850 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1851 location_t to dump_location_t.
1852 (vect_slp_bb): Update for change in type of vect_location.
1853 * tree-vectorizer.c (vect_location): Convert from source_location
1854 to dump_user_location_t.
1855 (try_vectorize_loop_1): Update for change in vect_location's type.
1856 (vectorize_loops): Likewise.
1857 (increase_alignment): Likewise.
1858 * tree-vectorizer.h (vect_location): Convert from source_location
1859 to dump_user_location_t.
1860 (find_loop_location): Convert return type from source_location to
1861 dump_user_location_t.
1862 (check_reduction_path): Convert 1st param from location_t to
1863 dump_user_location_t.
1864 * value-prof.c (check_counter): Update call to dump_printf_loc to
1865 pass in a dump_user_location_t rather than a location_t; update
1866 call to error_at for change in type of "locus".
1867 (check_ic_target): Update call to dump_printf_loc to
1868 pass in a dump_user_location_t rather than a location_t, via the
1871 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1873 * config/s390/s390.h (enum processor_flags): Do not use
1874 default tune parameter when -march was specified.
1876 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1879 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1880 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1883 2018-06-26 Richard Biener <rguenther@suse.de>
1885 PR tree-optimization/86287
1887 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1888 (vect_analyze_loop): Initialize n_stmts.
1890 2018-06-26 Richard Biener <rguenther@suse.de>
1893 * fold-const.c (fold_convertible_p): Pointer extension
1896 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1899 * dwarf2out.c (loc_list_has_views): Adjust comments.
1900 (dw_loc_list): Split single cross-partition range with
1903 2018-06-25 Jeff Law <law@redhat.com>
1905 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1906 on -mbig-switch by default.
1908 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1910 (const_float_0_operand): Remove unused predicate.
1911 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1912 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1913 (recipsf2): New expander. Original pattern now called
1915 (recipdf2, recipdf2_insn): Similarly.
1916 (rsqrtsf2, rsqrtsf2_insn): Similarly
1917 (rsqrtdf2, rsqrtdf2_insn): Similarly
1919 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1921 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1922 Simplify logic for FreeBSD (twice).
1924 2018-06-25 Martin Sebor <msebor@redhat.com>
1926 PR tree-optimization/86204
1927 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1928 a strnlen result if it's less than the length of the string.
1930 2018-06-25 Martin Sebor <msebor@redhat.com>
1932 PR tree-optimization/85700
1933 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1934 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1935 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1937 2018-06-25 Martin Sebor <msebor@redhat.com>
1939 * doc/extend.texi (Zero-length arrays): Update and clarify.
1941 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1943 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1944 added IEEE/IBM long double multilib support on PowerPC little
1945 endian Linux systems.
1946 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1947 (MULTILIB_DEFAULTS): Likewise.
1948 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1950 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1951 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1952 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1954 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1957 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1958 (REORDER_45): Likewise.
1960 2018-06-25 Jeff Law <law@redhat.com>
1962 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1963 dividend to 32 bits. Adjust length.
1964 (udivmodhi4): Cleanup output template. Fix length.
1966 2018-06-25 Carl Love <cel@us.ibm.com>
1968 * config/rs6000/vsx.md: Change word selector to prefered location.
1970 2018-06-25 Richard Biener <rguenther@suse.de>
1972 PR tree-optimization/86304
1973 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1974 epilogue-if-converted loops as well.
1976 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1978 * lto-section-out.c (lto_begin_section): Do not print section
1979 name for noaddr and unnumbered dumps.
1981 2018-06-25 Richard Biener <rguenther@suse.de>
1983 * tree-vectorizer.h (struct vec_info_shared): New structure
1984 with parts split out from struct vec_info and loop_nest from
1985 struct _loop_vec_info.
1986 (struct vec_info): Adjust accordingly.
1987 (struct _loop_vec_info): Likewise.
1988 (LOOP_VINFO_LOOP_NEST): Adjust.
1989 (LOOP_VINFO_DATAREFS): Likewise.
1990 (LOOP_VINFO_DDRS): Likewise.
1991 (struct _bb_vec_info): Likewise.
1992 (BB_VINFO_DATAREFS): Likewise.
1993 (BB_VINFO_DDRS): Likewise.
1994 (struct _stmt_vec_info): Add dr_aux member.
1995 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1996 (DR_MISALIGNMENT_UNINITIALIZED): New.
1997 (set_dr_misalignment): Adjust.
1998 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1999 (vect_analyze_loop): Adjust prototype.
2000 (vect_analyze_loop_form): Likewise.
2001 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2002 Compute dependences lazily.
2003 (vect_record_base_alignments): Use shared datarefs/ddrs.
2004 (vect_verify_datarefs_alignment): Likewise.
2005 (vect_analyze_data_refs_alignment): Likewise.
2006 (vect_analyze_data_ref_accesses): Likewise.
2007 (vect_analyze_data_refs): Likewise.
2008 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2009 constructor parameter for shared part.
2010 (vect_analyze_loop_form): Pass in shared part and adjust.
2011 (vect_analyze_loop_2): Pass in storage for the number of
2012 stmts. Move loop nest finding to the caller. Compute
2014 (vect_analyze_loop): Pass in shared part.
2015 (vect_transform_loop): Verify shared datarefs are unchanged.
2016 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2017 constructor parameter for shared part.
2018 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2019 (vect_slp_bb): Verify shared datarefs are unchanged before
2021 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2023 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2024 DR_MISALIGNMENT_UNINITIALIZED.
2025 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2026 parameter for shared part.
2027 (vec_info::~vec_info): Adjust.
2028 (vec_info_shared::vec_info_shared): New.
2029 (vec_info_shared::~vec_info_shared): Likewise.
2030 (vec_info_shared::save_datarefs): Likewise.
2031 (vec_info_shared::check_datarefs): Likewise.
2032 (try_vectorize_loop_1): Construct shared part live for analyses
2033 of a single loop for multiple vector sizes.
2034 * tree-parloops.c (gather_scalar_reductions): Adjust.
2036 2018-06-25 Richard Biener <rguenther@suse.de>
2038 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2039 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2040 (vect_analyze_data_refs): Remove similar code from here and
2041 simplify accordingly.
2043 2018-06-25 Richard Biener <rguenther@suse.de>
2045 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2046 for reverse storage order accesses rather than asserting
2047 they cannot happen here.
2049 2018-06-25 Tom de Vries <tdevries@suse.de>
2052 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2053 Use data16 instead of .byte for insn prefix.
2055 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2058 * parser.c (make_char_string_pack): Pass this literal chars
2059 through cpp_interpret_string.
2060 (cp_parser_userdef_numeric_literal): Check the result of
2061 make_char_string_pack.
2063 2018-06-24 Maya Rashish <coypu@sdf.org>
2065 * ginclude/stddef.h: Simplify conditions around avoiding
2066 re-definition of __size_t.
2068 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2070 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2071 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2073 2018-06-22 Maya Rashish <coypu@sdf.org>
2075 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2076 made to OpenBSD and NetBSD through the years.
2078 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2080 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2081 behavior of vec_pack (vector double, vector double) to match
2082 behavior of vec_float2 (vector double, vector double).
2084 2018-06-22 Olivier Hainque <hainque@adacore.com>
2086 * gimplify.c (gimplify_function_tree): Prevent creation
2087 of a trampoline for the address of the current function
2088 passed to entry/exit instrumentation hooks.
2090 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2093 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2096 2018-06-22 Martin Liska <mliska@suse.cz>
2098 PR tree-optimization/86263
2099 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2100 Bail out if is_enabled is false.
2101 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2103 (jump_table_cluster::is_enabled): New function.
2105 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2107 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2108 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2109 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2110 (lto_input_ts_binfo_tree_pointers): Likewise.
2111 * tree-streamer-out.c (streamer_write_tree_bitfields,
2112 write_ts_binfo_tree_pointers): Likewise.
2113 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2115 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2117 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2119 2018-06-22 Martin Liska <mliska@suse.cz>
2121 * symbol-summary.h (get): Make it pure and inline move
2122 functionality from ::get function.
2123 (get): Remove and inline into ::get and ::get_create.
2124 (get_create): Move code from ::get function.
2126 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2129 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2130 -x assembler-with-cpp.
2132 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2134 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2135 _FILE_OFFSET_BITS=64 for C++.
2137 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2139 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2140 conversion insn that shows up when pr85657-3.c is compiled using
2141 IEEE 128-bit long double.
2142 (neg<mode>2_internal): Use the correct mode to check whether the
2143 mode is IBM extended.
2144 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2145 multiply and divide external functions from being created more
2148 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2150 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2152 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2153 the edge can be forwarded.
2154 (cfg_layout_merge_blocks): Likewise.
2156 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2158 * except.c (finish_eh_generation): Commit edge insertions only after
2159 the EH edges have been redirected from post-landing to landing pads.
2161 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2163 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2164 create_tmp_var_for to create the FRAME decl.
2165 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2167 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2169 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2170 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2171 debug statement when resetting its value.
2172 (expand_call_inline): Copy the locus of the call onto the assignment
2173 of the return value, if any. Use local variable in more cases.
2175 2018-06-21 Martin Liska <mliska@suse.cz>
2177 * ipa-pure-const.c (propagate_nothrow): Use
2178 funct_state_summaries->get.
2179 (dump_malloc_lattice): Likewise.
2180 (propagate_malloc): Likewise.
2182 2018-06-21 Richard Biener <rguenther@suse.de>
2184 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2185 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2186 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2188 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2189 BLOCK_ABSTRACT_ORIGIN unconditionally.
2191 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2193 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2195 * ipa-reference.c (ipa_reference_c_finalize): Delete
2196 ipa_ref_opt_sum_summaries and set it to NULL.
2198 2018-06-21 Tom de Vries <tdevries@suse.de>
2200 PR tree-optimization/85859
2201 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2202 test with comment from bb_no_side_effects_p.
2204 2018-06-21 Richard Biener <rguenther@suse.de>
2206 PR tree-optimization/86232
2207 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2208 max for constant niter.
2210 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2212 * config/aarch64/aarch64-simd.md
2213 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2215 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2217 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2218 Make opernads of the unspec commutative.
2220 2018-06-21 Richard Biener <rguenther@suse.de>
2222 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2223 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2224 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2225 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2227 (vect_record_base_alignments): Do not record base alignment
2228 for gathers or scatters.
2229 (vect_compute_data_ref_alignment): Drop return value that is always
2230 true. Bail out early for gathers or scatters.
2231 (vect_enhance_data_refs_alignment): Bail out early for gathers
2233 (vect_find_same_alignment_drs): Likewise.
2234 (vect_analyze_data_refs_alignment): Remove dead code.
2235 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2236 (vect_analyze_data_refs): For possible gathers or scatters do
2237 not create an alternate DR, just check their possible validity
2238 and mark them. Adjust DECL_NONALIASED handling to not rely
2240 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2241 update inits of gathers or scatters.
2242 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2243 Also copy gather/scatter flag to pattern vinfo.
2245 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2247 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2248 behavior of vec_packsu (vector unsigned long long, vector unsigned
2249 long long) to match behavior of vec_packs with same signature.
2251 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2252 Thomas Schwinge <thomas@codesourcery.com>
2253 Cesar Philippidis <cesar@codesourcery.com>
2255 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2256 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2257 (gimplify_adjust_omp_clauses): Likewise.
2258 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2259 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2260 (gimplify_omp_target_update): Update handling of acc update and
2262 * omp-low.c (install_var_field): Remove unused parameter
2263 base_pointers_restrict.
2264 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2265 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2267 (omp_target_base_pointers_restrict_p): Delete.
2268 (scan_omp_target): Update call to scan_sharing_clauses.
2269 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2271 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2272 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2273 (convert_local_omp_clauses): Likewise.
2274 * tree-pretty-print.c (dump_omp_clause): Likewise.
2275 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2277 (omp_clause_code_name): Likewise.
2279 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2282 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2285 PR tree-optimization/86231
2286 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2287 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2289 2018-06-20 Tom de Vries <tdevries@suse.de>
2291 PR tree-optimization/86097
2292 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2293 iv type if signedness of iv type is not the same as that of *nit.
2295 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2297 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2298 EDGE_EH edges, verify they are all EDGE_EH.
2300 2018-06-20 Maya Rashish <coypu@sdf.org>
2302 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2304 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2306 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2307 * config/aarch64/aarch64.c (xgene1_tunings): Add
2308 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2309 (aarch64_mode_valid_for_sched_fusion_p):
2310 Allow 16-byte modes.
2311 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2312 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2314 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2316 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2317 * config/aarch64/iterators.md (VQ2): New mode iterator.
2319 2018-06-20 Martin Liska <mliska@suse.cz>
2321 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2322 Change default ratio from 10 to 8.
2324 2018-06-20 Martin Liska <mliska@suse.cz>
2326 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2328 (bit_test_cluster::find_bit_tests): Likewise.
2329 (switch_decision_tree::analyze_switch_statement): Find clusters.
2330 * tree-switch-conversion.h (struct jump_table_cluster): Document
2333 2018-06-20 Martin Liska <mliska@suse.cz>
2335 * tree-switch-conversion.c (switch_conversion::collect):
2336 Record m_uniq property.
2337 (switch_conversion::expand): Bail out for special conditions.
2338 (group_cluster::~group_cluster): New.
2339 (group_cluster::group_cluster): Likewise.
2340 (group_cluster::dump): Likewise.
2341 (jump_table_cluster::emit): New.
2342 (switch_decision_tree::fix_phi_operands_for_edges): New.
2343 (struct case_node): Remove struct.
2344 (jump_table_cluster::can_be_handled): New.
2345 (case_values_threshold): Moved to header.
2346 (reset_out_edges_aux): Likewise.
2347 (jump_table_cluster::is_beneficial): New.
2348 (bit_test_cluster::can_be_handled): Likewise.
2349 (add_case_node): Remove.
2350 (bit_test_cluster::is_beneficial): New.
2351 (case_bit_test::cmp): New.
2352 (bit_test_cluster::emit): New.
2353 (expand_switch_as_decision_tree_p): Remove.
2354 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2355 (fix_phi_operands_for_edge): Likewise.
2356 (switch_decision_tree::analyze_switch_statement): New.
2357 (compute_cases_per_edge): Move ...
2358 (switch_decision_tree::compute_cases_per_edge): ... here.
2359 (try_switch_expansion): Likewise.
2360 (switch_decision_tree::try_switch_expansion): Likewise.
2361 (record_phi_operand_mapping): Likewise.
2362 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2363 (emit_case_decision_tree): Likewise.
2364 (switch_decision_tree::emit): Likewise.
2365 (balance_case_nodes): Likewise.
2366 (switch_decision_tree::balance_case_nodes): Likewise.
2367 (dump_case_nodes): Likewise.
2368 (switch_decision_tree::dump_case_nodes): Likewise.
2369 (emit_jump): Likewise.
2370 (switch_decision_tree::emit_jump): Likewise.
2371 (emit_cmp_and_jump_insns): Likewise.
2372 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2373 (emit_case_nodes): Likewise.
2374 (switch_decision_tree::emit_case_nodes): Likewise.
2375 (conditional_probability): Remove.
2376 * tree-switch-conversion.h (enum cluster_type): New.
2378 (struct cluster): Likewise.
2379 (cluster::cluster): Likewise.
2380 (struct simple_cluster): Likewise.
2381 (simple_cluster::simple_cluster): Likewise.
2382 (struct group_cluster): Likewise.
2383 (struct jump_table_cluster): Likewise.
2384 (struct bit_test_cluster): Likewise.
2385 (struct min_cluster_item): Likewise.
2386 (struct case_tree_node): Likewise.
2387 (case_tree_node::case_tree_node): Likewise.
2388 (jump_table_cluster::case_values_threshold): Likewise.
2389 (struct case_bit_test): Likewise.
2390 (struct switch_decision_tree): Likewise.
2391 (struct switch_conversion): Likewise.
2392 (switch_decision_tree::reset_out_edges_aux): Likewise.
2394 2018-06-20 Martin Liska <mliska@suse.cz>
2396 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2397 (hoist_edge_and_branch_if_true): Likewise.
2398 (expand_switch_using_bit_tests_p): Likewise.
2399 (struct case_bit_test): Likewise.
2400 (case_bit_test_cmp): Likewise.
2401 (emit_case_bit_tests): Likewise.
2402 (switch_conversion::switch_conversion): New class.
2403 (struct switch_conv_info): Remove old struct.
2404 (collect_switch_conv_info): More to ...
2405 (switch_conversion::collect): ... this.
2406 (check_range): Likewise.
2407 (switch_conversion::check_range): Likewise.
2408 (check_all_empty_except_final): Likewise.
2409 (switch_conversion::check_all_empty_except_final): Likewise.
2410 (check_final_bb): Likewise.
2411 (switch_conversion::check_final_bb): Likewise.
2412 (create_temp_arrays): Likewise.
2413 (switch_conversion::create_temp_arrays): Likewise.
2414 (free_temp_arrays): Likewise.
2415 (gather_default_values): Likewise.
2416 (switch_conversion::gather_default_values): Likewise.
2417 (build_constructors): Likewise.
2418 (switch_conversion::build_constructors): Likewise.
2419 (constructor_contains_same_values_p): Likewise.
2420 (switch_conversion::contains_same_values_p): Likewise.
2421 (array_value_type): Likewise.
2422 (switch_conversion::array_value_type): Likewise.
2423 (build_one_array): Likewise.
2424 (switch_conversion::build_one_array): Likewise.
2425 (build_arrays): Likewise.
2426 (switch_conversion::build_arrays): Likewise.
2427 (gen_def_assigns): Likewise.
2428 (switch_conversion::gen_def_assigns): Likewise.
2429 (prune_bbs): Likewise.
2430 (switch_conversion::prune_bbs): Likewise.
2431 (fix_phi_nodes): Likewise.
2432 (switch_conversion::fix_phi_nodes): Likewise.
2433 (gen_inbound_check): Likewise.
2434 (switch_conversion::gen_inbound_check): Likewise.
2435 (process_switch): Use the newly created class.
2436 (switch_conversion::expand): New.
2437 (switch_conversion::~switch_conversion): New.
2438 * tree-switch-conversion.h: New file.
2440 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2442 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2443 tree-vect-patterns.c.
2444 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2445 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2446 (vect_recog_sad_pattern): Likewise.
2447 (vect_recog_widen_sum_pattern): Likewise.
2448 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2449 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2450 (vect_recog_rotate_pattern): Likewise.
2451 (vect_recog_mult_pattern): Likewise.
2452 (vect_recog_vector_vector_shift_pattern): Likewise.
2453 (vect_recog_divmod_pattern): Likewise.
2454 (vect_recog_mixed_size_cond_pattern): Likewise.
2455 (vect_recog_bool_pattern): Likewise.
2456 (vect_recog_mask_conversion_pattern): Likewise.
2457 (vect_try_gather_scatter_pattern): Likewise.
2458 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2459 (vect_recog_over_widening_pattern): Likewise.
2460 (vect_recog_gather_scatter_pattern): Likewise.
2461 (vect_recog_func_ptr): Move from tree-vectorizer.h
2462 (vect_vect_recog_func_ptrs): Move further down the file.
2463 (vect_recog_func): Likewise. Remove the third argument.
2464 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2465 (vect_pattern_recog_1): Expect the pattern function to do any
2466 necessary target tests. Also expect it to provide a vector type.
2467 Remove the type_in handling.
2469 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2471 * tree-vect-patterns.c (vect_pattern_detected): New function.
2472 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2473 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2474 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2475 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2476 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2477 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2478 (vect_recog_mask_conversion_pattern)
2479 (vect_try_gather_scatter_pattern): Likewise.
2481 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2483 * tree-vect-patterns.c (vect_get_internal_def): New function.
2484 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2485 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2486 (search_type_for_mask_1): Use it.
2488 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2490 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2491 redundant WIDEN_SUM_EXPR handling.
2492 (vect_recog_sad_pattern): Likewise.
2494 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2496 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2497 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2498 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2499 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2500 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2502 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2504 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2505 use the stmt_vec_info of the original bb statement for the
2506 new zero assignment, even if the call is part of a pattern.
2508 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2510 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2511 that the sequence is attached to the original statement rather
2512 than the pattern statement.
2513 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2514 PATTERN_DEF_SEQ from the original statement rather than
2515 the main pattern statement.
2516 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2517 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2518 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2520 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2522 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2523 definition statements before the early exit for statements that aren't
2525 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2527 (vect_transform_loop): ...here. Process pattern definition
2528 statements without first checking whether the main pattern
2529 statement is live or relevant.
2531 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2533 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2534 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2536 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2538 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2539 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2540 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2541 (expand_block_compare): Change select_block_compare_mode call.
2542 (expand_strncmp_align_check): Use new functions, fix comment.
2543 (emit_final_str_compare_gpr): New function.
2544 (expand_strn_compare): Refactor and clean up code.
2545 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2547 2018-06-19 Tony Reix <tony.reix@atos.com>
2548 Damien Bergamini <damien.bergamini@atos.com>
2549 David Edelsohn <dje.gcc@gmail.com>
2551 * collect2.c (static_obj): New variable.
2552 (static_libs): New variable.
2553 (is_in_list): Uncomment declaration.
2554 (main): Track AIX libraries linked statically.
2555 (is_in_list): Uncomment definition.
2556 (scan_prog_file): Don't add AIX shared libraries initializer
2557 to constructor list if linking statically.
2559 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2561 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2563 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2565 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2567 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2570 2018-06-19 Martin Liska <mliska@suse.cz>
2572 * config/i386/i386.c (ix86_can_inline_p): Do not use
2573 ipa_fn_summaries::get_create.
2574 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2576 (devirtualization_time_bonus): Likewise.
2577 (ipcp_propagate_stage): Likewise.
2578 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2579 (edge_set_predicate): Likewise.
2580 (evaluate_conditions_for_known_args): Likewise.
2581 (evaluate_properties_for_edge): Likewise.
2582 (ipa_call_summary::reset): Tranform to ...
2583 (ipa_call_summary::~ipa_call_summary): ... this.
2584 (ipa_fn_summary::reset): Transform to ...
2585 (ipa_fn_summary::~ipa_fn_summary): ... this.
2586 (ipa_fn_summary_t::remove): Rename to ...
2587 (ipa_fn_summary_t::remove_callees): ... this.
2588 (ipa_fn_summary_t::duplicate): Use placement new
2589 instead of memory copy.
2590 (ipa_call_summary_t::duplicate): Likewise.
2591 (ipa_call_summary_t::remove): Remove.
2592 (dump_ipa_call_summary): Change get_create to get.
2593 (ipa_dump_fn_summary): Dump only when summary exists.
2594 (analyze_function_body): Use symbol_summary::get instead
2596 (compute_fn_summary): Likewise.
2597 (estimate_edge_devirt_benefit): Likewise.
2598 (estimate_edge_size_and_time): Likewise.
2599 (inline_update_callee_summaries): Likewise.
2600 (remap_edge_change_prob): Likewise.
2601 (remap_edge_summaries): Likewise.
2602 (ipa_merge_fn_summary_after_inlining): Likewise.
2603 (write_ipa_call_summary): Likewise.
2604 (ipa_fn_summary_write): Likewise.
2605 (ipa_free_fn_summary): Likewise.
2606 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2607 (struct ipa_call_summary): Likewise.
2608 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2610 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2611 (estimate_size_after_inlining): Likewise.
2612 (estimate_growth): Likewise.
2613 (growth_likely_positive): Likewise.
2614 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2615 (inline_call): Likewise.
2616 * ipa-inline.c (caller_growth_limits): Likewise.
2617 (can_inline_edge_p): Likewise.
2618 (can_inline_edge_by_limits_p): Likewise.
2619 (compute_uninlined_call_time): Likewise.
2620 (compute_inlined_call_time): Likewise.
2621 (want_inline_small_function_p): Likewise.
2622 (edge_badness): Likewise.
2623 (update_caller_keys): Likewise.
2624 (update_callee_keys): Likewise.
2625 (inline_small_functions): Likewise.
2626 (inline_to_all_callers_1): Likewise.
2627 (dump_overall_stats): Likewise.
2628 (early_inline_small_functions): Likewise.
2629 (early_inliner): Likewise.
2630 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2631 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2632 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2633 * ipa-split.c (execute_split_functions): Likewise.
2634 * symbol-summary.h: Likewise.
2635 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2637 2018-06-19 Richard Biener <rguenther@suse.de>
2639 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2640 (vectorize_loops): ... here. Fix dbgcnt handling.
2641 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2643 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2646 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2647 ieee128 argument takes up only one (vector) register, not two (floating
2650 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2652 * gimplify.c (gimplify_init_constructor): Really never clear for an
2653 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2655 2018-06-19 Richard Biener <rguenther@suse.de>
2657 PR tree-optimization/86179
2658 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2659 after failed recognition.
2661 2018-06-18 Martin Sebor <msebor@redhat.com>
2664 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2665 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2666 Handle integer subtraction.
2667 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2668 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2670 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2672 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2673 param from rtx to rtx_insn *.
2674 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2676 (frv_ifcvt_modify_insn): Likwise.
2677 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2678 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2679 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2680 as_a <rtx_insn *> cast to local "unprotected_region" once
2681 it's been established that it's not NULL or pc_rtx.
2682 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2683 param "sethi" from rtx to rtx_insn *.
2684 (nds32_group_float_insns): Likewise for param "insn".
2685 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2687 (vax_output_int_subtract): Likewise.
2688 * config/vax/vax.c (vax_output_int_add): Likewise for param
2690 (vax_output_int_subtract): Likewise.
2691 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2692 (emit_pattern_after): Likewise for param "after".
2693 (emit_insn_after): Likewise.
2694 (emit_jump_insn_after): Likewise.
2695 (emit_call_insn_after): Likewise.
2696 (emit_debug_insn_after): Likewise.
2697 (emit_pattern_before): Likewise for param "before".
2698 (emit_insn_before): Likewise.
2699 (emit_jump_insn_before): Likewise.
2700 * final.c (get_insn_template): Likewise for param "insn", removing
2702 * output.h (get_insn_template): Likewise for 2nd param.
2703 * rtl.h (emit_insn_before): Likewise.
2704 (emit_jump_insn_before): Likewise.
2705 (emit_debug_insn_before_noloc): Likewise.
2706 (emit_insn_after): Likewise.
2707 (emit_jump_insn_after): Likewise.
2708 (emit_call_insn_after): Likewise.
2709 (emit_debug_insn_after): Likewise.
2710 (set_insn_deleted): Likewise for param.
2712 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2715 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2716 floating point modes, so that IFmode is numerically greater than
2717 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2718 to declare the ordering. This prevents IFmode from being
2719 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2720 machine. Include rs6000-modes.h to share the fractional values
2721 between genmodes* and the rest of the compiler.
2725 * config/rs6000/rs6000-modes.h: New file.
2726 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2727 meaning of rs6000_long_double_size so that 126..128 selects an
2728 appropriate 128-bit floating point type.
2729 (rs6000_option_override_internal): Likewise.
2730 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2731 (TARGET_LONG_DOUBLE_128): Change the meaning of
2732 rs6000_long_double_size so that 126..128 selects an appropriate
2733 128-bit floating point type.
2734 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2735 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2736 source and destination to match the standard usage.
2737 (truncifkf2): Likewise.
2738 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2739 ISA 2.07 to use an explicit clobber, instead of passing in a
2741 (copysign<mode>3_soft): Likewise.
2743 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2745 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2746 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2747 (vect_slp_analyze_instance_dependence): Likewise.
2748 (vect_enhance_data_refs_alignment): Likewise.
2749 (vect_analyze_data_refs_alignment): Likewise.
2750 (vect_slp_analyze_and_verify_instance_alignment
2751 (vect_analyze_data_ref_accesses): Likewise.
2752 (vect_prune_runtime_alias_test_list): Likewise.
2753 (vect_analyze_data_refs): Likewise.
2754 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2755 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2756 (vect_analyze_scalar_cycles_1): Likewise.
2757 (vect_get_loop_niters): Likewise.
2758 (vect_analyze_loop_form_1): Likewise.
2759 (vect_update_vf_for_slp): Likewise.
2760 (vect_analyze_loop_operations): Likewise.
2761 (vect_analyze_loop): Likewise.
2762 (vectorizable_induction): Likewise.
2763 (vect_transform_loop): Likewise.
2764 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2765 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2766 (vect_make_slp_decision): Likewise.
2767 (vect_detect_hybrid_slp): Likewise.
2768 (vect_slp_analyze_operations): Likewise.
2769 (vect_slp_bb): Likewise.
2770 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2771 (vectorizable_bswap): Likewise.
2772 (vectorizable_call): Likewise.
2773 (vectorizable_simd_clone_call): Likewise.
2774 (vectorizable_conversion): Likewise.
2775 (vectorizable_assignment): Likewise.
2776 (vectorizable_shift): Likewise.
2777 (vectorizable_operation): Likewise.
2778 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2780 2018-06-18 Martin Sebor <msebor@redhat.com>
2782 PR tree-optimization/81384
2783 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2784 * builtins.c (expand_builtin_strnlen): New function.
2785 (expand_builtin): Call it.
2786 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2787 * builtins.def (BUILT_IN_STRNLEN): New.
2788 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2789 Warn for bounds in excess of maximum object size.
2790 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2791 single-value ranges. Handle strnlen.
2792 (handle_builtin_strlen): Handle strnlen.
2793 (strlen_check_and_optimize_stmt): Same.
2794 * doc/extend.texi (Other Builtins): Document strnlen.
2796 2018-06-18 Maya Rashish <coypu@sdf.org>
2798 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2799 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2800 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2802 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2804 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2806 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2808 * tree.c (escaped_string::escape): Replace cast to char * by
2809 const_cast<char *> (unescaped).
2811 2018-06-18 Nick Clifton <nickc@redhat.com>
2814 * tree.c (escaped_string): New class. Converts an unescaped
2815 string into its escaped equivalent.
2816 (warn_deprecated_use): Use the new class to convert the
2817 deprecation message, if present.
2818 (test_escaped_strings): New self test.
2819 (test_c_tests): Add test_escaped_strings.
2820 * doc/extend.texi (deprecated): Add a note that the
2821 deprecation message is affected by the -fmessage-length
2822 option, and that control characters will be escaped.
2823 (#pragma GCC error): Document this pragma.
2824 (#pragma GCC warning): Likewise.
2825 * doc/invoke.texi (-fmessage-length): Document this option's
2826 effect on the #warning and #error preprocessor directives and
2827 the deprecated attribute.
2829 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2831 * tree.c (decl_value_expr_lookup): Revert latest change.
2832 (decl_value_expr_insert): Likewise.
2834 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2836 * gimplify.c (nonlocal_vlas): Delete.
2837 (nonlocal_vla_vars): Likewise.
2838 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2840 (gimplify_body): Do not create and destroy nonlocal_vlas.
2841 * tree-nested.c: Include diagnostic.h.
2842 (use_pointer_in_frame): Tweak.
2843 (lookup_field_for_decl): Add assertion and declare the transformation.
2844 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2845 internal error when the reference is in a wrong context. Do not
2846 create a debug decl by default.
2847 (note_nonlocal_block_vlas): Delete.
2848 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2849 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2850 create a debug decl by default.
2851 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2852 call is in a wrong context.
2853 (fixup_vla_decls): New function.
2854 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2855 debug variables were created.
2856 * tree.c (decl_value_expr_lookup): Add checking assertion.
2857 (decl_value_expr_insert): Likewise.
2859 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2862 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2863 * tree-scalar-evolution.c (interpret_expr): Likewise.
2864 (expression_expensive_p): Likewise.
2865 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2866 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2867 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2868 (ssa_defined_by_minus_one_stmt_p): New.
2870 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2873 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2874 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2875 * dojump.c (do_jump): Likewise.
2876 * expr.c (expand_expr_real_2): Check operand type's sign.
2877 * fold-const.c (const_unop): Handle ABSU_EXPR.
2878 (fold_abs_const): Likewise.
2879 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2880 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2881 (strip_sign_op_1): Likesise.
2882 * match.pd: Add new pattern to generate ABSU_EXPR.
2883 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2884 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2885 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2886 * tree-inline.c (estimate_operator_cost): Likewise.
2887 * tree-pretty-print.c (dump_generic_node): Likewise.
2888 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2889 * tree.def (ABSU_EXPR): New.
2891 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2894 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2895 documented as preserved for backward compatibility only.
2896 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2898 PR rtl-optimization/86108
2899 * bb-reorder.c (create_forwarder_block): Renamed to ...
2900 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2901 jump from new landing pad to the second part.
2902 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2905 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2908 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2909 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2910 Only call store_expr for halves if the mode is the same.
2913 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2914 Fix up comment formatting.
2916 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2918 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2919 (typed_splay_tree::closure,
2920 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2921 (typed_splay_tree::typed_splay_tree,
2922 typed_splay_tree::operator =): Declared private.
2923 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2924 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2925 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2926 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2927 typed_splay_tree::splay_tree_splay,
2928 typed_splay_tree::splay_tree_foreach_helper,
2929 typed_splay_tree::splay_tree_insert,
2930 typed_splay_tree::splay_tree_remove,
2931 typed_splay_tree::splay_tree_lookup,
2932 typed_splay_tree::splay_tree_predecessor,
2933 typed_splay_tree::splay_tree_successor,
2934 typed_splay_tree::splay_tree_min,
2935 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2936 (typed_splay_tree::root, typed_splay_tree::comp,
2937 typed_splay_tree::delete_key,
2938 typed_splay_tree::delete_value): New data members.
2939 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2940 typed_splay_tree::remove.
2942 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2944 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2945 -mginv and -mno-ginv to the assembler.
2946 * config/mips/mips.opt (-mcrc): New option.
2948 * doc/invoke.text (-mcrc): Document.
2951 2018-06-15 Nick Clifton <nickc@redhat.com>
2954 * tree.c (escaped_string): New class. Converts an unescaped
2955 string into its escaped equivalent.
2956 (warn_deprecated_use): Use the new class to convert the
2957 deprecation message, if present.
2958 (test_escaped_strings): New self test.
2959 (test_c_tests): Add test_escaped_strings.
2960 * doc/extend.texi (deprecated): Add a note that the
2961 deprecation message is affected by the -fmessage-length
2962 option, and that control characters will be escaped.
2963 (#pragma GCC error): Document this pragma.
2964 (#pragma GCC warning): Likewise.
2965 * doc/invoke.texi (-fmessage-length): Document this option's
2966 effect on the #warning and #error preprocessor directives and
2967 the deprecated attribute.
2969 2018-06-15 Richard Biener <rguenther@suse.de>
2971 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2972 here, also noting vector size used.
2973 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2974 size used in MSG_OPTIMIZED_LOCATIONS dump.
2975 (pass_slp_vectorize::execute): Adjust.
2977 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2980 * config/arc/arc.c (arc_return_address_register): Fix
2983 2018-06-15 Richard Biener <rguenther@suse.de>
2986 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2987 leave useless conversion stripping to force_gimple_operand_gsi.
2988 (gimplify_build2): Likewise.
2989 (gimplify_build1): Likewise.
2991 2018-06-15 Richard Biener <rguenther@suse.de>
2994 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2995 before adjusting their block.
2997 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2999 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3000 multilibs for *-*-rtems*.
3001 * config/riscv/t-rtems: New file.
3003 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3006 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3007 unsigned_type_for returns NULL.
3010 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3011 subregs of multi-word pseudos unless the float mode has word size.
3013 2018-06-14 Richard Biener <rguenther@suse.de>
3016 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3017 duplicate and harmful type_hash_canon.
3018 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3020 2018-06-14 Richard Biener <rguenther@suse.de>
3023 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3026 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3028 * config/rtems.h (STDINT_LONG32): Define.
3030 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3031 Prachi Godbole <prachi.godbole@imgtec.com>
3033 * config/mips/mips-cpus.def: Define P6600.
3034 * config/mips/mips-tables.opt: Regenerate.
3035 * config/mips/mips.c (mips_ucbranch_type): New enum.
3036 (mips_rtx_cost_data): Add support for P6600.
3037 (mips_issue_rate): Likewise.
3038 (mips_multipass_dfa_lookahead): Likewise.
3039 (mips_avoid_hazard): Likewise.
3040 (mips_reorg_process_insns): Likewise.
3041 (mips_classify_branch_p6600): New function.
3042 * config/mips/mips.h (TUNE_P6600): New define.
3043 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3044 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3045 * config/mips/mips.md: Include p6600.md.
3046 (processor): Add p6600.
3047 * config/mips/p6600.md: New file.
3048 * doc/invoke.texi: Add p6600 to supported architectures.
3050 2018-06-13 Martin Sebor <msebor@redhat.com>
3052 PR tree-optimization/86114
3053 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3055 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3057 2018-06-13 Richard Biener <rguenther@suse.de>
3059 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3060 Properly set vector type of the intermediate stmt.
3061 * tree-vect-stmts.c (vectorizable_operation): The destination
3062 var always has vectype_out type.
3064 2018-06-13 Jeff Law <law@redhat.com>
3066 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3067 integer 0 for argument to print_rtl_with_bb.
3068 (rl78_reorg): Likewise.
3070 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3072 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3073 from rtx to rtx_insn *.
3074 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3076 (add_sched_insns_for_speculation): Likewise for local "target",
3077 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3078 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3079 from rtx_insn ** to rtx_code_label **.
3080 (reorg_emit_nops): Likewise.
3081 (c6x_reorg): Likewise for local "call_labels".
3082 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3084 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3085 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3086 the loops over LABEL_REFS.
3087 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3089 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3090 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3091 (split_branches): Strengthen local "olabel" from rtx to
3092 rtx_insn *, adding a safe_as_a cast.
3093 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3095 (add_insn_after): Likewise for first two params.
3096 (add_insn_before): Likewise.
3097 (remove_insn): Likewise for param.
3098 (emit_pattern_before_noloc): Likewise for second and third params.
3099 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3100 (emit_call_insn_before_noloc): Likewise.
3101 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3103 (emit_barrier_before): Likewise.
3104 (emit_label_before): Strengthen "label" param from "rtx" to
3105 "rtx_code_label *". Strengthen "before" param from "rtx" to
3107 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3109 (emit_pattern_after_noloc): Likewise.
3110 (emit_insn_after_noloc): Likewise.
3111 (emit_jump_insn_after_noloc): Likewise.
3112 (emit_call_insn_after_noloc): Likewise.
3113 (emit_debug_insn_after_noloc): Likewise.
3114 (emit_barrier_after): Likewise.
3115 (emit_label_after): Likewise for both params.
3116 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3117 "loc" param from "int" to "location_t".
3118 (emit_insn_after_setloc): Likewise.
3119 (emit_jump_insn_after_setloc): Likewise.
3120 (emit_call_insn_after_setloc): Likewise.
3121 (emit_debug_insn_after_setloc): Likewise.
3122 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3123 "loc" param from "int" to "location_t".
3124 (emit_pattern_before): Convert NULL_RTX to NULL.
3125 (emit_insn_before_setloc): Convert "loc" param from "int" to
3127 (emit_jump_insn_before_setloc): Likewise.
3128 (emit_call_insn_before_setloc): Likewise.
3129 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3130 rtx_insn *. Convert "loc" param from "int" to "location_t".
3131 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3132 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3133 Convert 3rd param from "int" to "location_t".
3134 (emit_barrier_before, emit_barrier_after, next_real_insn):
3135 Strengthen param from rtx to rtx_insn *.
3136 (emit_label_before): Strengthen 1st param from "rtx" to
3137 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3139 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3140 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3141 Strengthen 2nd param from "rtx" to "rtx_insn *".
3142 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3143 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3144 Likewise. Convert 3rd param from "int" to "location_t".
3145 (emit_label_after): Strengthen 1st param from "rtx" to
3147 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3149 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3150 from "rtx" to "rtx_insn *".
3152 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3154 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3155 bodies streamed in with -Q.
3156 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3157 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3158 * gimple-streamer-out.c: Include gimple-pretty-print.h
3159 (output_bb): Dump stmts streamed.
3160 * lto-section-out.c: Include print-tree.h
3161 (lto_begin_section): Dump sections created.
3162 (lto_output_decl_index): Dump decl encoded.
3163 * lto-streamer-out.c: Include print-tree.h
3164 (create_output_block): Dump output block created.
3165 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3166 (output_function): Dump function output.
3167 (output_constructor): Dump constructor streamed.
3168 (write_global_stream): Output indexes encoded.
3169 (produce_asm_for_decls): Dump streams encoded.
3170 * lto-streamer.c (streamer_dump_file): New global var.
3171 * lto-streamer.h (streamer_dump_file): Declare.
3172 * passes.c (ipa_write_summaries): Initialize streamer dump.
3173 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3176 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3179 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3180 offsets for register save directives. Emit a second batch of save
3181 directives, if need be, when the function accesses prior frames.
3183 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3185 * config/arc/fpu.md (fmasf4): Force operand to register.
3186 (fnmasf4): Likewise.
3188 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3190 * config/arc/arc-protos.h (arc_pad_return): Remove.
3191 * config/arc/arc.c (machine_function): Remove force_short_suffix
3193 (arc_print_operand): Adjust printing of '&'.
3194 (arc_verify_short): Remove conditional printing of short suffix.
3195 (arc_final_prescan_insn): Remove reference to size_reason.
3196 (pad_return): New function.
3197 (arc_reorg): Call pad_return.
3198 (arc_pad_return): Remove.
3199 (arc_init_machine_status): Remove reference to force_short_suffix.
3200 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3201 (attr length): When attribute iscompact is true force to 2
3202 regardless; in the case of maybe check if we want to force the
3203 instruction to have 4 bytes length.
3204 (nopv): Change it to generate 4 byte long nop as well.
3205 (blockage): New pattern.
3206 (simple_return): Remove call to arc_pad_return.
3207 (p_return_i): Likewise.
3209 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3211 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3213 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3215 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3218 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3220 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3221 for ARC700 and ARCv2.
3223 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3226 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3227 operands[2] instead of operands[1].
3230 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3232 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3233 case, check whether the outer register overlaps an unallocatable
3234 register, not just whether it fits the required class.
3236 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3238 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3240 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3241 (memrefs_conflict_p): Likewise.
3242 (init_alias_analysis): Likewise.
3243 * cfgexpand.c (expand_debug_expr): Likewise.
3244 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3245 * cse.c (fold_rtx): Likewise.
3246 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3247 * expr.c (emit_block_move_hints): Likewise.
3248 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3249 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3250 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3251 (emit_group_store): Likewise.
3252 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3253 to read the PRE/POST_MODIFY increment.
3254 * calls.c (store_one_arg): Use strip_offset.
3255 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3257 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3259 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3260 (simplify_binary_operation_1): Extend CONST_INT handling to
3262 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3263 than a HOST_WIDE_INT.
3264 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3266 (adjust_mems, add_stores): Update accodingly.
3267 (vt_canonicalize_addr): Track polynomial offsets.
3268 (emit_note_insn_var_location): Likewise.
3269 (vt_add_function_parameter): Likewise.
3270 (vt_initialize): Likewise.
3272 2018-06-12 Jeff Law <law@redhat.com>
3274 * config.gcc (alpha*-*-freebsd*): Remove.
3275 * config/alpha/freebsd.h: Remove.
3277 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3280 * spellcheck-tree.c (levenshtein_distance): Rename to...
3281 (get_edit_distance): ...this, and update for underlying renaming.
3282 * spellcheck-tree.h (levenshtein_distance): Rename to...
3283 (get_edit_distance): ...this.
3284 * spellcheck.c (levenshtein_distance): Rename to...
3285 (get_edit_distance): ...this. Convert from Levenshtein distance
3286 to Damerau-Levenshtein distance by supporting transpositions of
3287 adjacent characters. Rename "v1" to "v_next" and "v0" to
3289 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3290 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3291 update for underlying renaming.
3292 (selftest::levenshtein_distance_unit_test): Rename to...
3293 (selftest::test_get_edit_distance_unit): ...this, and update for
3294 underlying renaming.
3295 (selftest::test_find_closest_string): Add example from PR 69968
3296 where transposition helps
3297 (selftest::test_metric_conditions): Update for renaming.
3298 (selftest::test_metric_conditions): Likewise.
3299 (selftest::spellcheck_c_tests): Likewise.
3300 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3301 (get_edit_distance): ...this.
3302 (best_match::consider): Update for renaming.
3304 2018-06-12 Martin Sebor <msebor@redhat.com>
3306 PR tree-optimization/85259
3307 * builtins.c (compute_objsize): Handle constant offsets.
3308 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3309 true iff a warning has been issued.
3310 * gimple.h (gimple_nonartificial_location): New function.
3311 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3312 gimple_nonartificial_location and handle -Wno-system-headers.
3313 (handle_builtin_stxncpy): Same.
3315 2018-06-12 Martin Sebor <msebor@redhat.com>
3318 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3320 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3322 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3323 BUILTIN_VEC_XST entries for pointer to double and long long.
3325 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3328 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3330 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3333 2018-06-12 Martin Liska <mliska@suse.cz>
3335 * doc/options.texi: Document IntegerRange.
3337 2018-06-12 Martin Liska <mliska@suse.cz>
3339 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3340 * opt-functions.awk: Handle Deprecated flag.
3341 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3343 (read_cmdline_option): Report warning for a deprecated option.
3344 * opts.h (struct cl_option): Add new field cl_deprecated.
3345 (CL_ERR_DEPRECATED): New.
3347 2018-06-12 Martin Liska <mliska@suse.cz>
3349 * doc/options.texi: Document Deprecated option flag.
3351 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3353 * config/arc/arc-arch.h (arc_extras): New enum.
3354 (arc_cpu_t):Add field extra.
3355 (arc_cpu_types): Consider the extras.
3356 * config/arc/arc-cpus.def: Add extras info.
3357 * config/arc/arc-opts.h (processor_type): Consider extra field.
3358 * config/arc/arc.c (arc_override_options): Handle extra field.
3360 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3362 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3363 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3365 * config/arc/arc.c (arc_init): Update pic warning.
3366 (irq_range): Update irq range parsing warnings.
3367 (arc_override_options): Update various warning messages.
3368 (arc_handle_aux_attribute): Likewise.
3370 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3372 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3374 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3376 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3377 compiler messages directives.
3379 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3381 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3382 * config/mips/mips-tables.opt: Regenerate.
3383 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3385 * doc/invoke.texi: Document -march=i6500.
3387 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3389 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3390 (i6400_gpmul): Add cpu_unit.
3391 (i6400_gpdiv): Likewise.
3392 (i6400_msa_add_d): Update reservations.
3393 (i6400_msa_int_add) Likewise.
3394 (i6400_msa_short_logic3) Likewise.
3395 (i6400_msa_short_logic2) Likewise.
3396 (i6400_msa_short_logic) Likewise.
3397 (i6400_msa_move) Likewise.
3398 (i6400_msa_cmp) Likewise.
3399 (i6400_msa_short_float2) Likewise.
3400 (i6400_msa_div_d) Likewise.
3401 (i6400_msa_long_logic1) Likewise.
3402 (i6400_msa_long_logic2) Likewise.
3403 (i6400_msa_mult) Likewise.
3404 (i6400_msa_long_float2) Likewise.
3405 (i6400_msa_long_float4) Likewise.
3406 (i6400_msa_long_float5) Likewise.
3407 (i6400_msa_long_float8) Likewise.
3408 (i6400_fpu_fadd): Include frint type.
3409 (i6400_fpu_store): New define_insn_reservation.
3410 (i6400_fpu_load): Likewise.
3411 (i6400_fpu_move): Likewise.
3412 (i6400_fpu_fcmp): Likewise.
3413 (i6400_fpu_fmadd): Likewise.
3414 (i6400_int_mult): Include imul3nc type and update reservation.
3415 (i6400_int_div): Include idiv3 type and update reservation.
3416 (i6400_int_load): Update to check type not move_type.
3417 (i6400_int_store): Likewise.
3418 (i6400_int_prefetch): Set zero latency.
3420 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3422 * gcc.c: Document new %@{...} sequence.
3423 (LINK_COMMAND_SPEC): Use it for the -L switches.
3424 (cpp_unique_options): Use it for the -I switches.
3425 (at_file_argbuf): New global variable.
3426 (in_at_file): Likewise.
3427 (alloc_args): Create at_file_argbuf.
3428 (clear_args): Truncate at_file_argbuf.
3429 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3430 (open_at_file): New function.
3431 (close_at_file): Likewise.
3432 (create_at_file): Delete.
3433 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3436 (validate_switches_from_spec): Deal with %@{...} sequence.
3437 (validate_switches): Likewise.
3438 (driver::finalize): Call clear_args.
3440 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3442 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3444 2018-06-11 Martin Sebor <msebor@redhat.com>
3446 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3447 (Wno-class-memaccess): Add @opindex.
3448 (Wno-templates, Wno-multiple-inheritance): Same.
3449 (Wno-virtual-inheritance, Wno-namespaces): Same.
3450 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3451 (Wno-format-overflow, Wno-format-truncation): Same.
3452 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3453 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3454 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3455 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3456 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3457 (Wno-misspelled-isr): Same.
3459 2018-06-11 Martin Sebor <msebor@redhat.com>
3461 * PR tree-optimization/86083
3462 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3464 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3466 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3468 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3471 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3472 on the correct operand.
3473 (*movdi_internal64): Ditto.
3475 2018-06-11 Martin Liska <mliska@suse.cz>
3477 PR tree-optimization/86089
3478 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3480 2018-06-11 Julia Koval <julia.koval@intel.com>
3482 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3483 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3484 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3486 2018-06-11 Olivier Hainque <hainque@adacore.com>
3488 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3489 for Ada with strict dwarf2.
3491 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3494 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3497 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3499 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3501 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3503 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3506 2018-06-08 Martin Liska <mliska@suse.cz>
3508 * tree-cfg.h (debug_function): Fix argument type to match
3511 2018-06-08 Martin Liska <mliska@suse.cz>
3513 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3514 Remove usage of MPX-related (and removed) fields.
3515 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3517 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3519 * cfg.c (debug): Use TDF_NONE rather than 0.
3520 * cfghooks.c (debug): Likewise.
3521 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3522 (struct dump_option_value_info): Convert to...
3523 (struct kv_pair): ...this template type.
3524 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3526 (optinfo_verbosity_options): Likewise.
3527 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3529 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3530 than int for "optgroup_flags" param.
3531 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3533 (dump_dec): Likewise.
3534 (dump_finish): Use TDF_NONE rather than 0.
3535 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3536 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3537 than 0. Update for change to option_ptr.
3538 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3539 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3540 0. Update for changes to optinfo_verbosity_options and
3542 (opt_info_switch_p): Convert optgroup_flags from int to
3544 (dump_basic_block): Use dump_flags_t rather than int
3545 for "dump_kind" param.
3546 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3547 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3548 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3549 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3550 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3551 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3552 TDF_NONE): Convert from macros to...
3553 (enum dump_flag): ...this new enum.
3554 (dump_flags_t): Update to use enum.
3555 (operator|, operator&, operator~, operator|=, operator&=):
3556 Implement for dump_flags_t.
3557 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3558 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3559 Convert from macros to...
3560 (enum optgroup_flag): ...this new enum.
3561 (optgroup_flags_t): New typedef.
3562 (operator|, operator|=): Implement for optgroup_flags_t.
3563 (struct dump_file_info): Convert field "alt_flags" to
3564 dump_flags_t. Convert field "optgroup_flags" to
3566 (dump_basic_block): Use dump_flags_t rather than int for param.
3567 (dump_generic_expr_loc): Likewise.
3568 (dump_dec): Likewise.
3569 (dump_register): Convert param "optgroup_flags" to
3571 (opt_info_enable_passes): Likewise.
3572 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3574 * gimple-pretty-print.c (debug): Likewise.
3575 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3576 (merged_store_group::apply_stores): Likewise.
3577 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3578 * gimple.c (verify_gimple_pp): Likewise.
3579 * graphite-poly.c (print_pbb_body): Likewise.
3580 * passes.c (pass_manager::register_one_dump_file): Convert
3581 local "optgroup_flags" to optgroup_flags_t.
3582 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3584 (debug_body): Likewise.
3585 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3586 to optgroup_flags_t.
3587 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3589 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3590 (convert_mult_to_fma): Likewise.
3591 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3592 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3593 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3594 "dump_kind" to dump_flags_t.
3596 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3598 * config/rs6000/rs6000.c (min, max): Delete.
3600 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3602 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3603 -mabi=spe and -mabi=no-spe.
3605 2018-06-08 Martin Liska <mliska@suse.cz>
3607 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3608 where we expect an existing summary.
3610 2018-06-08 Martin Liska <mliska@suse.cz>
3612 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3613 * ipa-inline.h (estimate_edge_growth): Likewise.
3615 2018-06-08 Martin Liska <mliska@suse.cz>
3617 * cgraph.c (function_version_hasher::hash): Use
3618 cgraph_node::get_uid ().
3619 (function_version_hasher::equal):
3620 * cgraph.h (cgraph_node::get_uid): New method.
3621 * ipa-inline.c (update_caller_keys): Use
3622 cgraph_node::get_uid ().
3623 (update_callee_keys): Likewise.
3624 * ipa-utils.c (searchc): Likewise.
3625 (ipa_reduced_postorder): Likewise.
3626 * lto-cgraph.c (input_node): Likewise.
3627 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3628 * symbol-summary.h (symtab_insertion): Likewise.
3629 (symtab_removal): Likewise.
3630 (symtab_duplication): Likewise.
3631 * tree-pretty-print.c (dump_function_header): Likewise.
3632 * tree-sra.c (convert_callers_for_node): Likewise.
3634 2018-06-08 Martin Liska <mliska@suse.cz>
3636 * cgraph.c (symbol_table::create_edge): Always assign a new
3638 (symbol_table::free_edge): Do not recycle numbers.
3639 * cgraph.h (cgraph_edge::get): New method.
3640 * symbol-summary.h (symtab_removal): Use it.
3641 (symtab_duplication): Likewise.
3642 (call_summary::hashable_uid): Remove.
3644 2018-06-08 Martin Liska <mliska@suse.cz>
3646 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3647 (initialize_growth_caches): Remove.
3648 (free_growth_caches): Likewise.
3649 (do_estimate_edge_time): Use edge_growth_cache.
3650 (do_estimate_edge_size): Likewise.
3651 (do_estimate_edge_hints): Likewise.
3652 * ipa-inline.c (reset_edge_caches): Likewise.
3653 (recursive_inlining): Likewise.
3654 (inline_small_functions): Likewise.
3655 * ipa-inline.h (initialize_growth_caches): Remove.
3656 (estimate_edge_size): Likewise.
3657 (estimate_edge_time): Likewise.
3658 (estimate_edge_hints): Likewise.
3659 (reset_edge_growth_cache): Likewise.
3660 * symbol-summary.h (call_summary::remove): New method.
3662 2018-06-08 Martin Liska <mliska@suse.cz>
3664 * ipa-cp.c (class edge_clone_summary): New summary.
3665 (grow_edge_clone_vectors): Remove.
3666 (ipcp_edge_duplication_hook): Remove.
3667 (class edge_clone_summary_t): New call_summary class.
3668 (ipcp_edge_removal_hook): Remove.
3669 (edge_clone_summary_t::duplicate): New function.
3670 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3671 (create_specialized_node): Likewise.
3672 (ipcp_driver): Initialize edge_clone_summaries and do not
3675 2018-06-08 Martin Liska <mliska@suse.cz>
3677 * symbol-summary.h (get): New function.
3678 (call_summary::m_initialize_when_cloning): New class member.
3680 2018-06-08 Martin Liska <mliska@suse.cz>
3682 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3683 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3684 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3685 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3686 (remove_cgraph_node_from_order): Use the removed_nodes set.
3687 (do_per_function_toporder): Likwise.
3688 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3689 instead of summary_uid.
3690 (symtab_removal): Likewise.
3691 (symtab_duplication): Likewise.
3693 2018-06-08 Martin Liska <mliska@suse.cz>
3695 * ipa-cp.c (ipcp_store_bits_results): Use
3696 ipcp_transformation_sum.
3697 (ipcp_store_vr_results): Likewise.
3698 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3700 (ipcp_transformation_initialize): ... this.
3701 (ipa_set_node_agg_value_chain):
3702 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3703 (write_ipcp_transformation_info): Likewise.
3704 (read_ipcp_transformation_info): Likewise.
3705 (ipcp_update_bits): Likewise.
3706 (ipcp_update_vr): Likewise.
3707 (ipcp_transform_function): Likewise.
3708 * ipa-prop.h: Rename ipcp_transformation_summary to
3709 ipcp_transformation.
3710 (class ipcp_transformation_t): New function summary.
3711 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3712 (ipa_get_agg_replacements_for_node): Likewise.
3714 2018-06-08 Martin Liska <mliska@suse.cz>
3716 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3718 (class funct_state_summary_t): New function_summary class.
3719 (has_function_state): Remove.
3720 (get_function_state): Likewise.
3721 (set_function_state): Likewise.
3722 (add_new_function): Likewise.
3723 (funct_state_summary_t::insert): New function.
3724 (duplicate_node_data): Remove.
3725 (remove_node_data): Remove.
3726 (funct_state_summary_t::duplicate): New function.
3727 (register_hooks): Create new funct_state_summaries.
3728 (pure_const_generate_summary): Use it.
3729 (pure_const_write_summary): Likewise.
3730 (pure_const_read_summary): Likewise.
3731 (propagate_pure_const): Likewise.
3732 (propagate_nothrow): Likewise.
3733 (dump_malloc_lattice): Likewise.
3734 (propagate_malloc): Likewise.
3735 (execute): Do not register hooks, just remove summary
3737 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3740 2018-06-08 Martin Liska <mliska@suse.cz>
3742 * ipa-reference.c (remove_node_data): Remove.
3743 (duplicate_node_data): Likewise.
3744 (class ipa_ref_var_info_summary_t): New class.
3745 (class ipa_ref_opt_summary_t): Likewise.
3746 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3747 (get_reference_optimization_summary): Use
3748 ipa_ref_opt_sum_summaries.
3749 (set_reference_vars_info): Remove.
3750 (set_reference_optimization_summary): Likewise.
3751 (ipa_init): Create summaries.
3752 (init_function_info): Use function summary.
3753 (ipa_ref_opt_summary_t::duplicate): New function.
3754 (ipa_ref_opt_summary_t::remove): New function.
3755 (get_read_write_all_from_node): Fix GNU coding style.
3756 (propagate): Use function summary.
3757 (write_node_summary_p): Fix GNU coding style.
3758 (stream_out_bitmap): Likewise.
3759 (ipa_reference_read_optimization_summary): Use function summary.
3760 (ipa_reference_c_finalize): Do not release hooks.
3762 2018-06-08 Martin Liska <mliska@suse.cz>
3764 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3765 (analyze_function_body): Extract multiple calls of get_create.
3766 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3767 * ipa-inline.c (recursive_inlining): Use ::get method.
3768 * ipa-inline.h (estimate_edge_growth): Likewise.
3770 2018-06-08 Martin Liska <mliska@suse.cz>
3772 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3774 (hsa_function_summary::hsa_function_summary): Use the new enum
3776 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3777 * hsa-gen.c (hsa_get_host_function): Likewise.
3778 (get_brig_function_name): Likewise.
3779 * ipa-hsa.c (process_hsa_functions): Likewise.
3780 (ipa_hsa_write_summary): Likewise.
3781 * symbol-summary.h (symtab_duplication): Use ::get function/
3782 (get): New function.
3784 2018-06-08 Martin Liska <mliska@suse.cz>
3786 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3788 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3789 (hsa_register_kernel): Likewise.
3790 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3791 * hsa-gen.c (hsa_get_host_function): Likewise.
3792 (get_brig_function_name): Likewise.
3793 (generate_hsa): Likewise.
3794 (pass_gen_hsail::execute): Likewise.
3795 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3796 (devirtualization_time_bonus): Likewise.
3797 (ipcp_propagate_stage): Likewise.
3798 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3799 (edge_set_predicate): Likewise.
3800 (evaluate_conditions_for_known_args): Likewise.
3801 (evaluate_properties_for_edge): Likewise.
3802 (ipa_fn_summary::reset): Likewise.
3803 (ipa_fn_summary_t::duplicate): Likewise.
3804 (dump_ipa_call_summary): Likewise.
3805 (ipa_dump_fn_summary): Likewise.
3806 (analyze_function_body): Likewise.
3807 (compute_fn_summary): Likewise.
3808 (estimate_edge_devirt_benefit): Likewise.
3809 (estimate_edge_size_and_time): Likewise.
3810 (estimate_calls_size_and_time): Likewise.
3811 (estimate_node_size_and_time): Likewise.
3812 (inline_update_callee_summaries): Likewise.
3813 (remap_edge_change_prob): Likewise.
3814 (remap_edge_summaries): Likewise.
3815 (ipa_merge_fn_summary_after_inlining): Likewise.
3816 (ipa_update_overall_fn_summary): Likewise.
3817 (read_ipa_call_summary): Likewise.
3818 (inline_read_section): Likewise.
3819 (write_ipa_call_summary): Likewise.
3820 (ipa_fn_summary_write): Likewise.
3821 (ipa_free_fn_summary): Likewise.
3822 * ipa-hsa.c (process_hsa_functions): Likewise.
3823 (ipa_hsa_write_summary): Likewise.
3824 (ipa_hsa_read_section): Likewise.
3825 * ipa-icf.c (sem_function::merge): Likewise.
3826 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3827 (do_estimate_edge_time): Likewise.
3828 (estimate_size_after_inlining): Likewise.
3829 (estimate_growth): Likewise.
3830 (growth_likely_positive): Likewise.
3831 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3832 (inline_call): Likewise.
3833 * ipa-inline.c (caller_growth_limits): Likewise.
3834 (can_inline_edge_p): Likewise.
3835 (can_inline_edge_by_limits_p): Likewise.
3836 (compute_uninlined_call_time): Likewise.
3837 (compute_inlined_call_time): Likewise.
3838 (want_inline_small_function_p): Likewise.
3839 (edge_badness): Likewise.
3840 (update_caller_keys): Likewise.
3841 (update_callee_keys): Likewise.
3842 (recursive_inlining): Likewise.
3843 (inline_small_functions): Likewise.
3844 (inline_to_all_callers_1): Likewise.
3845 (dump_overall_stats): Likewise.
3846 (early_inline_small_functions): Likewise.
3847 (early_inliner): Likewise.
3848 * ipa-inline.h (estimate_edge_growth): Likewise.
3849 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3850 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3851 * ipa-prop.h (IPA_NODE_REF): Likewise.
3852 (IPA_EDGE_REF): Likewise.
3853 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3854 (propagate_malloc): Likewise.
3855 * ipa-split.c (execute_split_functions): Likewise.
3856 * symbol-summary.h: Rename get to get_create.
3858 (get_create): Likewise.
3859 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3861 2018-06-08 Martin Liska <mliska@suse.cz>
3863 * symbol-summary.h (release): Move definition out of class
3865 (symtab_removal): Likewise.
3866 (symtab_duplication): Likewise.
3868 2018-06-08 Martin Liska <mliska@suse.cz>
3870 * symbol-summary.h (function_summary): Move constructor
3871 implementation out of class declaration.
3872 (release): Likewise.
3873 (symtab_insertion): Likewise.
3874 (symtab_removal): Likewise.
3875 (symtab_duplication): Likewise.
3878 2018-06-08 Martin Liska <mliska@suse.cz>
3880 * Makefile.in: Remove support for MPX (macros, related functions,
3881 fields in cgraph_node, ...).
3882 * builtin-types.def (BT_BND): Likewise.
3883 (BT_FN_BND_CONST_PTR): Likewise.
3884 (BT_FN_CONST_PTR_BND): Likewise.
3885 (BT_FN_VOID_PTR_BND): Likewise.
3886 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3887 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3888 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3889 (expand_builtin_mempcpy_with_bounds): Likewise.
3890 (expand_builtin_memset_with_bounds): Likewise.
3891 (expand_builtin_memset_args): Likewise.
3892 (std_expand_builtin_va_start): Likewise.
3893 (expand_builtin): Likewise.
3894 (expand_builtin_with_bounds): Likewise.
3895 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3896 (DEF_LIB_BUILTIN_CHKP): Likewise.
3897 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3898 (DEF_CHKP_BUILTIN): Likewise.
3899 (BUILT_IN_MEMCPY): Likewise.
3900 (BUILT_IN_MEMMOVE): Likewise.
3901 (BUILT_IN_MEMPCPY): Likewise.
3902 (BUILT_IN_MEMSET): Likewise.
3903 (BUILT_IN_STPCPY): Likewise.
3904 (BUILT_IN_STRCAT): Likewise.
3905 (BUILT_IN_STRCHR): Likewise.
3906 (BUILT_IN_STRCPY): Likewise.
3907 (BUILT_IN_STRLEN): Likewise.
3908 (BUILT_IN_MEMCPY_CHK): Likewise.
3909 (BUILT_IN_MEMMOVE_CHK): Likewise.
3910 (BUILT_IN_MEMPCPY_CHK): Likewise.
3911 (BUILT_IN_MEMSET_CHK): Likewise.
3912 (BUILT_IN_STPCPY_CHK): Likewise.
3913 (BUILT_IN_STRCAT_CHK): Likewise.
3914 (BUILT_IN_STRCPY_CHK): Likewise.
3915 * calls.c (store_bounds): Likewise.
3916 (emit_call_1): Likewise.
3917 (special_function_p): Likewise.
3918 (maybe_warn_nonstring_arg): Likewise.
3919 (initialize_argument_information): Likewise.
3920 (finalize_must_preallocate): Likewise.
3921 (compute_argument_addresses): Likewise.
3922 (expand_call): Likewise.
3923 * cfgexpand.c (expand_call_stmt): Likewise.
3924 (expand_return): Likewise.
3925 (expand_gimple_stmt_1): Likewise.
3926 (pass_expand::execute): Likewise.
3927 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3928 (cgraph_node::remove): Likewise.
3929 (cgraph_node::dump): Likewise.
3930 (cgraph_node::verify_node): Likewise.
3931 * cgraph.h (chkp_function_instrumented_p): Likewise.
3932 (symtab_node::get_alias_target): Likewise.
3933 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3934 (cgraph_local_p): Likewise.
3935 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3936 (cgraph_edge::rebuild_references): Likewise.
3937 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3938 (walk_polymorphic_call_targets): Likewise.
3939 (cgraph_node::expand_thunk): Likewise.
3940 (symbol_table::output_weakrefs): Likewise.
3941 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3942 (ix86_handle_option): Likewise.
3943 * config/i386/constraints.md: Likewise.
3944 * config/i386/i386-builtin-types.def (BND): Likewise.
3948 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3949 (BDESC_FIRST): Likewise.
3951 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3952 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3953 * config/i386/i386.c (enum reg_class): Likewise.
3954 (ix86_target_string): Likewise.
3955 (ix86_option_override_internal): Likewise.
3956 (ix86_conditional_register_usage): Likewise.
3957 (ix86_valid_target_attribute_inner_p): Likewise.
3958 (ix86_set_indirect_branch_type): Likewise.
3959 (ix86_set_current_function): Likewise.
3960 (ix86_function_arg_regno_p): Likewise.
3961 (init_cumulative_args): Likewise.
3962 (ix86_function_arg_advance): Likewise.
3963 (ix86_function_arg): Likewise.
3964 (ix86_pass_by_reference): Likewise.
3965 (ix86_function_value_regno_p): Likewise.
3966 (ix86_function_value_1): Likewise.
3967 (ix86_function_value_bounds): Likewise.
3968 (ix86_return_in_memory): Likewise.
3969 (ix86_setup_incoming_vararg_bounds): Likewise.
3970 (ix86_va_start): Likewise.
3971 (indirect_thunk_need_prefix): Likewise.
3972 (print_reg): Likewise.
3973 (ix86_print_operand): Likewise.
3974 (ix86_expand_call): Likewise.
3975 (ix86_output_function_return): Likewise.
3976 (reg_encoded_number): Likewise.
3977 (BDESC_VERIFYS): Likewise.
3978 (ix86_init_mpx_builtins): Likewise.
3979 (ix86_init_builtins): Likewise.
3980 (ix86_emit_cmove): Likewise.
3981 (ix86_emit_move_max): Likewise.
3982 (ix86_expand_builtin): Likewise.
3983 (ix86_builtin_mpx_function): Likewise.
3984 (ix86_get_arg_address_for_bt): Likewise.
3985 (ix86_load_bounds): Likewise.
3986 (ix86_store_bounds): Likewise.
3987 (ix86_load_returned_bounds): Likewise.
3988 (ix86_store_returned_bounds): Likewise.
3989 (ix86_class_likely_spilled_p): Likewise.
3990 (ix86_hard_regno_mode_ok): Likewise.
3991 (x86_order_regs_for_local_alloc): Likewise.
3992 (ix86_mitigate_rop): Likewise.
3993 (ix86_bnd_prefixed_insn_p): Likewise.
3994 (ix86_mpx_bound_mode): Likewise.
3995 (ix86_make_bounds_constant): Likewise.
3996 (ix86_initialize_bounds): Likewise.
3997 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3998 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3999 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4000 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4001 (TARGET_CHKP_BOUND_MODE): Likewise.
4002 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4003 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4004 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4005 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4006 * config/i386/i386.h (TARGET_MPX): Likewise.
4007 (TARGET_MPX_P): Likewise.
4008 (VALID_BND_REG_MODE): Likewise.
4009 (FIRST_BND_REG): Likewise.
4010 (LAST_BND_REG): Likewise.
4011 (enum reg_class): Likewise.
4012 (BND_REG_P): Likewise.
4013 (BND_REGNO_P): Likewise.
4014 (BNDmode): Likewise.
4015 (ADJUST_INSN_LENGTH): Likewise.
4016 * config/i386/i386.md: Likewise.
4017 * config/i386/i386.opt: Likewise.
4018 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4019 (defined): Likewise.
4020 (LINK_MPX): Likewise.
4021 (MPX_SPEC): Likewise.
4022 (LIBMPX_SPEC): Likewise.
4023 (LIBMPXWRAPPERS_SPEC): Likewise.
4024 (CHKP_SPEC): Likewise.
4025 * config/i386/predicates.md: Likewise.
4026 * dbxout.c (dbxout_type): Likewise.
4027 * doc/extend.texi: Likewise.
4028 * doc/invoke.texi: Likewise.
4029 * doc/md.texi: Likewise.
4030 * doc/tm.texi: Likewise.
4031 * doc/tm.texi.in: Likewise.
4032 * dwarf2out.c (is_base_type): Likewise.
4033 (gen_formal_types_die): Likewise.
4034 (gen_subprogram_die): Likewise.
4035 (gen_type_die_with_usage): Likewise.
4036 (gen_decl_die): Likewise.
4037 (dwarf2out_late_global_decl): Likewise.
4038 * expr.c (expand_assignment): Likewise.
4039 (emit_storent_insn): Likewise.
4040 (store_expr_with_bounds): Likewise.
4041 (store_expr): Likewise.
4042 (expand_expr_real_1): Likewise.
4043 * expr.h (store_expr_with_bounds): Likewise.
4044 * function.c (use_register_for_decl): Likewise.
4045 (struct bounds_parm_data): Likewise.
4046 (assign_parms_augmented_arg_list): Likewise.
4047 (assign_parm_find_entry_rtl): Likewise.
4048 (assign_parm_is_stack_parm): Likewise.
4049 (assign_parm_load_bounds): Likewise.
4050 (assign_bounds): Likewise.
4051 (assign_parms): Likewise.
4052 (expand_function_start): Likewise.
4053 * gcc.c (CHKP_SPEC): Likewise.
4054 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4055 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4056 (wrestrict_dom_walker::check_call): Likewise.
4057 * gimple.c (gimple_build_call_from_tree): Likewise.
4058 * gimple.h (enum gf_mask): Likewise.
4059 (gimple_call_with_bounds_p): Likewise.
4060 (gimple_call_set_with_bounds): Likewise.
4061 * gimplify.c (gimplify_init_constructor): Likewise.
4062 * ipa-cp.c (initialize_node_lattices): Likewise.
4063 (propagate_constants_across_call): Likewise.
4064 (find_more_scalar_values_for_callers_subset): Likewise.
4065 * ipa-hsa.c (process_hsa_functions): Likewise.
4066 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4067 * ipa-icf.c (sem_function::merge): Likewise.
4068 * ipa-inline.c (early_inliner): Likewise.
4069 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4070 (warn_function_cold): Likewise.
4071 (propagate_pure_const): Likewise.
4072 * ipa-ref.h (enum GTY): Likewise.
4073 * ipa-split.c (find_retbnd): Likewise.
4074 (consider_split): Likewise.
4075 (split_function): Likewise.
4076 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4077 * ipa.c (walk_polymorphic_call_targets): Likewise.
4078 (symbol_table::remove_unreachable_nodes): Likewise.
4079 (process_references): Likewise.
4080 (cgraph_build_static_cdtor_1): Likewise.
4081 * lto-cgraph.c (lto_output_node): Likewise.
4082 (output_refs): Likewise.
4083 (compute_ltrans_boundary): Likewise.
4084 (input_overwrite_node): Likewise.
4085 (input_node): Likewise.
4086 (input_cgraph_1): Likewise.
4087 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4088 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4089 (class pass_chkp_instrumentation_passes): Likewise.
4090 (make_pass_chkp_instrumentation_passes): Likewise.
4091 * passes.def: Likewise.
4092 * rtl.h (struct GTY): Likewise.
4093 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4094 * stor-layout.c (layout_type): Likewise.
4095 * symtab.c: Likewise.
4096 * target.def: Likewise.
4097 * targhooks.c (default_chkp_bound_type): Likewise.
4098 (default_chkp_bound_mode): Likewise.
4099 (default_builtin_chkp_function): Likewise.
4100 (default_chkp_function_value_bounds): Likewise.
4101 (default_chkp_make_bounds_constant): Likewise.
4102 (default_chkp_initialize_bounds): Likewise.
4103 * targhooks.h (default_chkp_bound_type): Likewise.
4104 (default_chkp_bound_mode): Likewise.
4105 (default_builtin_chkp_function): Likewise.
4106 (default_chkp_function_value_bounds): Likewise.
4107 (default_chkp_make_bounds_constant): Likewise.
4108 (default_chkp_initialize_bounds): Likewise.
4109 * toplev.c (compile_file): Likewise.
4110 (process_options): Likewise.
4111 * tree-core.h (DEF_BUILTIN): Likewise.
4112 (DEF_BUILTIN_CHKP): Likewise.
4113 * tree-inline.c (declare_return_variable): Likewise.
4114 (remap_gimple_stmt): Likewise.
4115 (copy_bb): Likewise.
4116 (initialize_inlined_parameters): Likewise.
4117 (expand_call_inline): Likewise.
4118 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4119 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4120 (make_pass_ipa_chkp_produce_thunks): Likewise.
4121 (make_pass_chkp): Likewise.
4122 (make_pass_chkp_opt): Likewise.
4123 (make_pass_chkp_instrumentation_passes): Likewise.
4124 * tree-pretty-print.c (dump_generic_node): Likewise.
4125 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4126 * tree-ssa-dce.c (propagate_necessity): Likewise.
4127 (eliminate_unnecessary_stmts): Likewise.
4128 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4129 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4130 * tree-ssa-sccvn.h: Likewise.
4131 * tree-ssa-strlen.c (get_string_length): Likewise.
4132 (valid_builtin_call): Likewise.
4133 (adjust_last_stmt): Likewise.
4134 (handle_builtin_strchr): Likewise.
4135 (handle_builtin_strcpy): Likewise.
4136 (handle_builtin_stxncpy): Likewise.
4137 (handle_builtin_memcpy): Likewise.
4138 (handle_builtin_strcat): Likewise.
4139 (strlen_check_and_optimize_stmt): Likewise.
4140 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4141 * tree-streamer-in.c: Likewise.
4142 * tree-streamer.c (record_common_node): Likewise.
4143 * tree.c (tree_code_size): Likewise.
4144 (wide_int_to_tree_1): Likewise.
4145 (type_contains_placeholder_1): Likewise.
4146 (build_common_tree_nodes): Likewise.
4147 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4148 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4149 (POINTER_BOUNDS_P): Likewise.
4150 (BOUNDED_TYPE_P): Likewise.
4151 (BOUNDED_P): Likewise.
4152 (CALL_WITH_BOUNDS_P): Likewise.
4153 (pointer_bounds_type_node): Likewise.
4154 * value-prof.c (gimple_ic): Likewise.
4155 * var-tracking.c (vt_add_function_parameters): Likewise.
4156 * varasm.c (make_decl_rtl): Likewise.
4157 (assemble_start_function): Likewise.
4158 (output_constant): Likewise.
4159 (maybe_assemble_visibility): Likewise.
4160 * varpool.c (ctor_for_folding): Likewise.
4161 * chkp-builtins.def: Remove.
4162 * ipa-chkp.c: Remove.
4163 * ipa-chkp.h: Remove.
4164 * rtl-chkp.c: Remove.
4165 * rtl-chkp.h: Remove.
4166 * tree-chkp-opt.c: Remove.
4167 * tree-chkp.c: Remove.
4168 * tree-chkp.h: Remove.
4170 2018-06-07 Carl Love <cel@us.ibm.com>
4172 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4173 vextract_fp_from_shortl): Add BE support.
4175 2018-06-07 Paul Koning <ni1d@arrl.net>
4177 * compare-elim.c (try_merge_compare): Don't merge compare if
4178 address contains a side effect.
4179 (try_eliminate_compare): Likewise.
4181 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4183 * config.gcc: Support "tremont".
4184 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4185 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4187 * config/i386/i386.c (m_TREMONT): Define.
4188 (processor_target_table): Add "tremont".
4189 (PTA_TREMONT): Define.
4190 (ix86_lea_outperforms): Add TARGET_TREMONT.
4191 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4192 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4193 and M_INTEL_GOLDMONT_PLUS.
4194 (fold_builtin_cpu): Add "tremont".
4195 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4196 (ix86_option_override_internal): Add "tremont".
4197 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4198 (processor_type): Add PROCESSOR_TREMONT.
4199 * config/i386/x86-tune.def: Add m_TREMONT.
4200 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4202 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4204 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4205 symbol defined for msp430i* devices to be lower case.
4207 gcc/testsuite/gcc.target/msp430/
4208 * msp430i-device-symbol.c: New test.
4209 * msp430f-device-symbol.c: New test.
4210 * msp430.h: New test header file.
4212 2018-06-07 Richard Biener <rguenther@suse.de>
4214 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4215 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4216 Properly wrap signed arithmetic if overflow wraps.
4218 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4220 PR tree-optimization/69615
4221 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4222 of a cast from a same precision integral SSA_NAME in a bb dominated
4223 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4224 cast to utype if rhs2 has already a compatible type.
4226 2018-06-07 Richard Biener <rguenther@suse.de>
4228 PR tree-optimization/85935
4229 * graphite-scop-detection.c (find_params_in_bb): Analyze
4230 condition operands with respect to the correct loop. Assert
4231 the analysis doesn't fail.
4233 2018-06-04 Carl Love <cel@us.ibm.com>
4235 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4236 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4237 as it is slightly cheaper.
4238 (first_match_or_eos_index_<mode>):
4239 Calculate index using natural element order.
4240 (first_match_index_<mode>):
4241 Calculate index using natural element order.
4242 (first_match_or_eos_index_<mode>):
4243 Calculate index using natural order.
4244 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4245 for BE and LE modes.
4246 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4247 P9V_BUILTIN_VCLZLSBB_V16QI.
4248 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4251 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4253 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4254 indentation and line wrap for many prototypes. Add missing
4255 @smallexample directives around block of prototypes for vec_xl and
4258 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4260 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4261 track if we pass or return IEEE 128-bit floating point.
4262 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4263 C++ mangling that is compatible with GCC 8.1.
4264 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4265 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4266 floating point types.
4267 (rs6000_function_arg_advance_1): Likewise.
4268 (rs6000_mangle_type): Optionally generate mangled names that match
4269 what GCC 8.1 generated for IEEE 128-bit floating point types.
4270 (rs6000_globalize_decl_name): If we have an external function that
4271 passes or returns IEEE 128-bit types, generate a weak reference
4272 from the mangled name used in GCC 8.1 to the current mangled
4274 (rs6000_init_builtins): Make __ibm128 use the long double type if
4275 long double is IBM extended double. Make __float128 use the long
4276 double type if long double is IEEE 128-bit.
4279 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4280 macro for __ibm128 built-in functions.
4281 (PACK_IF): Add __ibm128 pack/unpack functions.
4282 (UNPACK_IF): Likewise.
4283 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4284 enable long double built-in functions if long double is IEEE
4285 128-bit floating point.
4286 (rs6000_invalid_builtin): Update long double built-in function
4288 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4289 functions, adjust the built-in function to use the long double
4290 built-in function if __ibm128 and long double are the same type.
4291 * doc/extend.texi (PowerPC builtins): Update documention for
4292 __builtin_{,un}pack_longdouble. Add documentation for
4293 __builtin_{,un}pack_ibm128.
4295 2018-06-06 Jim Wilson <jimw@sifive.com>
4297 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4298 (struct machine_function): New field interrupt_mode.
4299 (riscv_handle_type_attribute): New function. Add forward declaration.
4300 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4301 (riscv_expand_epilogue): Check interrupt_mode field.
4302 (riscv_set_current_function): Check interrupt attribute args and
4303 set interrupt_mode field.
4304 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4305 (riscv_sret, riscv_uret): New.
4306 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4307 new arguments to interrupt attribute.
4309 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4312 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4313 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4315 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4317 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4318 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4319 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4320 several redundant entries.
4322 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4324 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4325 type from "rtx" to "rtx_insn *".
4326 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4327 for local "call_insn", removing cast.
4328 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4330 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4332 PR tree-optimization/86066
4333 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4334 for BIT_INSERT_EXPR stores.
4336 2018-06-06 Richard Biener <rguenther@suse.de>
4338 PR tree-optimization/86062
4339 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4340 component refs ontop
4341 of to be offsetted base.
4343 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4345 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4346 to be static and remove check on interrupt attribute name.
4348 gcc/testsuite/gcc.target/msp430/
4349 * function-attributes-4.c: New test.
4350 * static-interrupts.c: New test.
4352 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4354 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4355 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4357 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4360 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4362 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4363 Remove second argument, change how error is called.
4364 (aarch64_layout_arg): Remove second argument from
4365 aarch64_err_no_fpadvsimd call.
4366 (aarch64_init_cumulative_args): Ditto.
4367 (aarch64_gimplify_va_arg_expr): Ditto.
4368 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4370 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4372 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4373 (*simple_return_indirect_internal<mode>): Rename from
4374 simple_return_indirect_internal. Use W mode iterator.
4375 (rstorssp): New expander.
4376 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4377 (clrssbsy): New expander.
4378 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4380 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4382 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4384 (cmse_check_pointed_object): Likewise.
4386 2018-06-05 Martin Liska <mliska@suse.cz>
4388 PR gcov-profile/47618
4389 * doc/invoke.texi: Document how -fprofile-dir format
4392 2018-06-05 Richard Biener <rguenther@suse.de>
4394 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4395 removal pretend DOM info isn't available so we do not update
4396 it and only remove edges, not dominated blocks. Actually free
4397 DOM info in case we removed something. Remove unreachable blocks.
4398 (mfb_keep_latches): Work with either DOM info or marked backedges.
4399 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4400 first. Mark backedges if DOM info isn't available.
4401 (Re-)compute DOM info after cleanup_control_flow_pre.
4403 2018-06-05 Richard Biener <rguenther@suse.de>
4405 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4406 (locus_discrim_hasher::hash): Adjust.
4407 (locus_discrim_hasher::equal): Likewise.
4408 (next_discriminator_for_locus): Work on line directly.
4409 (same_line_p): Pass in expanded locus1 as well.
4410 (assign_discriminators): Avoid redundant location expansions.
4412 2018-06-05 Richard Biener <rguenther@suse.de>
4414 PR tree-optimization/86046
4415 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4416 if required after clearing TREE_ADDRESSABLE.
4418 2018-06-05 Richard Biener <rguenther@suse.de>
4420 PR tree-optimization/86047
4421 * tree-ssa-loop.c (for_each_index): Glob handling of all
4422 decls and constants and really handle all of them.
4424 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4427 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4428 qualifier_void_pointer and qualifier_const_void_pointer.
4429 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4430 (arm_init_builtins): Handle the above.
4431 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4432 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4433 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4436 2018-06-05 Martin Liska <mliska@suse.cz>
4438 * auto-profile.c (read_autofdo_file): Do not use
4439 gcov_ctr_summary struct.
4440 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4441 * coverage.c (struct counts_entry): Likewise.
4442 (read_counts_file): Read just single summary entry.
4443 (get_coverage_counts): Use gcov_summary struct.
4444 * coverage.h (get_coverage_counts): Likewise.
4445 * gcov-dump.c (dump_working_sets): Likewise.
4446 (tag_summary): Dump just single summary.
4447 * gcov-io.c (gcov_write_summary): Write just histogram
4449 (gcov_read_summary): Read just single summary.
4450 (compute_working_sets): Use gcov_summary struct.
4451 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4452 of GCOV_COUNTERS_SUMMABLE.
4453 (GCOV_COUNTERS_SUMMABLE): Remove.
4454 (GCOV_FIRST_VALUE_COUNTER): Replace with
4455 GCOV_COUNTER_V_INTERVAL.
4456 (struct gcov_ctr_summary): Remove.
4457 (struct gcov_summary): Directly use fields of former
4459 (compute_working_sets): Use gcov_summary struct.
4460 * gcov.c (read_count_file): Do not use ctrs fields.
4461 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4463 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4465 * profile.c: Likewise.
4466 * profile.h: Likewise.
4468 2018-06-05 Martin Liska <mliska@suse.cz>
4470 PR gcov-profile/84846
4471 * gcov.c (output_lines): Print working directory only
4472 in intermediate format.
4474 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4476 * config/s390/s390-builtin-types.def: Add void function type.
4477 * config/s390/s390-builtins.def: Use the function type for the
4480 2018-06-04 Jim Wilson <jimw@sifive.com>
4482 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4484 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4485 and maybe_eh_return. Change regno to unsigned int. Use new args to
4486 handle EH_RETURN_DATA_REGNO registers properly.
4487 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4488 (riscv_expand_epilogue): Update comment. Change argument name and
4489 type. Update code to use new name and type. Pass new args to
4490 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4492 * config/riscv/riscv.md (NORMAL_RETURN): New.
4493 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4494 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4495 (eh_return): Call gen_eh_return_internal and emit barrier.
4496 (eh_return_internal): Call riscv_expand_epilogue.
4498 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4500 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4501 bit_insertion field and declare can_be_merged_into method.
4502 (merged_store_group::can_be_merged_into): New method.
4503 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4504 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4505 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4507 2018-06-04 Richard Biener <rguenther@suse.de>
4509 PR tree-optimization/85955
4510 * builtins.c (fold_builtin_sincos): Convert pointers to
4511 destination to appropriate type before dereferencing.
4513 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4515 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4517 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4519 * expr.c (expand_expr_real_1): Force the operand into memory if
4520 its TYPE_MODE is BLKmode and if there is no integer mode for
4521 the number of bits being extracted.
4523 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4527 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4528 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4530 2018-06-04 Richard Biener <rguenther@suse.de>
4532 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4533 (cleanup_tree_cfg_noloop): ... single caller. Do
4534 start_recording_case_labels later.
4536 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4538 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4539 to _IMMINTRIN_H_INCLUDED.
4540 * config/i386/pconfigintrin.h: Ditto.
4541 * config/i386/waitpkgintrin.h: Ditto.
4542 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4543 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4544 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4545 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4546 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4547 waitpkgintrin.h and cldemoteintrin.h.
4549 2018-06-04 Richard Biener <rguenther@suse.de>
4551 PR tree-optimization/86038
4552 * tracer.c (find_best_successor): Check probability for
4553 being initialized, bail out if not.
4555 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4558 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4559 of bits to ignore when comparing architectures.
4561 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4563 PR tree-optimization/69615
4564 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4565 maximum or minimum of the type, try to merge it also as if
4566 range1 is + [-, x - 1] or + [x + 1, -].
4569 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4571 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4573 PR tree-optimization/86034
4574 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4575 the unsigned bitfield type in a bit insertion sequence if it does not
4576 have a larger precision than the bitfield size.
4577 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4579 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4581 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4583 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4585 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4586 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4587 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4588 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4590 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4592 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4593 Disable -fdelete-null-pointer-checks for ELF toolchain.
4595 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4596 Kito Cheng <kito.cheng@gmail.com>
4598 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4599 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4600 (nds32*-*-*): Add glibc and uclibc conditions.
4601 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4602 (TARGET_EXCEPT_UNWIND_INFO): Define.
4603 * config/nds32/elf.h: New file.
4604 * config/nds32/linux.h: New file.
4605 * config/nds32/nds32-elf.opt: New file.
4606 * config/nds32/nds32-linux.opt: New file.
4607 * config/nds32/nds32-fp-as-gp.c
4608 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4609 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4611 (nds32_asm_file_end): Ditto.
4612 (nds32_print_operand): Ditto.
4613 (nds32_insert_attributes): Ditto.
4614 (nds32_init_libfuncs): New function.
4615 (TARGET_HAVE_TLS): Define.
4616 (TARGET_INIT_LIBFUNCS): Define.
4617 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4619 (TARGET_ELF): Apply different mcmodel setting.
4620 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4621 been migrated into elf.h and linux.h files.
4622 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4623 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4624 (mcmodel): The content has been migrated into nds32-elf.opt and
4625 nds32-linux.opt files.
4626 * config/nds32/t-elf: New file.
4627 * config/nds32/t-linux: New file.
4629 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4630 Shiva Chen <shiva0217@gmail.com>
4632 * config/nds32/constants.md (unspec_volatile_element): Add
4633 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4634 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4636 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4637 (make_pass_nds32_fp_as_gp): Declare.
4638 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4640 (nds32_asm_function_end_prologue): Remove unused asm output.
4641 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4642 (nds32_asm_file_start): Output necessary fp_as_gp information.
4643 (nds32_option_override): Adjust register usage.
4644 (nds32_expand_prologue): Consider fp_as_gp situation.
4645 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4646 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4649 (simple_return): Ditto.
4650 (omit_fp_begin): Output special directive for fp_as_gp.
4651 (omit_fp_end): Output special directive for fp_as_gp.
4652 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4653 mforbid-fp-as-gp): New options.
4655 2018-06-01 Mark Wielaard <mark@klomp.org>
4657 * dwarf2out.c (dwarf2out_finish): Remove generation of
4658 DW_AT_loclists_base.
4660 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4662 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4663 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4664 (struct merged_store_group): Add bit_insertion field.
4665 (dump_char_array): Use standard hexadecimal format.
4666 (merged_store_group::merged_store_group): Set bit_insertion to false.
4667 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4668 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4669 also print the mask in the dump file.
4670 (pass_store_merging::gate): Minor tweak.
4671 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4672 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4673 stores with INTEGER_CST stores.
4674 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4675 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4676 and use it throughout. Generate bit insertion sequences if need be.
4677 (pass_store_merging::process_store): Remove redundant condition.
4678 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4680 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4682 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4683 the 128-bit floating point types. Fix function comment.
4685 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4687 * config/aarch64/aarch64-simd.md
4688 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4690 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4693 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4695 PR tree-optimization/85989
4696 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4698 (backprop::intersect_uses): Check it when deciding whether this
4699 is a backedge reference.
4700 (backprop::process_block): Add each phi to m_visited_phis
4701 after visiting it, then clear it at the end.
4703 2018-06-01 Richard Biener <rguenther@suse.de>
4705 * tree-vectorizer.h (vect_dr_stmt): New function.
4706 (vect_get_load_cost): Adjust.
4707 (vect_get_store_cost): Likewise.
4708 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4709 Use vect_dr_stmt instead of DR_SMTT.
4710 (vect_record_base_alignments): Likewise.
4711 (vect_calculate_target_alignment): Likewise.
4712 (vect_compute_data_ref_alignment): Likewise and make static.
4713 (vect_update_misalignment_for_peel): Likewise.
4714 (vect_verify_datarefs_alignment): Likewise.
4715 (vector_alignment_reachable_p): Likewise.
4716 (vect_get_data_access_cost): Likewise. Pass down
4717 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4718 (vect_get_peeling_costs_all_drs): Likewise.
4719 (vect_peeling_hash_get_lowest_cost): Likewise.
4720 (vect_enhance_data_refs_alignment): Likewise.
4721 (vect_find_same_alignment_drs): Likewise.
4722 (vect_analyze_data_refs_alignment): Likewise.
4723 (vect_analyze_group_access_1): Likewise.
4724 (vect_analyze_group_access): Likewise.
4725 (vect_analyze_data_ref_access): Likewise.
4726 (vect_analyze_data_ref_accesses): Likewise.
4727 (vect_vfa_segment_size): Likewise.
4728 (vect_small_gap_p): Likewise.
4729 (vectorizable_with_step_bound_p): Likewise.
4730 (vect_prune_runtime_alias_test_list): Likewise.
4731 (vect_analyze_data_refs): Likewise.
4732 (vect_supportable_dr_alignment): Likewise.
4733 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4734 (vect_gen_prolog_loop_niters): Likewise.
4735 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4736 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4738 (vect_recog_mask_conversion_pattern): Likewise.
4739 (vect_try_gather_scatter_pattern): Likewise.
4740 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4741 to vect_get_store_cost.
4742 (vect_get_store_cost): Get stmt_info instead of DR.
4743 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4744 (vect_get_load_cost): Get stmt_info instead of DR.
4746 2018-06-01 Richard Biener <rguenther@suse.de>
4749 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4750 vars wrapped in MEM_REFs.
4752 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4754 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4755 Fix subreg tests so that we only return a choice between
4756 GENERAL_REGS and FP_REGS if the original classes included both.
4758 2018-06-01 Richard Biener <rguenther@suse.de>
4761 * tree-ssa-structalias.c (get_function_part_constraint):
4762 Handle NULL fi->decl.
4763 (find_func_aliases_for_call): Properly handle indirect
4764 fi from direct call.
4765 (find_func_clobbers): Likewise.
4766 (ipa_pta_execute): Likewise.
4767 (create_variable_info_for): For functions that are ifunc_resolver
4768 resolve to a varinfo that contains the result of the resolver call.
4769 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4772 2018-05-31 Michael Collison <michael.collison@arm.com>
4774 * config/aarch64/aarch64.md:
4775 (*fix_to_zero_extenddfdi2): New pattern.
4776 * gcc.target/aarch64/fix_extend1.c: New testcase.
4778 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4782 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4783 and BUILT_IN_STRNCMP_EQ.
4784 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4785 BUILT_IN_STRNCMP_EQ.
4786 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4787 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4788 (gimple_fold_builtin): Likewise.
4789 * tree-ssa-strlen.c (compute_string_length): New function.
4790 (determine_min_obsize): New function.
4791 (handle_builtin_string_cmp): New function to handle calls to
4792 string compare functions.
4793 (strlen_optimize_stmt): Add handling to builtin string compare
4795 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4796 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4797 * tree.c (build_common_builtin_nodes): Add new defines of
4798 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4800 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4803 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4804 functions with naked attribute.
4806 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4808 * config/i386/sse.md (avx_vec_concat<mode>):
4809 Substitute concat_tg_mode mode attribute with xtg_mode.
4810 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4811 (concat_tg_mode): Remove mode attribute.
4813 2018-05-31 Martin Sebor <msebor@redhat.com>
4816 * calls.c (alloc_max_size): Correct a logic error/typo.
4817 Treat excessive arguments as infinite. Warn for invalid arguments.
4818 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4820 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4823 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4824 and movx for Haswell.
4826 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4827 Cesar Philippidis <cesar@codesourcery.com>
4830 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4831 when emitting error on private/firstprivate reductions.
4832 * omp-low.c (lower_omp_target): Avoid reference-type processing
4833 on pointers for firstprivate clause.
4835 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4837 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4840 * config/aarch64/aarch64-simd.md
4841 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4842 (aarch64_ld1_x3_<mode>): Likewise
4843 (aarch64_st1x2<VALLDIF:mode>): Likewise
4844 (aarch64_st1_x2_<mode>): Likewise
4845 (aarch64_st1x3<VALLDIF:mode>): Likewise
4846 (aarch64_st1_x3_<mode>): Likewise
4847 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4848 (vld1_s8_x3): Likewise.
4849 (vld1_u16_x3): Likewise.
4850 (vld1_s16_x3): Likewise.
4851 (vld1_u32_x3): Likewise.
4852 (vld1_s32_x3): Likewise.
4853 (vld1_u64_x3): Likewise.
4854 (vld1_s64_x3): Likewise.
4855 (vld1_f16_x3): Likewise.
4856 (vld1_f32_x3): Likewise.
4857 (vld1_f64_x3): Likewise.
4858 (vld1_p8_x3): Likewise.
4859 (vld1_p16_x3): Likewise.
4860 (vld1_p64_x3): Likewise.
4861 (vld1q_u8_x3): Likewise.
4862 (vld1q_s8_x3): Likewise.
4863 (vld1q_u16_x3): Likewise.
4864 (vld1q_s16_x3): Likewise.
4865 (vld1q_u32_x3): Likewise.
4866 (vld1q_s32_x3): Likewise.
4867 (vld1q_u64_x3): Likewise.
4868 (vld1q_s64_x3): Likewise.
4869 (vld1q_f16_x3): Likewise.
4870 (vld1q_f32_x3): Likewise.
4871 (vld1q_f64_x3): Likewise.
4872 (vld1q_p8_x3): Likewise.
4873 (vld1q_p16_x3): Likewise.
4874 (vld1q_p64_x3): Likewise.
4875 (vst1_s64_x2): Likewise.
4876 (vst1_u64_x2): Likewise.
4877 (vst1_f64_x2): Likewise.
4878 (vst1_s8_x2): Likewise.
4879 (vst1_p8_x2): Likewise.
4880 (vst1_s16_x2): Likewise.
4881 (vst1_p16_x2): Likewise.
4882 (vst1_s32_x2): Likewise.
4883 (vst1_u8_x2): Likewise.
4884 (vst1_u16_x2): Likewise.
4885 (vst1_u32_x2): Likewise.
4886 (vst1_f16_x2): Likewise.
4887 (vst1_f32_x2): Likewise.
4888 (vst1_p64_x2): Likewise.
4889 (vst1q_s8_x2): Likewise.
4890 (vst1q_p8_x2): Likewise.
4891 (vst1q_s16_x2): Likewise.
4892 (vst1q_p16_x2): Likewise.
4893 (vst1q_s32_x2): Likewise.
4894 (vst1q_s64_x2): Likewise.
4895 (vst1q_u8_x2): Likewise.
4896 (vst1q_u16_x2): Likewise.
4897 (vst1q_u32_x2): Likewise.
4898 (vst1q_u64_x2): Likewise.
4899 (vst1q_f16_x2): Likewise.
4900 (vst1q_f32_x2): Likewise.
4901 (vst1q_f64_x2): Likewise.
4902 (vst1q_p64_x2): Likewise.
4903 (vst1_s64_x3): Likewise.
4904 (vst1_u64_x3): Likewise.
4905 (vst1_f64_x3): Likewise.
4906 (vst1_s8_x3): Likewise.
4907 (vst1_p8_x3): Likewise.
4908 (vst1_s16_x3): Likewise.
4909 (vst1_p16_x3): Likewise.
4910 (vst1_s32_x3): Likewise.
4911 (vst1_u8_x3): Likewise.
4912 (vst1_u16_x3): Likewise.
4913 (vst1_u32_x3): Likewise.
4914 (vst1_f16_x3): Likewise.
4915 (vst1_f32_x3): Likewise.
4916 (vst1_p64_x3): Likewise.
4917 (vst1q_s8_x3): Likewise.
4918 (vst1q_p8_x3): Likewise.
4919 (vst1q_s16_x3): Likewise.
4920 (vst1q_p16_x3): Likewise.
4921 (vst1q_s32_x3): Likewise.
4922 (vst1q_s64_x3): Likewise.
4923 (vst1q_u8_x3): Likewise.
4924 (vst1q_u16_x3): Likewise.
4925 (vst1q_u32_x3): Likewise.
4926 (vst1q_u64_x3): Likewise.
4927 (vst1q_f16_x3): Likewise.
4928 (vst1q_f32_x3): Likewise.
4929 (vst1q_f64_x3): Likewise.
4930 (vst1q_p64_x3): Likewise.
4932 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4934 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4935 user_label_prefix to name.
4937 * tree-core.h: Update comment about the format of NAME string
4938 passed to handler in attribute_spec.
4940 * config/msp430/msp430.md: Remove erroneous subreg expression from
4941 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4942 zero_extend{q,h}isi2.
4944 2018-05-30 Borislav Petkov <bp@suse.de>
4946 * doc/extend.texi: Document some architecture specific
4947 constraints and sort entries.
4949 2018-05-30 Martin Sebor <msebor@redhat.com>
4952 * builtins.c (expand_builtin_stpcpy_1): New function.
4953 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4954 only if the former succeeds.
4956 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4958 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4961 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4963 * doc/invoke.texi (-flinker-output): Document
4965 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4967 * passes.c (ipa_write_summaries): Only modify statements if body
4969 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4970 incrementally linking.
4971 (ipa_passes): Likewise.
4972 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4973 pass down resolution info.
4974 * common.opt (flag_incremental_link): Update info.
4975 * gcc.c (plugin specs): Turn flinker-output=* to
4976 -plugin-opt=-linker-output-known
4977 * toplev.c (compile_file): Also cut compilation when doing incremental
4979 * flag-types. (enum lto_partition_model): Add
4980 LTO_LINKER_OUTPUT_NOLTOREL.
4981 (invoke.texi): Add -flinker-output docs.
4982 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4983 link same way as WPA; do not stream in dead initializers.
4985 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4986 darf2out_register_external_decl): Support incremental link.
4988 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4990 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4992 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4994 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4995 it down to simple_object_copy_lto_debug_sections.
4996 (run_gcc): Determine incremental LTO link time and configure
4997 lto1 into non-wpa mode, disable renaming of debug sections.
4999 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5001 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5002 descriptions of various incorrectly documented functions.
5004 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5007 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5008 address check not strict.
5010 2018-05-30 Richard Biener <rguenther@suse.de>
5012 PR tree-optimization/85964
5013 * tracer.c (better_p): Drop initialized count check, we only
5014 call the function with initialized counts now.
5015 (find_best_successor): Do find a best edge if one
5016 has uninitialized count.
5017 (find_best_predecessor): Likewise. Do BB frequency check only
5018 if count is initialized.
5020 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5022 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5023 (aarch64_ldrstr_offset_compare): New.
5024 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5025 load/store orderings.
5026 (aarch64_gen_adjusted_ldpstp): Likewise.
5028 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5030 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5031 Check for subset of GENERAL_REGS and FP_REGS.
5032 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5035 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5037 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5038 and wi::to_poly_offset. Add the current offset and then check
5039 whether the sum fits, rather than using an unchecked addition of
5040 a checked term. Check for a shwi rather than a uhwi.
5041 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5042 (store_constructor): Use poly_int_tree_p.
5043 (expand_expr_real_1): Likewise.
5044 * function.c (assign_temp): Likewise.
5045 * fold-const.c (const_binop): Use poly_int_tree_p and
5047 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5049 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5050 to_poly_offset for MEM offsets.
5051 * ipa-icf.c (sem_variable::equals): Likewise.
5052 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5053 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5054 wi::to_poly_offset for BIT_FIELD_REF offsets.
5055 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5057 * var-tracking.c (emit_note_insn_var_location): Use
5058 tree_to_poly_uint64.
5060 2018-05-29 Jim Wilson <jimw@sifive.com>
5062 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5064 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5067 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5068 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5070 (sse4_1_round<mode>2): Use nonimmediate_operand
5071 for operand 1 predicate.
5073 2018-05-29 Martin Sebor <msebor@redhat.com>
5074 Richard Biener <rguenther@suse.de>
5077 * calls.c (get_size_range): Call determine_value_range instead
5078 of get_value_range..
5079 * tree-vrp.h (determine_value_range): Declared new function.
5080 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5082 2018-05-29 Richard Biener <rguenther@suse.de>
5084 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5085 sure to use non-pattern stmts for get_earlier_stmt arguments.
5086 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5087 called on pattern stmts.
5088 (get_later_stmt): Likewise.
5090 2018-05-29 Martin Liska <mliska@suse.cz>
5092 PR gcov-profile/85759
5093 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5096 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5098 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5100 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5104 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5105 VEC_PACK_FLOAT_EXPR): New tree codes.
5106 * tree-pretty-print.c (op_code_prio): Handle
5107 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5108 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5109 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5110 * tree-inline.c (estimate_operator_cost): Likewise.
5111 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5112 * fold-const.c (const_binop): Likewise.
5113 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5114 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5115 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5116 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5117 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5118 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5119 * expr.c (expand_expr_real_2): Likewise.
5120 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5121 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5122 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5124 * optabs.c (expand_widen_pattern_expr): For
5125 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5126 sign from result type rather than operand's type.
5127 (expand_binop_directly): For vec_packu_float_optab and
5128 vec_packs_float_optab allow result type to be different from operand's
5130 * optabs-tree.c (optab_for_tree_code): Handle
5131 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5132 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5133 * tree-vect-generic.c (expand_vector_operations_1): Handle
5134 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5135 VEC_PACK_FLOAT_EXPR.
5136 * tree-vect-stmts.c (supportable_widening_operation): Handle
5138 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5139 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5140 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5141 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5142 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5144 (vec_pack<floatprefix>_float_<mode>): New expander.
5145 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5147 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5148 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5149 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5150 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5151 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5153 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5154 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5155 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5156 VEC_PACK_FLOAT_EXPR): Document.
5158 2018-05-29 Richard Biener <rguenther@suse.de>
5160 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5162 (stmt_vec_info_vec): Make pointer.
5163 (init_stmt_vec_info_vec): Remove.
5164 (free_stmt_vec_info_vec): Likewise.
5165 (set_stmt_vec_info_vec): New function.
5166 (free_stmt_vec_infos): Likewise.
5167 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5168 (set_vinfo_for_stmt): Likewise.
5169 (get_earlier_stmt): Likewise.
5170 (get_later_stmt): Likewise.
5171 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5172 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5173 (vec_info::~vec_info): Free stmt_vec_infos.
5174 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5175 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5176 (pass_slp_vectorize::execute): Likewise.
5177 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5178 (free_stmt_vec_info_vec): Likewise.
5179 (set_stmt_vec_info_vec): New function.
5180 (free_stmt_vec_infos): Likewise.
5181 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5182 the global stmt_vec_info_vec.
5183 * tree-parloops.c (gather_scalar_reductions): Use
5184 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5187 2018-05-29 Richard Biener <rguenther@suse.de>
5189 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5191 2018-05-29 Martin Liska <mliska@suse.cz>
5192 David Malcolm <dmalcolm@redhat.com>
5194 * vec.c (test_reverse): New.
5195 (vec_c_tests): Add new test.
5196 * vec.h (vl_ptr>::reverse): New function.
5198 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5200 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5202 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5205 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5207 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5209 2018-05-28 Richard Biener <rguenther@suse.de>
5211 PR tree-optimization/85933
5212 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5213 look at stmts marked as vectorizable.
5215 2018-05-28 Richard Biener <rguenther@suse.de>
5217 PR tree-optimization/85934
5218 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5219 vector boolean check before scalar optimization.
5221 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5223 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5226 2018-05-28 Mark Wielaard <mark@klomp.org>
5228 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5229 if it is an expression containing a minus sign.
5231 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5233 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5235 2018-05-27 Paul Koning <ni1d@arrl.net>
5237 * config/pdp11/pdp11.md (truncsihi2): Remove.
5239 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5240 Chung-Ju Wu <jasonwucj@gmail.com>
5242 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5244 (unaligned_store_dw): Ditto.
5245 * config/nds32/nds32-memory-manipulation.c
5246 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5247 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5248 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5249 (emit_setmem_word_loop): Rename to ...
5250 (emit_setmem_doubleword_loop): ... this.
5251 (nds32_gen_dup_4_byte_to_word_value): New function.
5252 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5253 (nds32_expand_setmem_loop): Refine implementation.
5254 (nds32_expand_setmem_loop_v3m): Ditto.
5255 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5258 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5260 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5262 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5264 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5265 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5266 machine->attr_no_prologue_p.
5267 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5268 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5269 (nds32_expand_epilogue): Consider attr_naked_p.
5270 (nds32_expand_epilogue_v3pop): Likewise.
5271 (nds32_can_use_return_insn): Likewise.
5272 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5273 attr_no_prologue_p fields.
5274 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5276 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5279 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5281 * config/i386/sse.md
5282 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5284 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5286 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5288 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5290 (*<floatsuffix>floatv2div2sf2): Rename to ...
5291 (*float<floatunssuffix>v2div2sf2): ... this.
5292 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5293 (float<floatunssuffix>v2div2sf2_mask): ... this.
5294 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5295 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5296 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5298 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5300 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5302 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5304 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5306 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5308 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5309 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5310 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5311 gen_ufix_truncv8dfv8si2.
5312 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5313 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5314 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5315 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5316 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5317 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5318 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5319 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5321 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5325 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5327 2018-05-25 Jim Wilson <jimw@sifive.com>
5329 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5330 * config/riscv/riscv.c (struct machine_function): Add
5331 interrupt_handler_p and attribute_checked_p fields.
5332 (riscv_attribute_table): Add interrupt.
5333 (riscv_interrupt_type_p): New.
5334 (riscv_save_reg_p): Save extra regs for interrupt handler.
5335 (riscv_use_save_libcall): Return false for interrupt handler.
5336 (riscv_first_stack_step): Add forward declaration.
5337 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5338 for interrupt handler with large frame. Use it for saved reg list.
5339 (riscv_expand_prologue): Move flag_stack_usage_info support to
5340 eliminate duplication.
5341 (riscv_expand_epilogue): Generate mret for interrupt handler.
5342 (riscv_epilogue_uses): New.
5343 (riscv_can_use_return_insn): Return false for interrupt handler.
5344 (riscv_function_ok_for_sibcall): Likewise.
5345 (riscv_set_current_function): Add interrupt handler support.
5346 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5347 * config/riscv/riscv.md (UNSPECV_MRET): New.
5349 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5351 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5353 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5355 PR tree-optimization/85712
5356 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5357 this candidate has already been replaced in-situ by a copy.
5359 2018-05-25 Jason Merrill <jason@redhat.com>
5361 PR c++/80485 - inline function non-zero address.
5362 * symtab.c (nonzero_address): Check DECL_COMDAT.
5364 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5367 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5368 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5369 extension of SImode operation. Use const123_operand predicate.
5370 (*saddsi_1): Remove.
5371 (*saddl_se_1): Ditto.
5373 (*ssubl_se_1): Ditto.
5374 * config/alpha/predicates.md (const123_operand): New predicate.
5375 * config/alpha/constraints.md (P): Use IN_RANGE.
5377 2018-05-25 Richard Biener <rguenther@suse.de>
5379 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5381 (ref_maybe_used_by_stmt_p): Likewise.
5382 (stmt_may_clobber_ref_p): Likewise.
5383 (stmt_may_clobber_ref_p_1): Likewise.
5384 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5386 (ref_maybe_used_by_stmt_p): Likewise.
5387 (stmt_may_clobber_ref_p): Likewise.
5388 (stmt_may_clobber_ref_p_1): Likewise.
5389 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5390 the alias oracle to disambiguate DRs with stmts DR analysis
5392 (vect_analyze_data_refs): Do not give up on not analyzable
5393 DRs for BB vectorization. Remove code truncating the dataref
5396 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5399 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5400 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5401 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5403 2018-05-25 Richard Biener <rguenther@suse.de>
5405 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5406 function, combining stmt data ref gathering and fatal analysis
5408 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5409 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5410 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5411 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5412 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5414 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5416 PR tree-optimization/85720
5417 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5418 SCC if all partitions are builtins.
5419 (version_loop_by_alias_check): New parameter. Generate cancelable
5420 runtime alias check if all partitions are builtins.
5421 (distribute_loop): Update call to above function.
5423 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5425 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5426 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5427 (parm_default_def_partition_arg): Ditto.
5428 (set_parm_default_def_partition): Ditto.
5429 (get_parm_default_def_partitions): Ditto and make it static.
5430 (get_undefined_value_partitions): Ditto and make it static.
5431 (remove_ssa_form): Refactor call to init_var_map here.
5432 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5433 computation for loop region.
5434 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5435 (register_default_def): Delete.
5436 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5437 (parm_default_def_partition_arg): Ditto.
5438 (set_parm_default_def_partition): Ditto.
5439 (get_parm_default_def_partitions): Ditto and make it static.
5440 (get_undefined_value_partitions): Ditto and make it static.
5441 (coalesce_with_default, coalesce_with_default): Update comment.
5442 (create_coalesce_list_for_region): New func factored out from
5443 create_outofssa_var_map.
5444 (populate_coalesce_list_for_outofssa): New func factored out from
5445 create_outofssa_var_map and coalesce_ssa_name.
5446 (create_outofssa_var_map): Delete.
5447 (coalesce_ssa_name): Refactor to support live range computation.
5448 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5449 (get_parm_default_def_partitions): Delete.
5450 (get_undefined_value_partitions): Ditto.
5451 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5452 computation for loop region.
5453 (new_tree_live_info, loe_visit_block): Ditto.
5454 (live_worklist, set_var_live_on_entry): Ditto.
5455 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5456 * tree-ssa-live.h (struct _var_map): New fields.
5457 (init_var_map): Change decl.
5458 (region_contains_p): New.
5460 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5462 * tree-ssa-live.h (live_merge_and_clear): Delete.
5464 2018-05-25 Richard Biener <rguenther@suse.de>
5467 * tree-dump.c (dequeue_and_dump): Remove access to removed
5468 operand 2 of a SWITCH_EXPR.
5470 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5472 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5473 multiplication and division.
5474 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5475 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5476 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5477 (cond_udiv_optab, cond_umod_optab): New optabs.
5478 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5479 (IFN_COND_RDIV): New internal functions.
5480 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5481 TRUNC_MOD_EXPR and RDIV_EXPR.
5482 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5483 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5485 (SVE_INT_BINARY): Include mult.
5486 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5487 (optab, sve_int_op): Handle mult.
5488 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5490 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5491 for SVE_INT_BINARY_SD.
5493 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5495 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5496 (optab, sve_int_op): Handle div and udiv.
5497 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5498 for SVE_INT_BINARY_SD.
5499 (*<optab><mode>3): New insn for the same.
5501 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5503 * tree-vect-patterns.c: Include predict.h.
5504 (vect_recog_divmod_pattern): Restrict check for division support
5505 to when optimizing for size.
5507 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5509 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5510 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5511 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5512 (gimple_match_op::set_op): Likewise.
5513 (gimple_resimplify4): Declare.
5514 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5515 (expr::gen_transform): Likewise.
5516 (decision_tree::gen): Generate a simplification routine for 4 operands.
5517 * gimple-match-head.c (gimple_simplify): Add an overload for
5518 4 operands. In the top-level function, handle up to 4 call
5519 arguments and call gimple_resimplify4.
5520 (gimple_resimplify4): New function.
5521 (build_call_internal): Pass a fourth operand.
5522 (maybe_push_to_seq): Likewise.
5523 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5524 Fold VEC_COND_EXPRs of an operation and a default value into
5525 an IFN_COND_* function if possible.
5526 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5528 (SVE_COND_FP_BINARY): Include them.
5529 (optab, sve_fp_op): Handle them.
5530 (SVE_INT_BINARY_REV): New code iterator.
5531 (SVE_COND_FP_BINARY_REV): New int iterator.
5532 (commutative): New int attribute.
5533 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5535 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5537 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5538 (*cond_<optab><mode>): New patterns for reversed operands.
5540 2018-05-25 Richard Biener <rguenther@suse.de>
5542 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5543 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5544 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5545 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5546 * tree-vect-data-refs.c (everywhere): Adjust users.
5547 * tree-vect-loop.c (everywhere): Likewise.
5548 * tree-vect-slp.c (everywhere): Likewise.
5549 * tree-vect-stmts.c (everywhere): Likewise.
5550 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5552 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5554 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5556 (gcc_cv_as_section_exclude): ... this.
5557 Try Solaris as #exclude syntax.
5558 * configure: Regenerate.
5559 * config.in: Regenerate.
5560 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5562 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5563 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5565 * varasm.c (default_elf_asm_named_section): Don't check if
5566 HAVE_GAS_SECTION_EXCLUDE is defined.
5568 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5570 * doc/md.texi: Update the documentation of the cond_* optabs
5571 to mention the new final operand. Fix GET_MODE_NUNITS call.
5572 Describe the scalar case too.
5573 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5574 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5576 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5577 (get_conditional_internal_fn): Update comment.
5578 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5579 accumulator value as a final argument to conditional functions.
5580 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5581 a define_expand and add an "else" operand. Assert for now that
5582 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5583 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5584 (*cond_<optab><mode>): New patterns.
5585 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5586 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5587 (UNSPEC_COND_EOR): Delete.
5588 (optab): Remove associated mappings.
5589 (SVE_INT_BINARY): New code iterator.
5590 (sve_int_op): Remove int attribute and add "minus" to the code
5592 (SVE_COND_INT_OP): Delete.
5593 (SVE_COND_FP_OP): Rename to...
5594 (SVE_COND_FP_BINARY): ...this.
5596 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5598 * optabs.c (can_reuse_operands_p): New function.
5599 (maybe_legitimize_operands): Try to reuse the results for
5602 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5604 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5605 Add {q} suffix to insn mnemonic.
5607 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5609 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5610 (msp430_warn_func_return): New.
5612 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5614 * fold-const.c (tree_nonzero_bits): New function.
5615 * fold-const.h (tree_nonzero_bits): Likewise.
5616 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5617 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5619 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5623 * varasm.c (assemble_alias): Check ifunc_resolver only on
5626 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5629 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5630 when memory input operand is handled.
5632 2018-05-24 Luis Machado <luis.machado@linaro.org>
5634 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5636 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5638 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5640 * match.pd: Delay FMA folds until after vectorization.
5642 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5645 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5646 address check not strict.
5648 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5650 * gimple-match.h (gimple_match_op): New class.
5651 (mprts_hook): Replace parameters with a gimple_match_op *.
5652 (maybe_build_generic_op): Likewise.
5653 (gimple_simplified_result_is_gimple_val): Replace parameters with
5654 a const gimple_match_op *.
5655 (gimple_simplify): Replace code_helper * and tree * parameters with
5656 a gimple_match_op * parameter.
5657 (gimple_resimplify1): Replace code_helper *, tree and tree *
5658 parameters with a gimple_match_op * parameter.
5659 (gimple_resimplify2): Likewise.
5660 (gimple_resimplify3): Likewise.
5661 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5662 parameters with a gimple_match_op * parameter.
5663 * gimple-match-head.c (gimple_simplify): Change prototypes of
5664 auto-generated functions to take a gimple_match_op * instead of
5665 separate code_helper * and tree * parameters. Make the same
5666 change in the top-level overload and update calls to the
5667 gimple_resimplify routines. Update calls to the auto-generated
5668 functions and to maybe_push_res_to_seq in the publicly-facing
5669 operation-specific gimple_simplify overloads.
5670 (gimple_match_op::MAX_NUM_OPS): Define.
5671 (gimple_resimplify1): Replace rcode and ops with a single res_op
5672 parameter. Update call to gimple_simplify.
5673 (gimple_resimplify2): Likewise.
5674 (gimple_resimplify3): Likewise.
5675 (mprts_hook): Replace parameters with a gimple_match_op *.
5676 (maybe_build_generic_op): Likewise.
5677 (build_call_internal): Replace type, nargs and ops with
5678 a gimple_match_op *.
5679 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5680 with a single gimple_match_op *. Update calls to mprts_hook,
5681 build_call_internal and gimple_simplified_result_is_gimple_val.
5682 Factor out code that is common to the tree_code and combined_fn cases.
5683 * genmatch.c (expr::gen_transform): Replace tem_code and
5684 tem_ops with a gimple_match_op called tem_op. Update calls
5685 to the gimple_resimplify functions and maybe_push_res_to_seq.
5686 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5687 res_ops. Update call to the gimple_resimplify functions.
5688 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5689 (decision_tree::gen): Make the functions take a gimple_match_op *
5690 called res_op instead of separate res_code and res_ops parameters.
5691 Update call accordingly.
5692 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5693 and ops with a single res_op parameter. Update calls to
5694 maybe_build_generic_op and maybe_push_res_to_seq.
5695 (fold_stmt_1): Update calls to gimple_simplify and
5696 replace_stmt_with_simplification.
5697 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5698 and gimple_simplified_result_is_gimple_val.
5699 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5701 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5702 with a gimple_match_op *.
5703 (vn_nary_build_or_lookup): Likewise. Update call to
5704 vn_nary_build_or_lookup_1.
5705 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5706 gimple_match_op *. Update calls to the gimple_resimplify routines
5707 and to gimple_simplified_result_is_gimple_val.
5708 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5709 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5710 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5711 (visit_nary_op): Likewise.
5712 (visit_reference_op_load): Likewise.
5714 2018-05-23 Luis Machado <luis.machado@linaro.org>
5716 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5717 modifier for printing the step amount.
5719 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5722 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5725 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5727 * doc/sourcebuild.texi (Endianness): New subsubsection.
5729 2018-05-23 Luis Machado <luis.machado@linaro.org>
5731 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5732 <prefetch_dynamic_strides>: New const bool field.
5733 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5734 prefetch_dynamic_strides.
5735 (exynosm1_prefetch_tune): Likewise.
5736 (thunderxt88_prefetch_tune): Likewise.
5737 (thunderx_prefetch_tune): Likewise.
5738 (thunderx2t99_prefetch_tune): Likewise.
5739 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5741 (aarch64_override_options_internal): Update to set
5742 PARAM_PREFETCH_DYNAMIC_STRIDES.
5743 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5744 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5745 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5746 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5747 prefetch-dynamic-strides setting.
5749 2018-05-23 Luis Machado <luis.machado@linaro.org>
5751 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5752 <minimum_stride>: New const int field.
5753 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5754 minimum_stride field defaulting to -1.
5755 (exynosm1_prefetch_tune): Likewise.
5756 (thunderxt88_prefetch_tune): Likewise.
5757 (thunderx_prefetch_tune): Likewise.
5758 (thunderx2t99_prefetch_tune): Likewise.
5759 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5760 <default_opt_level>: Set to 3.
5761 (aarch64_override_options_internal): Update to set
5762 PARAM_PREFETCH_MINIMUM_STRIDE.
5763 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5764 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5765 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5766 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5767 stride is constant and is below the minimum stride threshold.
5769 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5771 * config/arm/arm-cpus.in (mode26): Delete.
5772 (armv4): Delete mode26 reference.
5773 * config/arm/arm.c (arm_configure_build_target): Delete use of
5776 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5778 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5780 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5781 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5783 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5784 (floatunsdidf2): Ditto.
5786 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5788 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5789 (fixuns_trunc<mode>si2_avx512f): Ditto.
5790 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5791 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5792 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5794 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5796 PR rtl-optimization/79985
5797 * df-scan.c (df_insn_refs_collect): Remove special case for
5798 global registers and asm statements.
5800 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5802 * extend.texi (Global Register Variables): Rewrite the bullet list.
5803 Note that the register is available for allocation. Note that access
5804 via inline asm must use constraints. Add note about async-signal
5805 handlers. Remove paragraph about automagic register selection.
5807 2018-05-23 Richard Biener <rguenther@suse.de>
5809 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5810 of fixed offset from memset VN.
5812 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5814 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5816 (alloc_cand_and_find_basis): Initialize first_interp field.
5817 (slsr_process_mul): Modify first_interp field.
5818 (slsr_process_add): Likewise.
5819 (slsr_process_cast): Modify first_interp field for each new
5821 (slsr_process_copy): Likewise.
5822 (dump_candidate): Dump first_interp field.
5823 (replace_mult_candidate): Process all interpretations, not just
5825 (replace_rhs_if_not_dup): Likewise.
5826 (replace_one_candidate): Likewise.
5828 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5830 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5832 (aarch64_needs_frame_chain): New function.
5833 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5835 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5838 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5839 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5840 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5841 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5842 as true for strict-align.
5843 (aarch64_can_inline_p): Perform checks even when callee has no
5844 attributes to check for strict alignment.
5845 * doc/extend.texi (AArch64 Function Attributes): Document
5847 * doc/invoke.texi: (AArch64 Options): Likewise.
5849 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5851 PR tree-optimization/85853
5852 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5853 the handling of the root of the node to...
5854 (vect_slp_analyze_node_operations_1): ...this new function,
5855 and run the whole thing with the child nodes' def types
5856 set according to their SLP node's def type.
5858 2018-05-23 Richard Biener <rguenther@suse.de>
5861 * tree-data-ref.c (create_runtime_alias_checks): Defer
5862 and ignore overflow warnings.
5864 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5866 PR tree-optimization/85822
5867 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5870 2018-05-23 Richard Biener <rguenther@suse.de>
5872 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5873 memset constants via native_interpret_expr.
5875 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5878 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5880 (cgraph_node::create_alias): Likewise.
5881 (cgraph_node::get_availability): Check ifunc_resolver instead
5882 of looking up ifunc attribute.
5883 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5884 * varasm.c (do_assemble_alias): Likewise.
5885 (assemble_alias): Likewise.
5886 (default_binds_local_p_3): Likewise.
5887 * cgraph.h (cgraph_node): Add ifunc_resolver.
5888 (cgraph_node::only_called_directly_or_aliased_p): Return false
5890 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5892 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5893 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5894 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5895 instead of looking up ifunc attribute.
5897 2018-05-22 Luis Machado <luis.machado@linaro.org>
5899 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5901 2018-05-22 Martin Sebor <msebor@redhat.com>
5904 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5905 only when expasion succeeds.
5906 (expand_builtin_strcmp): Same.
5907 (expand_builtin_strncmp): Same.
5909 2018-05-22 Martin Sebor <msebor@redhat.com>
5911 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5913 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5914 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5916 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5917 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5919 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5920 Add check for legitimate_address.
5921 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5922 (aarch64_swap_ldrstr_operands): New.
5923 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5926 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5927 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5929 * config/aarch64/aarch64.md: New patterns to generate stp
5931 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5932 single words and double words.
5933 (load_pair_sw, load_pair_dw): Likewise.
5934 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5936 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5938 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5939 for different mode ldpstp and add peephole for merged zero stores.
5941 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5943 (aarch64_gen_store_pair): Rename calls to match new patterns.
5944 (aarch64_gen_load_pair): Rename calls to match new patterns.
5945 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5946 (load_pair<DREG:mode><DREG2:mode>): ... This.
5947 (store_pair<mode>): Rename to...
5948 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5949 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5951 (V_INT_EQUIV): Handle SImode.
5952 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5955 2018-05-22 Martin Sebor <msebor@redhat.com>
5958 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5959 or ajust the presumed bound on an operation to avoid unnecessary
5962 2018-05-22 Martin Sebor <msebor@redhat.com>
5964 PR tree-optimization/85826
5965 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5966 assuming that a DECL necesarily has a constant size.
5968 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5971 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5973 2018-05-22 Richard Biener <rguenther@suse.de>
5975 PR tree-optimization/85834
5976 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5977 non-constant and non-zero memset arguments.
5979 2018-05-22 Martin Liska <mliska@suse.cz>
5982 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5984 2018-05-22 Richard Biener <rguenther@suse.de>
5986 PR tree-optimization/85863
5987 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5988 comparisons when vectype is specified.
5989 (vectorizable_condition): Do not specify vectype for
5990 vect_is_simple_cond when SLP vectorizing.
5992 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5995 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5996 define __ibm128 as long double.
5997 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5998 as a distinct type when IEEE 128-bit support is enabled.
5999 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6000 128-bit types to use the correct functions.
6001 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6002 convert between 128-bit floating point types that have different
6003 modes but the same representation, instead of using gen_lowpart to
6005 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6007 (IFKF_reg): New attributes to give the register constraints for
6009 (extend<mode>tf2_internal): New insns to mark an explicit
6010 conversion between 128-bit floating point types that have a
6011 different mode but share the same representation.
6013 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6015 PR tree-optimization/85814
6016 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6017 a null return from get_strinfo when unsharing the next
6018 strinfo in the chain.
6020 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6023 * varasm.c (weak_finish): Clean up weak_decls.
6025 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6027 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6028 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6029 UNSPEC_UADALP values.
6030 * config/aarch64/iterators.md (ABAL): New int iterator.
6033 (sur): Add mappings for the above.
6034 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6036 (aarch64_<sur>abal<mode>_4): Likewise.
6037 (aarch64_<sur>adalp<mode>_3): Likewise.
6038 (<sur>sadv16qi): New define_expand.
6040 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6042 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6043 (*movdf_internal): Ditto.
6044 (*rcpsf2_sse): Ditto.
6045 (*rsqrtsf2_sse): Ditto.
6046 (*sqrt<mode>2_sse): Ditto.
6048 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6050 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6052 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6053 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6054 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6055 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6057 * config/aarch64/arm_neon.h: Likewise.
6058 * config/aarch64/iterators.md (VQ_I): New.
6060 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6062 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6064 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6067 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6069 * config/nds32/constraints.md (S): New constraint.
6070 * config/nds32/nds32.md (call_internal): Use constraint S.
6071 (call_value_internal): Likewise.
6072 (sibcall_internal): Likewise.
6073 (sibcall_value_internal): Likewise.
6075 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6076 Chung-Ju Wu <jasonwucj@gmail.com>
6078 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6081 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6082 Chung-Ju Wu <jasonwucj@gmail.com>
6084 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6085 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6086 (nds32_rtx_costs_impl): Simplify.
6087 (nds32_address_cost_impl): Simplify.
6088 (nds32_init_rtx_costs): New function.
6089 (nds32_rtx_costs_speed_prefer): Likewise.
6090 (nds32_rtx_costs_size_prefer): Likewise.
6091 (nds32_address_cost_speed_prefer): Likewise.
6092 (nds32_address_cost_speed_fwprop): Likewise.
6093 (nds32_address_cost_size_prefer): Likewise.
6094 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6095 * config/nds32/nds32.c (nds32_option_override): Use
6096 nds32_init_rtx_costs function.
6098 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6100 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6101 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6102 (TARGET_PIPELINE_N8): Likewise.
6103 (TARGET_PIPELINE_N10): Likewise.
6104 (TARGET_PIPELINE_N13): Likewise.
6105 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6107 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6109 * config/nds32/nds32-fpu.md: Update copyright year.
6111 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6113 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6115 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6117 * config/nds32/nds32.c
6118 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6119 * config/nds32/nds32.opt (minline-asm-r15): New option.
6121 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6123 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6125 * config/nds32/nds32.md (abssi2): New pattern.
6127 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6129 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6130 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6131 Merge insn pattern from sse_cvtsi2ss<round_name> and
6132 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6133 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6134 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6135 using SWI48 mode iterator.
6136 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6137 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6138 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6139 pattern from sse_cvttss2si<round_saeonly_name>
6140 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6141 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6142 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6143 using SWI48 mode iterator.
6144 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6145 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6146 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6147 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6148 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6149 using SWI48 mode iterator.
6150 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6151 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6152 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6153 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6154 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6155 SWI48 mode iterator.
6156 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6157 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6158 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6159 pattern from sse_cvttsd2si<round_saeonly_name>
6160 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6162 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6164 * config/nds32/nds32-md-auxiliary.c
6165 (nds32_valid_smw_lwm_base_p): Refine.
6166 (nds32_output_smw_single_word): Refine.
6167 (nds32_output_smw_double_word): New.
6168 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6170 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6172 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6173 (nds32_output_stack_pop): Refine.
6174 (nds32_expand_unaligned_load): Refine.
6175 (nds32_expand_unaligned_store): Refine.
6177 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6178 Chung-Ju Wu <jasonwucj@gmail.com>
6180 * config/nds32/constants.md: Add TP_REGNUM constant.
6181 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6182 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6184 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6185 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6186 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6187 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6188 and PIC code generation.
6189 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6191 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6193 * config/nds32/nds32.md: Support TLS and PIC.
6194 * config/nds32/nds32.c: Support TLS and PIC.
6195 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6196 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6199 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6201 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6202 mode with E_ prefix.
6204 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6205 Chung-Ju Wu <jasonwucj@gmail.com>
6207 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6208 * config/nds32/nds32-md-auxiliary.c
6209 (symbolic_reference_mentioned_p): New.
6210 (nds32_legitimize_ict_address): New.
6211 (nds32_expand_ict_move): New.
6212 (nds32_indirect_call_referenced_p): New.
6213 (nds32_symbol_binds_local_p): Delete.
6214 (nds32_long_call_p): Modify.
6215 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6216 * config/nds32/nds32-protos.h
6217 (symbolic_reference_mentioned_p): Declare.
6218 (nds32_legitimize_ict_address): Declare.
6219 (nds32_expand_ict_move): Declare.
6220 (nds32_indirect_call_referenced_p): Declare.
6221 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6222 (nds32_relax_group): Use nds32_ict_const_p as condition.
6223 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6224 (nds32_asm_file_start): Output ict_model directive in asm code.
6225 (nds32_legitimate_address_p): Consider indirect call.
6226 (nds32_print_operand): Consider indirect call.
6227 (nds32_print_operand_address): Consider indirect call.
6228 (nds32_insert_attributes): Handle "indirect_call" attribute.
6229 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6230 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6231 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6232 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6233 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6234 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6235 (TARGET_ICT_MODEL_SMALL): Define.
6236 (TARGET_ICT_MODEL_LARGE): Define.
6237 * config/nds32/nds32.md (movsi): Consider ict model.
6238 (call, call_value): Consider ict model.
6239 (sibcall, sibcall_value): Consider ict model.
6240 * config/nds32/nds32.opt (mict-model): New option.
6241 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6244 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6245 Monk Chiang <sh.chiang04@gmail.com>
6246 Jim Wilson <jimw@sifive.com>
6248 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6249 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6250 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6251 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6252 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6253 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6254 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6255 compute save_libcall_adjustment properly.
6256 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6257 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6258 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6259 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6260 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6261 (ABI_SPEC): Handle mabi=ilp32e.
6262 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6263 (RVE): Add RVE mask.
6264 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6265 <-march>: Add rv32e as an example.
6267 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6270 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6271 (intra_create_variable_infos): Handle C++ constructors.
6273 2018-05-18 Martin Liska <mliska@suse.cz>
6275 * passes.def: Remove a redundant pass.
6277 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6280 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6282 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6284 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6286 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6288 (armv2, armv2a, armv3, armv3m): Delete architectures.
6289 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6290 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6291 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6293 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6294 (*mulsidi3adddi): Likewise.
6295 (mulsidi3): Likewise.
6296 (*mulsidi3_nov6): Likewise.
6297 (umulsidi3): Likewise.
6298 (umulsidi3_nov6): Likewise.
6299 (umaddsidi4): Likewise.
6300 (*umulsidi3adddi): Likewise.
6301 (smulsi3_highpart): Likewise.
6302 (*smulsi3_highpart_nov6): Likewise.
6303 (umulsi3_highpart): Likewise.
6304 (*umulsi3_highpart_nov6): Likewise.
6305 * config/arm/arm.h (arm_arch3m): Delete.
6306 * config/arm/arm.c (arm_arch3m): Delete.
6307 (arm_option_override_internal): Update armv3-related comment.
6308 (arm_configure_build_target): Delete use of isa_bit_mode32.
6309 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6310 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6311 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6313 (mulusa3): Likewise.
6314 * config/arm/arm-protos.h (arm_arch3m): Delete.
6315 * config/arm/arm-tables.opt: Regenerate.
6316 * config/arm/arm-tune.md: Likewise.
6317 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6318 deleted architectures.
6320 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6322 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6323 (armv5t, armv5te): New features.
6324 (ARMv5, ARMv5e): Delete fgroups.
6325 (ARMv5t, ARMv5te): Adjust for above changes.
6327 (armv5, armv5e): Delete arches.
6328 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6330 (*call_reg_arm): Likewise.
6331 (*call_value_reg_armv5): Likewise.
6332 (*call_value_reg_arm): Likewise.
6333 (*call_symbol): Likewise.
6334 (*call_value_symbol): Likewise.
6335 (*sibcall_insn): Likewise.
6336 (*sibcall_value_insn): Likewise.
6338 (prefetch): Likewise.
6339 (define_split and define_peephole2 dependent on arm_arch5):
6341 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6343 (TARGET_ARM_QBIT): Likewise.
6344 (TARGET_DSP_MULTIPLY): Likewise.
6345 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6346 (arm_arch5, arm_arch5e): Delete.
6347 (arm_arch5t, arm_arch5te): Declare.
6348 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6349 (arm_arch5t): Declare.
6350 (arm_option_reconfigure_globals): Update for the above.
6351 (arm_options_perform_arch_sanity_checks): Update comment, replace
6352 use of arm_arch5 with arm_arch5t.
6353 (use_return_insn): Likewise.
6354 (arm_emit_call_insn): Likewise.
6355 (output_return_instruction): Likewise.
6356 (arm_final_prescan_insn): Likewise.
6357 (arm_coproc_builtin_available): Likewise.
6358 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6359 arm_arch5e with arm_arch5t and arm_arch5te.
6360 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6361 (arm_arch5t, arm_arch5te): Declare.
6362 * config/arm/arm-tables.opt: Regenerate.
6363 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6364 * config/arm/t-multilib: Likewise.
6365 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6366 instead of arm_arch5.
6367 (*call_reg_thumb1): Likewise.
6368 (*call_value_reg_thumb1_v5): Likewise.
6369 (*call_value_reg_thumb1): Likewise.
6370 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6372 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6374 2018-05-18 Martin Liska <mliska@suse.cz>
6376 PR gcov-profile/84846
6377 * doc/gcov.texi: Document -t option of gcov tool.
6379 2018-05-18 Martin Liska <mliska@suse.cz>
6381 PR gcov-profile/84846
6382 * gcov.c (print_usage): Add new -t option.
6383 (process_args): Handle the option.
6384 (generate_results): Use stdout as output when requested by
6387 2018-05-18 Martin Liska <mliska@suse.cz>
6389 PR gcov-profile/84846
6390 * coverage.c (coverage_init): Write PWD to .gcno file.
6391 * doc/gcov.texi: Document how working directory is printed.
6392 * gcov-dump.c (dump_gcov_file): Print PWD.
6393 * gcov.c (output_intermediate_file): Likewise.
6394 (read_graph_file): Read PWD string.
6395 (output_lines): Print PWD.
6397 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6400 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6401 for retval and return false if all args to phi are zero.
6403 2018-05-18 Richard Biener <rguenther@suse.de>
6405 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6407 (evrp_dom_walker::before_dom_children): Call it.
6409 2018-05-18 Richard Biener <rguenther@suse.de>
6411 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6412 results when processing array refs with variable index.
6414 2018-05-18 Toon Moene <toon@moene.org>
6416 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6417 directly after that of -floop-interchange. Indicate that both
6418 options are enabled by default when specifying -O3.
6420 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6422 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6423 iterator. Delete separate integer-mode vec_set<mode> expander.
6424 (aarch64_simd_vec_setv2di): Delete.
6425 (vec_setv2di): Delete.
6426 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6427 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6428 the "w, r" alternative.
6430 2018-05-18 Martin Liska <mliska@suse.cz>
6432 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6433 * tree-pass.h (make_pass_lower_switch_O0): New function.
6434 * tree-switch-conversion.c (node_has_low_bound): Remove.
6435 (node_has_high_bound): Likewise.
6436 (node_is_bounded): Likewise.
6437 (class pass_lower_switch): Make it a template type and create
6439 (pass_lower_switch::execute): Add template argument.
6440 (make_pass_lower_switch): New function.
6441 (make_pass_lower_switch_O0): New function.
6442 (do_jump_if_equal): Remove.
6443 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6444 all the hard work to tree optimization passes.
6446 2018-05-18 Martin Liska <mliska@suse.cz>
6448 * dbgcnt.c (limit_low): Renamed from limit.
6449 (limit_high): New variable.
6450 (dbg_cnt_is_enabled): Check for upper limit.
6451 (dbg_cnt): Adjust dumping.
6452 (dbg_cnt_set_limit_by_index): Add new argument for high
6454 (dbg_cnt_set_limit_by_name): Likewise.
6455 (dbg_cnt_process_single_pair): Parse new format.
6456 (dbg_cnt_process_opt): Use strtok.
6457 (dbg_cnt_list_all_counters): Remove 'value' and add
6459 * doc/invoke.texi: Document changes.
6461 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6463 * doc/sourcebuild.texi (scalar_all_fma): Document.
6464 * tree.def (FMA_EXPR): Delete.
6465 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6466 * internal-fn.c (ternary_direct): New macro.
6467 (expand_ternary_optab_fn): Likewise.
6468 (direct_ternary_optab_supported_p): Likewise.
6469 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6470 * builtins.c (fold_builtin_fma): Delete.
6471 (fold_builtin_3): Don't call it.
6472 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6473 * expr.c (expand_expr_real_2): Likewise.
6474 * fold-const.c (operand_equal_p): Likewise.
6475 (fold_ternary_loc): Likewise.
6476 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6477 * gimple.c (DEFTREECODE): Likewise.
6478 * gimplify.c (gimplify_expr): Likewise.
6479 * optabs-tree.c (optab_for_tree_code): Likewise.
6480 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6481 * tree-eh.c (operation_could_trap_p): Likewise.
6482 (stmt_could_throw_1_p): Likewise.
6483 * tree-inline.c (estimate_operator_cost): Likewise.
6484 * tree-pretty-print.c (dump_generic_node): Likewise.
6485 (op_code_prio): Likewise.
6486 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6487 * tree-ssa-operands.c (get_expr_operands): Likewise.
6488 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6489 * fold-const-call.h (fold_fma): Delete.
6490 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6491 CFN_FNMA and CFN_FNMS.
6493 * genmatch.c (combined_fn): New enum.
6494 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6495 (commutative_op): New function.
6496 (commutate): Use it. Handle more than 2 operands.
6497 (dt_operand::gen_gimple_expr): Use commutative_op.
6498 (parser::parse_expr): Allow :c to be used with non-binary
6499 operators if the commutative operand is known.
6500 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6501 CFN_FMS, CFN_FNMA and CFN_FNMS.
6502 (backprop::process_assign_use): Remove FMA_EXPR handling.
6503 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6504 (gen_hsa_fma): New function.
6505 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6506 IFN_FNMA and IFN_FNMS.
6507 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6508 * gimple-fold.h (follow_all_ssa_edges): Declare.
6509 * gimple-fold.c (follow_all_ssa_edges): New function.
6510 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6511 gimple_build interface and use follow_all_ssa_edges to fold the result.
6512 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6513 instead of checking for optabs directly.
6514 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6515 rather than FMA_EXPRs.
6516 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6517 call to IFN_FMA instead of an FMA_EXPR.
6519 2018-05-17 Jim Wilson <jimw@sifive.com>
6521 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6522 sign extended promoted subreg, and the range does not have the sign bit
6523 set, then do a sign extend.
6525 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6526 test, check for sign extended subreg and/or constant operands, and
6527 do a sign extend in that case.
6529 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6531 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6532 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6534 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6535 Change logics_shift_reg to logics_shift_imm.
6536 (thunderx2t99_fp_loadpair_basic): Delete.
6537 (thunderx2t99_fp_storepair_basic): Delete.
6538 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6539 (thunderx2t99_asimd_polynomial): Delete.
6540 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6541 and neon_fp_mul_d_scalar_q.
6542 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6543 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6544 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6545 (thunderx2t99_asimd_lut): Add missing tbl types.
6546 (thunderx2t99_asimd_ext): Delete.
6547 (thunderx2t99_asimd_load1_1_mult): Delete.
6548 (thunderx2t99_asimd_load1_2_mult): Delete.
6549 (thunderx2t99_asimd_load1_ldp): New.
6550 (thunderx2t99_asimd_load1): New.
6551 (thunderx2t99_asimd_load2): Add missing *load2* types.
6552 (thunderx2t99_asimd_load3): New.
6553 (thunderx2t99_asimd_load4): New.
6554 (thunderx2t99_asimd_store1_1_mult): Delete.
6555 (thunderx2t99_asimd_store1_2_mult): Delete.
6556 (thunderx2t99_asimd_store2_mult): Delete.
6557 (thunderx2t99_asimd_store2_onelane): Delete.
6558 (thunderx2t99_asimd_store_stp): New.
6559 (thunderx2t99_asimd_store1): New.
6560 (thunderx2t99_asimd_store2): New.
6561 (thunderx2t99_asimd_store3): New.
6562 (thunderx2t99_asimd_store4): New.
6564 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6566 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6567 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6569 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6570 Segher Boessenkool <segher@kernel.crashing.org>
6573 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6576 2018-05-17 Richard Biener <rguenther@suse.de>
6578 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6579 for pruning loop and prune defs feeding only already visited PHIs.
6581 2018-05-17 Richard Biener <rguenther@suse.de>
6583 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6585 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6586 Richard Biener <rguenther@suse.de>
6588 PR tree-optimization/85793
6589 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6590 for VMAT_ELEMENTWISE.
6592 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6594 * internal-fn.h (lookup_internal_fn): Declare
6595 * internal-fn.c (lookup_internal_fn): New function.
6596 * gimple.c (gimple_build_call_from_tree): Handle calls to
6598 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6599 internal function names.
6600 * tree-pretty-print.c (dump_generic_node): Likewise.
6601 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6603 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6605 * gimple-fold.h (gimple_build): Make the function forms take
6606 combined_fn rather than built_in_function.
6607 (gimple_simplify): Likewise.
6608 * gimple-match-head.c (gimple_simplify): Likewise.
6609 * gimple-fold.c (gimple_build): Likewise.
6610 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6611 rather than gimple_build_call_internal.
6612 (get_initial_defs_for_reduction): Likewise.
6613 (vect_create_epilog_for_reduction): Likewise.
6614 (vectorizable_live_operation): Likewise.
6616 2018-05-17 Martin Liska <mliska@suse.cz>
6618 * gimple-ssa-sprintf.c (format_directive): Do not use
6619 space in between 'G_' and '('.
6621 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6624 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6625 even if the mask is not all ones.
6628 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6630 (ix86_gimple_fold_builtin): Likewise.
6633 * config/i386/i386.c: Include tree-vector-builder.h.
6634 (ix86_vector_shift_count): New function.
6635 (ix86_fold_builtin): Fold shift builtins by scalar count.
6636 (ix86_gimple_fold_builtin): Likewise.
6638 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6639 _mm512_setzero): New intrinsics.
6641 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6642 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6644 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6645 code generation for cases where splatting a value is not useful.
6646 * simplify-rtx.c (simplify_ternary_operation): Simplify
6647 vec_merge across a vec_duplicate and a paradoxical subreg forming
6648 a vector mode to a vec_concat.
6650 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6652 * config.gcc: Support "goldmont-plus".
6653 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6655 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6656 PROCESSOR_GOLDMONT_PLUS.
6657 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6658 (processor_target_table): Add "goldmont-plus".
6659 (PTA_GOLDMONT_PLUS): Define.
6660 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6661 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6662 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6663 (fold_builtin_cpu): Add "goldmont-plus".
6664 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6665 (ix86_option_override_internal): Add "goldmont-plus".
6666 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6667 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6668 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6669 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6671 2018-05-17 Richard Biener <rguenther@suse.de>
6673 PR tree-optimization/85757
6674 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6675 remove defs that only feed that PHI from further processing.
6677 2018-05-16 Jim Wilson <jimw@sifive.com>
6679 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6681 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6683 2018-05-16 Mark Wielaard <mark@klomp.org>
6685 * dwarf2out.c (count_index_strings): New function.
6686 (output_indirect_strings): Call count_index_strings and generate
6687 header for dwarf_version >= 5.
6689 2018-05-16 Mark Wielaard <mark@klomp.org>
6691 * dwarf2out.c (dwarf_FORM): New function.
6692 (set_indirect_string): Use dwarf_FORM.
6693 (reset_indirect_string): Likewise.
6694 (size_of_die): Likewise.
6695 (value_format): Likewise.
6696 (output_die): Likewise.
6697 (add_skeleton_AT_string): Likewise.
6698 (output_macinfo_op): Likewise.
6699 (index_string): Likewise.
6700 (output_index_string_offset): Likewise.
6701 (output_index_string): Likewise.
6702 (count_index_strings): Likewise.
6704 2018-05-16 Carl Love <cel@us.ibm.com>
6706 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6707 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6709 2018-05-16 Martin Jambor <mjambor@suse.cz>
6711 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6712 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6714 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6716 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6717 (fnma<mode>4): Likewise.
6718 (fms<mode>4): Likewise.
6719 (fnms<mode>4): Likewise.
6720 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6721 (aarch64_fnma<mode>4): Likewise.
6722 (aarch64_fms<mode>4): Likewise.
6723 (aarch64_fnms<mode>4): Likewise.
6724 (aarch64_fnmadd<mode>4): Likewise.
6726 2018-05-16 Jason Merrill <jason@redhat.com>
6728 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6730 2018-05-16 Richard Biener <rguenther@suse.de>
6732 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6733 (dump_stmt_cost): Declare.
6734 (add_stmt_cost): Dump cost we add.
6735 (add_stmt_costs): New function.
6736 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6738 (vect_analyze_stmt): Adjust prototype.
6739 (vectorizable_condition): Likewise.
6740 (vectorizable_live_operation): Likewise.
6741 (vectorizable_reduction): Likewise.
6742 (vectorizable_induction): Likewise.
6743 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6744 cost vector to pass to vectorizable_ and record afterwards.
6745 (vect_model_reduction_cost): Take cost vector argument and adjust.
6746 (vect_model_induction_cost): Likewise.
6747 (vectorizable_reduction): Likewise.
6748 (vectorizable_induction): Likewise.
6749 (vectorizable_live_operation): Likewise.
6750 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6751 SLP_TREE_NUMBER_OF_VEC_STMTS.
6752 (vect_analyze_slp_cost_1): Remove.
6753 (vect_analyze_slp_cost): Likewise.
6754 (vect_slp_analyze_node_operations): Take visited args and
6755 a target cost vector. Avoid processing already visited stmt sets.
6756 (vect_slp_analyze_operations): Use a local cost vector to gather
6757 costs and register those of non-discarded instances.
6758 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6759 (vect_schedule_slp_instance): Remove copying of
6760 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6762 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6763 adding cost. Record cost entry location.
6764 (vect_prologue_cost_for_slp_op): Function to compute cost of
6765 a constant or invariant generated for SLP vect in the prologue,
6766 split out from vect_analyze_slp_cost_1.
6767 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6768 (vect_model_promotion_demotion_cost): Likewise.
6769 (vect_model_store_cost): Likewise, make static.
6770 (vect_model_load_cost): Likewise.
6771 (vectorizable_bswap): Add cost vector arg and adjust.
6772 (vectorizable_call): Likewise.
6773 (vectorizable_simd_clone_call): Likewise.
6774 (vectorizable_conversion): Likewise.
6775 (vectorizable_assignment): Likewise.
6776 (vectorizable_shift): Likewise.
6777 (vectorizable_operation): Likewise.
6778 (vectorizable_store): Likewise.
6779 (vectorizable_load): Likewise.
6780 (vectorizable_condition): Likewise.
6781 (vectorizable_comparison): Likewise.
6782 (can_vectorize_live_stmts): Likewise.
6783 (vect_analyze_stmt): Likewise.
6784 (vect_transform_stmt): Adjust calls to vectorizable_*.
6785 * tree-vectorizer.c: Include gimple-pretty-print.h.
6786 (dump_stmt_cost): New function.
6788 2018-05-16 Richard Biener <rguenther@suse.de>
6790 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6791 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6792 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6793 (check_name): New callback.
6794 (dse_classify_store): Track cycles via a visited bitmap of PHI
6795 defs and simplify handling of in-loop and across loop dead stores
6796 and properly fail for loop-variant refs. Handle byte-tracking with
6797 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6800 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6802 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6803 (vect_get_mask_type_for_stmt): Likewise.
6804 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6806 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6807 to determine the statement's vector type and the vector type that
6808 should be used for calculating nunits. Deal with cases in which
6809 the type has to be deferred.
6810 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6811 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6812 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6813 (vect_determine_vf_for_stmt): New functions, split out from...
6814 (vect_determine_vectorization_factor): ...here.
6815 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6816 (vect_get_mask_type_for_stmt): New functions, split out from
6817 vect_determine_vectorization_factor.
6819 2018-05-16 Richard Biener <rguenther@suse.de>
6821 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6822 verify the [VEC_]COND_EXPR embedded comparison.
6824 2018-05-15 Martin Sebor <msebor@redhat.com>
6826 PR tree-optimization/85753
6827 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6828 RECORD_TYPE in addition to ARRAY_TYPE.
6830 2018-05-15 Martin Sebor <msebor@redhat.com>
6833 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6835 2018-05-15 Richard Biener <rguenther@suse.de>
6837 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6838 add by_clobber_p one. Change algorithm to collect all defs
6839 representing uses we need to walk and try reducing them to
6840 a single one before failing.
6841 (dse_dom_walker::dse_optimize_stmt): Adjust.
6843 2018-05-13 Mark Wielaard <mark@klomp.org>
6845 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6846 (size_of_loc_descr): Likewise.
6847 (output_loc_operands): Likewise.
6848 (output_loc_operands_raw): Likewise.
6849 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6850 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6851 (hash_loc_operands): Likewise.
6852 (compare_loc_operands): Likewise.
6854 2018-05-14 Mark Wielaard <mark@klomp.org>
6856 * dwarf2out.c (count_index_addrs): New function.
6857 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6859 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6861 PR tree-optimization/83648
6862 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6863 return value as malloc candidate.
6865 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6868 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6869 param as true in call to suggest_attribute.
6871 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6873 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6874 -mreadonly-in-sdata.
6876 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6878 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6880 (aarch64_crypto_aesd_fused): Likewise.
6882 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6884 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6885 (movsi_aarch64): Likewise.
6886 (load_pairsi): Likewise.
6887 (load_pairdi): Likewise.
6888 (store_pairsi): Likewise.
6889 (store_pairdi): Likewise.
6890 (load_pairsf): Likewise.
6891 (load_pairdf): Likewise.
6892 (store_pairsf): Likewise.
6893 (store_pairdf): Likewise.
6894 (zero_extend): Likewise.
6895 (trunc): Swap alternatives.
6896 (fcvt_target): Add '?' to prefer w over r.
6898 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6901 * config/i386/i386.md: Disallow non-commutative arithmetics in
6902 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6903 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6904 in the peephole2 before it.
6906 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6908 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6909 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6910 (ix86_handle_option): Handle -mcldemote.
6911 * config.gcc: New header.
6912 * config/i386/cldemoteintrin.h: New file.
6913 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6914 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6916 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6917 OPTION_MASK_ISA_CLDEMOTE.
6918 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6919 (ix86_valid_target_attribute_inner_p): Ditto.
6920 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6921 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6922 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6923 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6924 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6926 * config/i386/i386.opt: Add -mcldemote.
6927 * config/i386/x86intrin.h: New header.
6928 * doc/invoke.texi: Add -mcldemote.
6930 2018-05-14 Richard Biener <rguenther@suse.de>
6932 * doc/match-and-simplify.texi: Adjust :s documentation.
6934 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6936 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6937 intended memcpy size.
6938 (REORDER_45): Likewise.
6940 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6942 * sort.cc: New file.
6943 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6944 * vec.c (qsort_chk): Use gcc_qsort.
6945 * Makefile.in (OBJS-libcommon): Add sort.o.
6946 (build/sort.o): New target. Use it...
6947 (BUILD_RTL): ... here, and...
6948 (build/gencfn-macros): ... here, and...
6949 (build/genmatch): ... here.
6951 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6952 Chung-Ju Wu <jasonwucj@gmail.com>
6954 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6955 * config/nds32/nds32-graywolf.md: New file.
6956 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6957 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6959 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6960 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6961 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6962 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6963 * config/nds32/pipelines.md: Include n15 settings.
6965 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6966 Chung-Ju Wu <jasonwucj@gmail.com>
6968 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6969 * config/nds32/nds32-n13.md: New file.
6970 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6971 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6973 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6974 * config/nds32/nds32.md (pipeline_model): Add n13.
6975 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6976 * config/nds32/pipelines.md: Include n13 settings.
6978 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6979 Chung-Ju Wu <jasonwucj@gmail.com>
6981 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6982 * config/nds32/nds32-n10.md: New file.
6983 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6984 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6986 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6987 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6988 * config/nds32/nds32.md (pipeline_model): Add n10.
6989 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6990 * config/nds32/pipelines.md: Include n10 settings.
6992 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6993 Kito Cheng <kito.cheng@gmail.com>
6994 Chung-Ju Wu <jasonwucj@gmail.com>
6996 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6997 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6998 Add enum values for DSP extension instructions.
6999 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7001 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7002 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7004 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7005 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7006 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7007 * config/nds32/nds32-intrinsic.md: Likewise.
7008 * config/nds32/nds32_intrinsic.h: Likewise.
7009 * config/nds32/nds32-md-auxiliary.c: Likewise.
7010 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7011 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7012 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7013 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7014 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7015 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7016 TYPE_DMAC in switch statement.
7017 * config/nds32/nds32.c: New checking and implementation for DSP
7018 extension instructions.
7019 * config/nds32/nds32.h: Likewise.
7020 * config/nds32/nds32.md: Likewise.
7021 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7022 * config/nds32/predicates.md: Implement new predicates for DSP
7025 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7027 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7028 Reformat alternatives and attributes so it is easier to identify
7029 which constraints/attributes go with which instruction.
7030 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7031 (mov<mode>_softfloat32, FMOVE64): Likewise.
7032 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7033 (mov<mode>_softfloat64, FMOVE64): Likewise.
7035 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7037 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7039 (Basic PowerPC Built-in Functions): The new name of the
7040 subsection previously known as "PowerPC Built-in Functions".
7041 (Basic PowerPC Built-in Functions Available on all Configurations):
7043 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7044 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7045 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7046 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7048 2018-05-11 Martin Jambor <mjambor@suse.cz>
7051 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7054 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7057 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7059 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7061 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7062 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7063 (ix86_handle_option): Handle -mwaitpkg.
7064 * config.gcc: New header.
7065 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7066 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7067 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7069 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7070 OPTION_MASK_ISA_WAITPKG.
7071 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7072 (ix86_option_override_internal): Add PTA_WAITPKG.
7073 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7074 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7075 IX86_BUILTIN_TPAUSE.
7076 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7077 __builtin_ia32_umwait and __builtin_ia32_tpause.
7078 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7079 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7080 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7081 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7082 UNSPECV_TPAUSE): New.
7083 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7084 * config/i386/i386.opt: Add -mwaitpkg.
7085 * config/i386/waitpkgintrin.h: New file.
7086 * config/i386/x86intrin.h: New header.
7087 * doc/invoke.texi: Add -mwaitpkg.
7089 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7092 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7094 (cortex-m0): Use armv6s-m isa.
7095 (cortex-m0plus): Likewise.
7096 (cortex-m1): Likewise.
7097 (cortex-m0.small-multiply): Likewise.
7098 (cortex-m0plus.small-multiply): Likewise.
7099 (cortex-m1.small-multiply): Likewise.
7101 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7102 Jakub Jelinek <jakub@redhat.com>
7104 PR tree-optimization/85692
7105 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7106 source permute as well.
7108 2018-05-11 Martin Liska <mliska@suse.cz>
7111 * doc/extend.texi: Document LLVM style format for no_sanitize
7114 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7116 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7117 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7118 (mode_supports_vsx_dform_quad): Likewise.
7119 (mode_supports_vmx_dform): Move these functions to be next to the
7120 other mode_supports functions.
7121 (mode_supports_dq_form): Likewise.
7122 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7123 mode_supports_dq_form.
7124 (reg_offset_addressing_ok_p): Likewise.
7125 (offsettable_ok_by_alignment): Likewise.
7126 (rs6000_legitimate_offset_address_p): Likewise.
7127 (legitimate_lo_sum_address_p): Likewise.
7128 (rs6000_legitimize_address): Likewise.
7129 (rs6000_legitimize_reload_address): Likewise.
7130 (rs6000_secondary_reload_inner): Likewise.
7131 (rs6000_preferred_reload_class): Likewise.
7132 (rs6000_output_move_128bit): Likewise.
7134 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7136 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7137 Generate SImode target register for null target.
7138 <case IX86_BUILTIN_XGETBV>: Ditto.
7139 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7140 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7142 2018-05-10 Carl Love <cel@us.ibm.com>
7144 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7145 dcbtt and dcbtstt if operands[2] is 0.
7147 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7150 * config/i386/sse.md (usadv64qi): New expander.
7152 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7154 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7155 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7156 -maltivec=be support.
7157 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7158 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7159 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7160 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7161 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7162 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7163 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7164 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7165 altivec_vsumsws): Adjust.
7166 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7167 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7168 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7170 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7171 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7172 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7173 (altivec_lve<VI_char>x): Delete expand.
7174 (*altivec_lve<VI_char>x_internal): Rename to...
7175 (altivec_lve<VI_char>x): ... this.
7176 (altivec_lvxl_<mode>): Delete expand.
7177 (*altivec_lvxl_<mode>_internal): Rename to ...
7178 (altivec_lvxl_<mode>): ... this.
7179 (altivec_stvxl_<mode>): Delete expand.
7180 (*altivec_stvxl_<mode>_internal): Rename to ...
7181 (altivec_stvxl_<mode>): ... this.
7182 (altivec_stve<VI_char>x): Delete expand.
7183 (*altivec_stve<VI_char>x_internal): Rename to ...
7184 (altivec_stve<VI_char>x): ... this.
7185 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7186 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7187 reduc_plus_scal_<mode>): Adjust.
7188 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7190 (rs6000_cpu_cpp_builtins): Adjust.
7191 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7192 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7193 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7194 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7195 -maltivec=be support.
7196 (rs6000_split_vec_extract_var): Adjust.
7197 (rs6000_split_v4si_init): Adjust.
7198 (swap_selector_for_mode): Delete.
7199 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7200 altivec_expand_stvex_be): Delete.
7201 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7202 -maltivec=be support.
7203 (rs6000_gimple_fold_builtin): Ditto.
7204 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7206 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7207 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7208 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7209 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7210 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7211 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7212 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7213 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7214 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7215 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7216 anonymous split): Adjust.
7217 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7218 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7220 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7222 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7223 when --with-gxx-include-dir is also specified.
7224 * configure: Regenerate.
7226 2018-05-09 Jim Wilson <jimw@sifive.com>
7229 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7230 * config/riscv/t-withmultilib: New.
7231 * config/riscv/withmultilib.h: New.
7232 * doc/install.texi: Document RISC-V --with-multilib-list support.
7234 2018-05-09 Richard Biener <rguenther@suse.de>
7236 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7238 (vect_bb_vectorization_profitable_p): Adjust. Compute
7239 actual scalar cost using the cost vector and the add_stmt_cost
7242 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7244 PR rtl-optimization/85645
7245 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7246 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7248 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7250 PR rtl-optimization/85645
7251 * shrink-wrap.c (spread_components): Return a boolean saying if
7252 anything was changed.
7253 (try_shrink_wrapping_separate): Iterate spread_components until
7254 nothing changes anymore.
7256 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7258 PR rtl-optimization/85645
7259 * regrename.c (build_def_use): Also kill the chains that include the
7260 destination of a REG_CFA_REGISTER note.
7262 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7264 PR rtl-optimization/85645
7265 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7266 insn that has a REG_CFA_REGISTER note.
7268 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7270 * cfgexpand.c (expand_clobber): New function.
7271 (expand_gimple_stmt_1): Use it.
7272 * tree-vect-stmts.c (vect_clobber_variable): New function,
7274 (vectorizable_simd_clone_call): ...here.
7275 (vectorizable_store): Emit a clobber either side of an
7276 IFN_STORE_LANES sequence.
7277 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7279 2018-05-09 Tom de Vries <tom@codesourcery.com>
7282 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7283 (define_insn "trap_if_false"): Add exit after trap.
7285 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7287 PR rtl-optimization/85638
7288 * bb-reorder.c: Include common/common-target.h.
7289 (create_forwarder_block): New function extracted from...
7290 (fix_up_crossing_landing_pad): ...here. Rename into...
7291 (dw2_fix_up_crossing_landing_pad): ...this.
7292 (sjlj_fix_up_crossing_landing_pad): New function.
7293 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7294 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7295 from both partitions and exit the loop after one iteration.
7297 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7300 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7302 (Basic PowerPC Built-in Functions): The new name of the
7303 subsection previously known as "PowerPC Built-in Functions".
7304 (Basic PowerPC Built-in Functions Available on all Configurations):
7306 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7308 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7310 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7312 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7315 2018-05-08 Jim Wilson <jimw@sifive.com>
7317 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7318 (LD_EMUL_SUFFIX): New.
7319 (LINK_SPEC): Use it.
7321 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7323 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7325 (Basic PowerPC Built-in Functions): The new name of the
7326 subsection previously known as "PowerPC Built-in Functions".
7327 (Basic PowerPC Built-in Functions Available on all Configurations):
7329 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7331 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7333 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7335 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7338 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7341 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7342 after cmpelim optimization.
7344 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7346 * config.gcc: Support "goldmont".
7347 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7348 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7350 * config/i386/i386.c (m_GOLDMONT): Define.
7351 (processor_target_table): Add "goldmont".
7352 (PTA_GOLDMONT): Define.
7353 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7354 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7355 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7356 (fold_builtin_cpu): Add "goldmont".
7357 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7358 (ix86_option_override_internal): Add "goldmont".
7359 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7360 (processor_type): Add PROCESSOR_GOLDMONT.
7361 * config/i386/i386.md: Add CPU "glm".
7362 * config/i386/glm.md: New file.
7363 * config/i386/x86-tune.def: Add m_GOLDMONT.
7364 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7366 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7369 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7371 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7372 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7373 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7376 * config/i386/i386.c (ix86_fold_builtin): Handle
7377 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7380 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7381 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7383 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7386 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7387 (check_arch): Likewise.
7388 (check_fpu): Return the result rather than printing it.
7389 (end arch): Fix operator precedence.
7390 (end cpu): Likewise.
7391 (END): Print the result from check_fpu.
7393 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7394 Alan Hayward <alan.hayward@arm.com>
7395 David Sherwood <david.sherwood@arm.com>
7397 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7398 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7399 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7400 (*fcmuo<mode>_and): New patterns.
7402 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7404 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7405 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7406 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7407 (cmp_op, sve_imm_con): New code attributes.
7408 (SVE_COND_INT_CMP, imm_con): Delete.
7409 (cmp_op): Remove above unspecs from int attribute.
7410 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7412 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7413 comparison-specific unspecs.
7414 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7415 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7416 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7417 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7418 (*vec_fcm<cmp_op><mode>): Rename to...
7419 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7420 (*vec_fcmuo<mode>): Rename to...
7421 (*fcmuo<mode>): ...this and adjust likewise.
7422 (*pred_fcm<cmp_op><mode>): New pattern.
7423 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7424 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7426 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7428 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7429 (aarch64_emit_sve_predicated_cond): New function.
7430 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7431 (aarch64_emit_unspec_cond_or): Replace with...
7432 (aarch64_emit_sve_or_conds): ...this new function. Use
7433 aarch64_emit_sve_ptrue_op for the individual comparisons and
7434 aarch64_emit_binop to OR them together.
7435 (aarch64_emit_inverted_unspec_cond): Replace with...
7436 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7437 aarch64_emit_sve_ptrue_op for the comparison and
7438 aarch64_emit_unop to invert the result.
7439 (aarch64_expand_sve_vec_cmp_float): Update after the above
7440 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7442 2018-05-07 Nathan Sidwell <nathan@acm.org>
7444 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7445 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7446 (Backwards Compatibility): Likewise.
7448 2018-05-07 Luis Machado <luis.machado@linaro.org>
7452 2018-05-07 Luis Machado <luis.machado@linaro.org>
7454 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7455 <prefetch_dynamic_strides>: New const bool field.
7456 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7457 prefetch_dynamic_strides.
7458 (exynosm1_prefetch_tune): Likewise.
7459 (thunderxt88_prefetch_tune): Likewise.
7460 (thunderx_prefetch_tune): Likewise.
7461 (thunderx2t99_prefetch_tune): Likewise.
7462 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7464 (aarch64_override_options_internal): Update to set
7465 PARAM_PREFETCH_DYNAMIC_STRIDES.
7466 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7467 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7468 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7469 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7470 prefetch-dynamic-strides setting.
7472 2018-05-07 Luis Machado <luis.machado@linaro.org>
7474 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7475 <minimum_stride>: New const int field.
7476 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7477 minimum_stride field.
7478 (exynosm1_prefetch_tune): Likewise.
7479 (thunderxt88_prefetch_tune): Likewise.
7480 (thunderx_prefetch_tune): Likewise.
7481 (thunderx2t99_prefetch_tune): Likewise.
7482 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7483 (aarch64_override_options_internal): Update to set
7484 PARAM_PREFETCH_MINIMUM_STRIDE.
7485 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7486 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7487 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7488 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7489 stride is constant and is below the minimum stride threshold.
7491 2018-05-07 Luis Machado <luis.machado@linaro.org>
7493 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7496 2018-05-07 Luis Machado <luis.machado@linaro.org>
7498 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7499 <prefetch_dynamic_strides>: New const bool field.
7500 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7501 prefetch_dynamic_strides.
7502 (exynosm1_prefetch_tune): Likewise.
7503 (thunderxt88_prefetch_tune): Likewise.
7504 (thunderx_prefetch_tune): Likewise.
7505 (thunderx2t99_prefetch_tune): Likewise.
7506 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7508 (aarch64_override_options_internal): Update to set
7509 PARAM_PREFETCH_DYNAMIC_STRIDES.
7510 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7511 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7512 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7513 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7514 prefetch-dynamic-strides setting.
7516 2018-05-07 Luis Machado <luis.machado@linaro.org>
7518 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7519 <minimum_stride>: New const int field.
7520 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7521 minimum_stride field.
7522 (exynosm1_prefetch_tune): Likewise.
7523 (thunderxt88_prefetch_tune): Likewise.
7524 (thunderx_prefetch_tune): Likewise.
7525 (thunderx2t99_prefetch_tune): Likewise.
7526 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7527 (aarch64_override_options_internal): Update to set
7528 PARAM_PREFETCH_MINIMUM_STRIDE.
7529 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7530 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7531 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7532 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7533 stride is constant and is below the minimum stride threshold.
7535 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7538 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7539 the type is addressable. Don't force op into register if it has
7542 2018-05-05 Roland McGrath <mcgrathr@google.com>
7545 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7546 any section for which we don't know a specific type it should have,
7547 regardless of name. Previously this was done only for the exact
7548 names ".init_array", ".fini_array", and ".preinit_array".
7549 (default_elf_asm_named_section): Add comment about
7550 relationship with default_section_type_flags and SECTION_NOTYPE.
7551 (get_section): Don't consider it a type conflict if one side has
7552 SECTION_NOTYPE and the other doesn't, as long as neither has the
7553 SECTION_BSS et al used in the default_section_type_flags logic.
7555 2018-05-05 Tom de Vries <tom@codesourcery.com>
7558 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7559 (workaround_barsyncs): New function.
7560 (nvptx_reorg): Use workaround_barsyncs.
7561 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7562 (define_expand "nvptx_membar_cta"): New define_expand.
7563 (define_insn "*nvptx_membar_cta"): New insn.
7565 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7567 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7568 To improve optimization opportunities.
7569 * builtin-types.def: The new needed builtin types for the above.
7571 2018-05-04 Richard Biener <rguenther@suse.de>
7573 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7574 * gimple-ssa-store-merging.c
7575 (imm_store_chain_info::output_merged_store): Remove redundant create,
7576 release split_store vector contents on failure.
7577 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7578 scalar stmt vector on cache hit.
7580 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7582 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7584 * config.gcc (powerpc-xilinx-eabi*): Remove.
7585 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7587 (fusion_addis_mem_combo_load): Ditto.
7588 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7590 (rs6000_cpu_cpp_builtins): Ditto.
7591 * config/rs6000/rs6000-linux.c
7592 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7593 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7594 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7596 (rs6000_setup_reg_addr_masks): Ditto.
7597 (rs6000_init_hard_regno_mode_ok): Ditto.
7598 (rs6000_option_override_internal): Ditto.
7599 (legitimate_lo_sum_address_p): Ditto.
7600 (rs6000_legitimize_address): Ditto.
7601 (rs6000_legitimize_reload_address): Ditto.
7602 (rs6000_legitimate_address_p): Ditto.
7603 (abi_v4_pass_in_fpr): Ditto.
7604 (setup_incoming_varargs): Ditto.
7605 (rs6000_gimplify_va_arg): Ditto.
7606 (rs6000_split_multireg_move): Ditto.
7607 (rs6000_savres_strategy): Ditto.
7608 (rs6000_emit_prologue_components): Ditto.
7609 (rs6000_emit_epilogue_components): Ditto.
7610 (rs6000_emit_prologue): Ditto.
7611 (rs6000_emit_epilogue): Ditto.
7612 (rs6000_elf_file_end): Ditto.
7613 (rs6000_function_value): Ditto.
7614 (rs6000_libcall_value): Ditto.
7615 * config/rs6000/rs6000.h: Ditto.
7616 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7617 (TARGET_MINMAX): ... this. New.
7618 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7619 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7620 (*movsi_internal1_single): Delete.
7621 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7622 mfpu=, mxilinx-fpu): Delete.
7623 * config/rs6000/singlefp.h: Delete.
7624 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7625 * config/rs6000/t-rs6000: Ditto.
7626 * config/rs6000/t-xilinx: Delete.
7627 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7628 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7629 (VStype_simple): Delete.
7630 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7631 * config/rs6000/xfpu.h: Delete.
7632 * config/rs6000/xfpu.md: Delete.
7633 * config/rs6000/xilinx.h: Delete.
7634 * config/rs6000/xilinx.opt: Delete.
7635 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7636 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7638 2018-05-04 Tom de Vries <tom@codesourcery.com>
7641 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7644 2018-05-04 Richard Biener <rguenther@suse.de>
7647 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7648 (expand_complex_div_wide): Likewise.
7649 (expand_complex_operations_1): Likewise.
7650 (expand_complex_libcall): Preserve EH info of the original stmt.
7651 (tree_lower_complex): Handle removed blocks.
7652 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7653 on complex multiplication and division libcall builtins.
7655 2018-05-04 Richard Biener <rguenther@suse.de>
7658 * fold-const.c (negate_expr_p): Restrict negation of operand
7659 zero of a division to when we know that can happen without
7661 (fold_negate_expr_1): Likewise.
7663 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7666 * real.h (real_nextafter): Declare.
7667 * real.c (real_nextafter): New function.
7668 * fold-const-call.c (fold_const_nextafter): New function.
7669 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7670 CASE_CFN_NEXTTOWARD.
7671 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7672 even when arg1_mode is different from arg0_mode.
7674 2018-05-03 Nathan Sidwell <nathan@acm.org>
7676 * doc/extend.texi (Deprecated Features): Remove
7678 (Backwards Compatibility): Likewise.
7679 * doc/invoke.texi (C++ Language Options): Likewise.
7680 (C++ Dialect Options): Likewise.
7682 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7685 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7686 _mm512_mask_mullox_epi64): New intrinsics.
7688 2018-05-03 Tom de Vries <tom@codesourcery.com>
7691 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7692 dump files): Add offload-tree.
7694 2018-05-03 Richard Biener <rguenther@suse.de>
7696 PR tree-optimization/85615
7697 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7698 to loops not nested in BBs loop father to avoid creating multi-entry
7701 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7703 PR tree-optimization/70291
7704 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7705 arguments. Change return type to tree. Emit libcall as a new
7706 statement rather than replacing existing one when inplace_p is true.
7707 (expand_complex_multiplication_components): New function.
7708 (expand_complex_multiplication): Expand floating-point complex
7709 multiplication using the above.
7710 (expand_complex_division): Rename inner_type parameter to type.
7711 Update expand_complex_libcall call-site.
7712 (expand_complex_operations_1): Update expand_complex_multiplication
7713 and expand_complex_division call-sites.
7715 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7718 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7719 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7720 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7721 the highest significant bit of the shift count mask is clear. In
7722 check whether and[sq]i3 is needed verify that all significant bits
7723 of the shift count other than the highest are set.
7725 2018-05-02 Tom de Vries <tom@codesourcery.com>
7728 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7729 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7730 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7731 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7732 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7733 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7734 __builtin_goacc_parlevel_size.
7736 2018-05-02 Richard Biener <rguenther@suse.de>
7738 PR tree-optimization/85597
7739 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7740 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7743 2018-05-02 Tom de Vries <tom@codesourcery.com>
7746 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7747 dump files): Add ltrans-tree.
7749 2018-05-02 Tom de Vries <tom@codesourcery.com>
7752 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7753 dump files): Add wpa-ipa.
7755 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7757 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7758 powerpc*-*-linux*paired* target.
7759 * config/rs6000/750cl.h: Delete.
7760 * config/rs6000/paired.h: Delete.
7761 * config/rs6000/paired.md: Delete.
7762 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7764 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7765 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7766 comment. Remove paired float support.
7767 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7768 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7770 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7771 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7773 * config/rs6000/rs6000.c: Remove paired float support.
7774 (paired_expand_vector_init, paired_expand_vector_move,
7775 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7776 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7777 paired_expand_builtin, paired_expand_predicate_builtin,
7778 paired_init_builtins): Delete.
7779 * config/rs6000/rs6000.h: Remove paired float support.
7780 * config/rs6000/rs6000.md: Remove paired float support.
7781 (move_from_CR_ov_bit): Delete.
7782 * config/rs6000/rs6000.opt (mpaired): Delete.
7783 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7784 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7786 2018-05-02 Richard Biener <rguenther@suse.de>
7789 * gimplify.c (gimplify_save_expr): When in SSA form allow
7790 SAVE_EXPRs to compute to SSA vars.
7792 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7795 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7796 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7797 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7798 clobber operands[2], instead use a new pseudo. Formatting fixes.
7800 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7802 PR tree-optimization/85586
7803 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7804 exit early for statements in the same group if the accesses are
7807 2018-05-02 Tom de Vries <tom@codesourcery.com>
7810 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7813 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7815 PR tree-optimization/85143
7816 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7818 2018-05-01 Tom de Vries <tom@codesourcery.com>
7821 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7822 not found" error message.
7824 2018-05-01 Tom de Vries <tom@codesourcery.com>
7827 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7828 * vec.c (test_ordered_remove_if): New function.
7829 (vec_c_tests): Call test_ordered_remove_if.
7830 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7831 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7832 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7833 VEC_ORDERED_REMOVE_IF.
7835 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7837 PR tree-optimization/82665
7838 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7839 pointer subtraction where arguments come from a memchr call.
7841 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7843 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7844 --push-state --as-needed and --pop-state instead of --as-needed and
7845 --no-as-needed if ld supports it.
7846 * configure: Regenerated.
7849 * doc/install.texi2html: Replace _002d with - and _002a with * in
7850 generated html files using sed.
7852 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7855 * gcc-rich-location.c (blank_line_before_p): New function.
7856 (use_new_line): New function.
7857 (gcc_rich_location::add_fixit_insert_formatted): New function.
7858 * gcc-rich-location.h
7859 (gcc_rich_location::add_fixit_insert_formatted): New function.
7861 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7863 * selftest.c (assert_streq): Rename "expected" and "actual" to
7864 "val1" and "val2". Extend NULL-handling to cover both inputs
7865 symmetrically, while still requiring both to be non-NULL for a pass.
7866 * selftest.h (assert_streq): Rename "expected" and "actual" to
7868 (ASSERT_EQ): Likewise.
7869 (ASSERT_EQ_AT): Likewise.
7870 (ASSERT_KNOWN_EQ): Likewise.
7871 (ASSERT_KNOWN_EQ_AT): Likewise.
7872 (ASSERT_NE): Likewise.
7873 (ASSERT_MAYBE_NE): Likewise.
7874 (ASSERT_MAYBE_NE_AT): Likewise.
7875 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7876 the assertion to pass.
7877 (ASSERT_STREQ_AT): Likewise.
7879 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7881 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7882 interaction with -pie.
7884 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7886 * selftest.h: Fix alphabetization of per-source-file selftest
7889 2018-04-30 Jason Merrill <jason@redhat.com>
7891 PR c++/61982 - dead stores to destroyed objects.
7892 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7895 2018-04-30 Jason Merrill <jason@redhat.com>
7897 * tree.c (build_clobber): New.
7898 * tree.h: Declare it.
7899 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7901 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7903 * diagnostic-show-locus.c (layout::layout): Update for
7904 location_get_source_line returning a char_span.
7905 (struct char_span): Move to input.h.
7906 (struct correction): Update for fields in char_span becoming
7908 (struct source_line): Update for location_get_source_line
7909 returning a char_span.
7910 (layout::print_line): Likewise.
7911 * edit-context.c (edited_file::print_content): Likewise.
7912 (edited_file::print_diff_hunk): Likewise.
7913 (edited_file::print_run_of_changed_lines): Likewise.
7914 (edited_file::get_num_lines): Likewise.
7915 (edited_line::edited_line): Likewise.
7916 * final.c (asm_show_source): Likewise.
7917 * input.c (location_get_source_line): Convert return type
7918 from const char * to char_span, losing the final "line_len"
7920 (dump_location_info): Update for the above.
7921 (get_substring_ranges_for_loc): Likewise. Use a char_span
7922 when handling the literal within the line.
7923 (test_reading_source_line): Update for location_get_source_line
7924 returning a char_span.
7925 * input.h (class char_span): Move here from
7926 diagnostic-show-locus.c, converting from a struct to a class.
7927 Make data members private.
7928 (char_span::operator bool): New.
7929 (char_span::length): New.
7930 (char_span::get_buffer): New.
7931 (char_span::operator[]): New.
7932 (char_span::subspan): Make const.
7933 (char_span::xstrdup): New.
7934 (location_get_source_line): Convert return type from const char *
7935 to char_span, losing the final "line_size" param.
7937 2018-04-30 Jan Hubicka <jh@suse.cz>
7939 * lto-wrapper.c (ltrans_priorities): New static var.
7940 (cmp_priority): New.
7941 (run_gcc): Read priorities and if doing parallel build order
7942 the Makefile by them.
7944 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7946 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7948 2018-04-30 Richard Biener <rguenther@suse.de>
7950 * tree-cfg.c (verify_address): Remove base argument, add
7951 flag whether to check TREE_ADDRESSABLE and do that.
7952 (verify_expr): Remove.
7953 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7954 (verify_gimple_assign_single): Likewise.
7955 (verify_gimple_switch): Likewise.
7956 (verify_expr_location_1): Dereference tp once. Add (disabled)
7957 piece from verify_expr.
7958 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7960 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7962 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7964 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7966 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7967 (small_data_pattern): Likewise.
7968 (arc_rewrite_small_data): Likewise.
7969 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7970 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7971 (get_symbol_alignment): New function.
7972 (legitimate_small_data_address_p): Likewise.
7973 (legitimate_scaled_address): Update, call
7974 legitimate_small_data_address_p.
7975 (output_sdata): New static variable.
7976 (arc_print_operand): Update how we handle small data operands.
7977 (arc_print_operand_address): Likewise.
7978 (arc_legitimate_address_p): Update, use
7979 legitimate_small_data_address_p.
7980 (arc_rewrite_small_data_p): Remove.
7981 (arc_rewrite_small_data_1): Likewise.
7982 (arc_rewrite_small_data): Likewise.
7983 (small_data_pattern): Likewise.
7984 (compact_sda_memory_operand): Update to use
7985 legitimate_small_data_address_p and get_symbol_alignment.
7986 (prepare_move_operands): Don't rewite sdata pattern.
7987 (prepare_extend_operands): Remove.
7988 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7990 (zero_extendqisi2): Likewise.
7991 (zero_extendhisi2): Likewise.
7992 (extendqihi2): Likewise.
7993 (extendqisi2): Likewise.
7994 (extendhisi2): Likewise.
7998 * config/arc/constraints.md (Usd): Change it to memory constraint.
8000 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8002 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8003 as source of std instructions.
8004 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8005 allow 6-bit constants as source for store instructions.
8006 (movdi_insn): Update instruction pattern to allow 6-bit constants
8007 as source for store instructions.
8009 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8011 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8013 2018-04-30 Nathan Sidwell <nathan@acm.org>
8014 Sandra Loosemore <sandra@codesourcery.com>
8016 * dumpfile.c (dump_open): Allow '-' for stdout.
8017 * doc/invoke.texi (Developer Options): Document dump filename
8018 determination early. Document stdin/stdout selection.
8020 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8022 Microblaze Target: PIC data text relative
8024 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8025 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8027 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8028 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8029 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8030 New addressing mode for data-text relative position indepenedent code.
8031 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8032 'ADDRESS_SYMBOLIC_TXT_REL'.
8033 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8034 (microblaze_legitimate_pic_operand): Exclude function calls from
8035 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8036 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8038 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8039 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8040 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8041 for 'address + offset'.
8042 (microblaze_expand_prologue): Add new function prologue call for
8044 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8045 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8046 table in case of TARGET_PIC_DATA_TEXT_REL.
8047 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8048 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8049 Add new macros 'UNSPEC_TEXT',
8050 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8051 + exclude function calls from 'UNSPEC_PLT' in case of data text
8053 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8054 new target hook for generating address diff vector tables in case of
8056 * doc/tm.texi : Regenerate.
8057 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8058 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8059 of addr diff vector generation.
8060 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8061 target hook definition.
8062 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8063 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8064 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8065 option for data text relative.
8067 2018-04-30 Richard Biener <rguenther@suse.de>
8069 * tree-chrec.h (evolution_function_is_constant_p): Remove
8071 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8073 2018-04-30 Richard Biener <rguenther@suse.de>
8076 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8078 2018-04-30 Richard Biener <rguenther@suse.de>
8080 PR tree-optimization/28364
8081 PR tree-optimization/85275
8082 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8083 copying first exit test.
8085 2018-04-28 Mark Wielaard <mark@klomp.org>
8087 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8089 (dwarf_AT): Handle DW_AT_addr_base.
8090 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8092 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8095 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8096 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8097 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8098 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8100 2018-04-28 Richard Biener <rguenther@suse.de>
8102 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8103 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8104 to reflect use. Only add interesting stmts.
8106 2018-04-27 Martin Jambor <mjambor@suse.cz>
8109 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8110 the jump function allows for passing through aggregate values.
8112 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8114 * input.h (in_system_header_at): Convert from macro to inline
8116 (from_macro_expansion_at): Likewise.
8117 (from_macro_definition_at): Likewise.
8119 2018-04-27 Jeff Law <law@redhat.com>
8121 * config.gcc: Mark tile* targets as deprecated/obsolete.
8123 2018-04-27 Richard Biener <rguenther@suse.de>
8125 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8128 2018-04-27 Richard Biener <rguenther@suse.de>
8130 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8132 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8134 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8135 with Yd constraint. Set "preferred_for_speed" attribute from
8136 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8138 (*movdi_internal): Ditto.
8139 (movti_interunit splitters): Remove
8140 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8141 (movdi_interunit splitters): Ditto.
8142 * config/i386/constraints.md (Ye): Remove.
8143 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8145 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8148 * config/aarch64/constraints.md (Usg): Limit to 31.
8151 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8153 PR tree-optimization/85529
8154 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8155 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8156 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8157 zero extension or masking of the MSB bit.
8158 (optimize_range_tests): Add FIRST_BB argument, pass it through
8159 to optimize_range_tests_var_bound.
8160 (maybe_optimize_range_tests, reassociate_bb): Adjust
8161 optimize_range_tests callers.
8163 2018-04-26 Richard Biener <rguenther@suse.de>
8164 Jakub Jelinek <jakub@redhat.com>
8166 * cgraph.h (symbol_table): Just declare debug method here.
8167 * symtab.c (symbol_table::debug): Define.
8169 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8171 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8173 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8175 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8176 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8177 (*movdi_internal): Substitute Yi and Yj constraint with x
8178 and Ym and Yn constraint with y constraint. Update "isa"
8179 attribute and set "preferred_for_speed" attribute from
8180 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8181 (*movsi_internal): Ditto.
8182 (*movdf_internal): Ditto.
8183 (*movsf_internal): Ditto.
8184 (*zero_extendsidi2): Ditto.
8185 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8186 (sse2_loadld): Ditto.
8187 (*vec_extract<ssevecmodelower>_0): Ditto.
8188 (*vec_extractv4si_0_zext_sse4): Ditto.
8189 (vec_concatv2di): Ditto.
8190 (*vec_dup<mode>): Ditto.
8191 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8192 * config/i386/constraints.md (Yi): Remove.
8197 2018-04-26 Nathan Sidwell <nathan@acm.org>
8199 * dumpfile.c (dump_open): New.
8200 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8201 (dump_finish): Detect stdio/stderr by value not name.
8203 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8205 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8207 2018-04-26 Tom de Vries <tom@codesourcery.com>
8210 * config/nvptx/nvptx.c (verify_neutering_jumps)
8211 (verify_neutering_labels): New function
8212 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8214 2018-04-26 Tom de Vries <tom@codesourcery.com>
8217 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8218 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8221 2018-04-26 Richard Biener <rguenther@suse.de>
8222 Tom de Vries <tom@codesourcery.com>
8225 * lto-streamer-out.c (output_function): Fixup loops if required to match
8226 discovery done in the reader.
8228 2018-04-26 Richard Biener <rguenther@suse.de>
8230 PR tree-optimization/85116
8231 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8232 have a loop exit from the single latch predecessor. Remove
8233 case of header with just condition.
8234 (ch_base::copy_headers): Exclude infinite loops from any
8236 (pass_ch::execute): Record exits.
8238 2018-04-26 Richard Biener <rguenther@suse.de>
8240 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8241 prologue cost vector and pass it to vect_get_load_cost.
8242 (vect_get_peeling_costs_all_drs): Likewise.
8243 (vect_peeling_hash_get_lowest_cost): Likewise.
8244 (vect_enhance_data_refs_alignment): Likewise.
8246 2018-04-26 Richard Biener <rguenther@suse.de>
8249 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8250 checking of integer<->pointer conversions.
8251 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8252 sign-/zero-extending pointer types.
8253 (expand_omp_for_static_chunk): Likewise.
8255 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8256 Jean Lee <xiaoyur347@gmail.com>
8258 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8259 (TARGET_ASAN_SHADOW_OFFSET): Define.
8260 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8261 true for -fsanitize=address.
8263 2018-04-25 Mark Wielaard <mark@klomp.org>
8265 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8268 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8270 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8271 than "alu", remove explicit "memory" and "imm_disp" attributes.
8272 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8275 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8276 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8277 gen_lowpart_no_emit.
8279 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8282 * config/i386/i386.c (ix86_expand_builtin): Change memory
8283 operand to XI, extend p0 to Pmode.
8284 * config/i386/i386.md: Change unspec volatile and operand
8285 1 mode to XI, change operand 0 mode to P.
8287 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8289 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8290 GET_MODE_MASK before any checking.
8291 (nds32_can_use_bset_p): Likewise.
8292 (nds32_can_use_btgl_p): Likewise.
8294 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8296 * config/nds32/nds32-doubleword.md: New define_split pattern for
8297 illegal register number.
8299 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8301 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8303 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8305 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8307 2018-04-25 Richard Biener <rguenther@suse.de>
8309 * lto-streamer.h (LTO_major_version): Bump to 8.
8311 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8313 * BASE-VER: Set to 9.0.0.
8315 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8317 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8318 in __abskf2 and __powikf2.
8320 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8323 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8324 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8325 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8326 Use the above on operand 2. Reindent.
8327 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8329 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8332 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8334 * config/i386/i386.opt (mcet): Removed.
8335 * doc/install.texi: Remove -mcet documentation.
8336 * doc/invoke.texi: Likewise.
8338 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8341 * doc/install.texi: Remove -mcet from bootstrap-cet.
8343 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8346 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8347 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8351 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8352 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8353 containing a CONST_VECTOR.
8355 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8357 * doc/install.texi: Update newlib dependency for nvptx.
8359 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8362 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8363 instead of INTVAL when shifting x left.
8365 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8367 PR tree-optimization/85478
8368 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8369 vect_grouped_store_supported for single element vectors.
8371 2018-04-24 Richard Biener <rguenther@suse.de>
8374 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8375 load cost increase to the case of non-constant step.
8377 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8380 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8381 destination if any_malformed_asm.
8383 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8386 * expr.c (store_field): In the bitfield case, if the value comes from
8387 a function call and is returned in registers by means of a PARALLEL,
8388 do not change the mode of the temporary unless BLKmode and VOIDmode.
8390 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8392 PR rtl-optimization/85423
8393 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8394 dependencies to debug insns when the previous insn is non-debug.
8396 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8398 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8399 enums into a single definition.
8400 (fls): Fix predicates and printing.
8403 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8405 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8406 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8407 and short u6 immediate.
8408 (check_if_valid_sleep_operand): Remove.
8409 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8411 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8413 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8414 flag_always_save_lp condition.
8415 * config/nds32/nds32.opt (malways-save-lp): New option.
8417 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8419 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8420 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8421 * config/nds32/nds32.h
8422 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8423 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8425 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8427 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8428 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8430 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8431 Chung-Ju Wu <jasonwucj@gmail.com>
8433 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8435 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8436 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8438 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8440 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8442 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8444 * config/nds32/nds32-protos.h (nds32_data_alignment,
8445 nds32_local_alignment): Declare.
8446 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8447 nds32_local_alignment): New functions.
8448 (TARGET_CONSTANT_ALIGNMENT): Define.
8449 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8451 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8453 * config/nds32/nds32.c
8454 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8455 (TARGET_MODES_TIEABLE_P): Likewise.
8457 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8459 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8462 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8463 Chung-Ju Wu <jasonwucj@gmail.com>
8465 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8466 for unaligned access.
8467 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8469 * config/nds32/nds32-intrinsic.md: Likewise.
8470 * config/nds32/nds32_intrinsic.h: Likewise.
8471 * config/nds32/nds32.h (nds32_builtins): Likewise.
8472 * config/nds32/nds32.opt (munaligned-access): New option.
8473 * config/nds32/nds32.c (nds32_asm_file_start): Display
8474 flag_unaligned_access status.
8476 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8478 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8479 -mno-relax is present.
8480 * config/riscv/linux.h (LINK_SPEC): Ditto.
8482 2018-04-20 Martin Sebor <msebor@redhat.com>
8485 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8487 (gimple_fold_builtin_stxcpy_chk): Same.
8488 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8490 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8493 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8494 __powikf2 when long double is IEEE 128-bit.
8496 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8498 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8499 step to make sure stack always aligned.
8501 2018-04-20 Carl Love <cel@us.ibm.com>
8504 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8505 size check for arg0.
8507 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8508 Tom de Vries <tom@codesourcery.com>
8511 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8512 Emit insns for calls too.
8513 (nvptx_find_par): Always look for worker-level predecessor insn.
8514 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8516 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8517 (nvptx_process_pars): Propagate frames for calls.
8519 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8522 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8524 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8525 (ix86_handle_option): Don't handle OPT_mibt.
8526 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8528 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8530 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8531 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8532 (ix86_target_macros): Define __CET__ with flag_cf_protection
8533 for -fcf-protection.
8534 * config/i386/i386.c (isa2_opts): Remove -mibt.
8535 * config/i386/i386.h (TARGET_IBT): Removed.
8536 (TARGET_IBT_P): Likewise.
8537 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8538 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8539 * config/i386/i386.opt (mcet): Update help message.
8542 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8543 -mcet as an alias for -mshstk.
8545 2018-04-20 Richard Biener <rguenther@suse.de>
8548 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8549 complexity by forcing a single use of the multiply operand.
8551 2018-04-20 Martin Jambor <mjambor@suse.cz>
8554 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8555 recursion dependency to only apply to non-clones.
8557 2018-04-20 Martin Jambor <mjambor@suse.cz>
8560 * ipa-cp.c (create_specialized_node): Check that clones of
8561 self-recursive edges exist during IPA-CP.
8563 2018-04-19 Toon Moene <toon@moene.org>
8565 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8568 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8570 PR tree-optimization/85467
8571 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8572 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8573 VECTOR_CST element to type.
8575 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8578 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8579 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8580 (builtin_longjmp): Likewise.
8581 (save_stack_nonlocal): New pattern.
8582 (restore_stack_nonlocal): Likewise.
8584 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8587 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8588 Replace ASM_OUTPUT_LABEL with fprintf.
8590 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8593 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8594 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8595 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8596 define __IBT__ and __SHSTK__ for -fcf-protection.
8597 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8599 (ix86_trampoline_init): Likewise.
8600 (x86_output_mi_thunk): Likewise.
8601 (ix86_notrack_prefixed_insn_p): Likewise.
8602 (ix86_option_override_internal): Don't disallow -fcf-protection.
8603 * config/i386/i386.md (rdssp<mode>): Also enable for
8605 (incssp<mode>): Likewise.
8606 (nop_endbr): Likewise.
8607 * config/i386/i386.opt (mcet): Change help message to built-in
8611 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8612 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8613 enable CET built-in functions.
8615 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8617 * common/config/i386/i386-common.c
8618 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8619 OPTION_MASK_ISA_MOVDIRI_UNSET,
8620 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8621 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8622 * config.gcc (movdirintrin.h): New header.
8623 * config/i386/cpuid.h (bit_MOVDIRI,
8624 bit_MOVDIR64B): New bits.
8625 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8627 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8628 (VOID, PVOID, PCVOID)): New function types.
8629 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8630 __builtin_ia32_directstoreu_u64,
8631 __builtin_ia32_movdir64b): New builtins.
8632 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8633 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8635 (ix86_valid_target_attribute_inner_p): Ditto.
8636 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8637 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8638 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8639 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8640 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8641 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8642 (movdiri<mode>, movdir64b_<mode>): New.
8643 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8644 * config/i386/immintrin.h: Include movdirintrin.h.
8645 * config/i386/movdirintrin.h: New file.
8646 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8648 2018-04-19 Richard Biener <rguenther@suse.de>
8651 * cfg.c (clear_bb_flags): When loop state says we have
8652 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8654 2018-04-19 Richard Biener <rguenther@suse.de>
8656 PR tree-optimization/84737
8657 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8658 copying restrict info.
8659 (vect_setup_realignment): Use it.
8660 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8661 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8662 the first DR to all generated stores.
8663 (vectorizable_load): Likewise for loads.
8665 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8667 PR tree-optimization/85446
8668 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8669 the integral and pointer types to have the same precision.
8671 * doc/install.texi: Document --disable-cet being the default and
8674 2018-04-18 Martin Liska <mliska@suse.cz>
8676 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8679 2018-04-18 Martin Liska <mliska@suse.cz>
8682 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8685 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8686 arguments if they are comparable.
8688 2018-04-18 Martin Liska <mliska@suse.cz>
8691 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8694 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8697 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8700 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8701 ENDBR after calling __morestack.
8703 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8706 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8707 by using gcc_base_ver to generate a gcc_driver_version, and use
8708 it when generating GCC_DRIVER_NAME.
8709 * configure: Regenerate.
8711 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8714 * config.gcc: Obsolete powerpc*-*-*spe*.
8716 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8719 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8720 (stabstr_D): Change type of unum from unsigned int to
8721 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8724 2018-04-17 Jim Wilson <jimw@sifive.com>
8727 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8728 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8729 Set arg_pointer_offset after using pretend_args_size.
8731 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8733 PR rtl-optimization/85431
8734 * dse.c (record_store): Ignore zero width stores.
8737 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8738 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8739 __builtin_stack_restore rather than after it.
8740 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8741 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8742 argument instead of virtual_dynamic_stack_rtx.
8744 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8746 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8748 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8749 Add note to error message to explain internal mapping of overloaded
8750 built-in function name to non-overloaded built-in function name.
8751 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8754 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8757 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8758 where the inputs overlap with the output.
8760 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8762 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8763 (=v, v) alternative and explicit "memory" attribute.
8764 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8765 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8767 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8768 "sselog1" type instead of "sselog".
8769 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8770 "sselog". Remove explicit "memory" attribute.
8771 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8772 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8774 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8775 "sselog1" type instead of "sselog". Remove explicit "memory"
8777 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8778 use "sselog1" type instead of "sselog". Remove explicit "memory"
8780 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8781 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8783 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8784 "sselog1" type instead of "sselog". Remove explicit "memory"
8786 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8787 use "sselog1" type instead of "sselog". Remove explicit "memory"
8791 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8794 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8797 2018-04-17 Martin Jambor <mjambor@suse.cz>
8800 * ipa-cp.c (create_specialized_node): Call
8801 expand_all_artificial_thunks if necessary.
8803 2018-04-17 Martin Liska <mliska@suse.cz>
8806 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8807 in message, remote space in between '_G' and '('.
8809 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8812 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8813 avx512f_vmcmp<mode>3<round_saeonly_name>,
8814 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8815 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8816 avx512f_rndscale<mode><round_saeonly_name>,
8817 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8818 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8819 Use %<iptr>2 instead of %2 for -masm=intel.
8820 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8821 avx512f_vcvttss2usi<round_saeonly_name>,
8822 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8824 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8825 avx512f_vcvttsd2usi<round_saeonly_name>,
8826 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8827 Use %q1 instead of %1 for -masm=intel.
8828 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8829 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8830 of %3 for -masm=intel.
8831 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8833 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8835 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8837 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8838 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8840 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8842 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8843 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8844 %g1 and one with %0 and %1.
8845 (avx512er_vmrcp28<mode><round_saeonly_name>,
8846 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8848 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8849 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8850 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8851 of %0 and %{%4%} for -masm=intel.
8852 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8853 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8854 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8855 order of %0 and %{%5%}%{z%} for -masm=intel.
8857 2018-04-17 Jan Hubicka <jh@suse.cz>
8860 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8862 2018-04-17 Martin Liska <mliska@suse.cz>
8865 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8866 for target_clone error message. Make default implementation
8867 clone to be a local declaration.
8868 (separate_attrs): Add new argument and check for an empty
8870 (expand_target_clones): Handle it.
8871 (ipa_target_clone): Make redirection just for target_clones
8874 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8875 Tom de Vries <tom@codesourcery.com>
8878 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8879 tiled basic blocks without omp continue statements.
8881 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8884 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8885 vec_extract expression as having side effects to make sure it gets
8888 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8891 * config/i386/i386.c (get_builtin_code_for_version): Check
8894 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8897 * gcc/config.gcc: Support "skylake".
8898 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8900 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8901 (processor_target_table): Add "skylake".
8902 (ix86_option_override_internal): Add "skylake".
8903 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8904 PROCESSOR_CANNONLAKE.
8905 (get_builtin_code_for_version): Fix priority for
8906 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8907 PROCESSOR_SKYLAKE-AVX512.
8908 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8909 (processor_type): Add PROCESSOR_SKYLAKE.
8911 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8912 Jason Merrill <jason@redhat.com>
8915 * convert.c (convert_to_integer_1): Use direct recursion for
8916 enumeral types and types with a precision less than the number
8917 of bits in their mode.
8919 2018-04-16 Julia Koval <julia.koval@intel.com>
8922 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8923 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8925 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8928 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8929 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8930 and -mno-direct-move.
8932 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8935 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8936 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8937 Ensure negative shifts result in {0}.
8939 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8941 PR rtl-optimization/79916
8942 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8943 regs (if any) to define how to gnerate SD moves when LRA is in
8946 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8948 PR rtl-optimization/85393
8949 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8950 * except.c (expand_dw2_landing_pad_for_region): Make static.
8951 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8952 a label and unconditional jump to old_bb, rather than
8953 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8956 PR rtl-optimization/85376
8957 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8958 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8959 instead of a specific value.
8961 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8962 Bin Cheng <bin.cheng@arm.com>
8964 PR tree-optimization/82965
8965 PR tree-optimization/83991
8966 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8967 by_profile_only parameter.
8968 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8969 information if the loop was predicted to iterate too many times.
8970 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8972 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8975 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8978 2018-04-13 Martin Liska <mliska@suse.cz>
8979 Jakub Jelinek <jakub@redhat.com>
8982 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8983 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8984 * builtins.c (expand_builtin_memory_copy_args): Use
8985 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8986 handle dest_addr == pc_rtx.
8988 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8991 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8992 asked to not generate direct moves.
8993 (fix_trunc<mode>si2_stfiwx): Similar.
8994 (fix_trunc<mode>si2_internal): Similar.
8996 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8999 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9000 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9001 lookup if dest in some wider mode is known to be const0_rtx and
9002 if so, record permanent equivalence for it to be ZERO_EXTEND of
9003 the narrower mode destination.
9005 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9007 * lto-streamer-out.c (output_function): Revert 259346.
9008 * omp-expand.c (expand_oacc_for): Likewise.
9010 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9012 PR rtl-optimization/85354
9013 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9014 * sel-sched.c (sel_global_init): ... here.
9016 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9019 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9020 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9021 mode for PE-COFF targets.
9022 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9023 (i386_pe_asm_lto_end): Likewise.
9024 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9025 (TARGET_ASM_LTO_END): Likewise.
9026 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9027 (i386_pe_asm_lto_start): New function.
9028 (i386_pe_asm_lto_end): Likewise.
9030 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9031 Richard Biener <rguenther@suse.de>
9034 * lto-streamer-out.c (output_function): Fix CFG loop state before
9036 * omp-expand.c (expand_oacc_for): Handle calls to internal
9037 functions like regular functions.
9039 2018-04-12 Richard Biener <rguenther@suse.de>
9042 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9043 for the early LTO debug to properly generate references to it
9044 during DIE emission. Do not re-use that for the skeleton for
9046 (dwarf2out_early_finish): Likewise.
9048 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9051 * config/i386/sse.md
9052 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9053 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9054 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9055 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9056 and output is a reg, avoid creating invalid lowpart subreg, but
9057 instead split into a 512-bit move. Don't split if not AVX512VL,
9058 input is xmm16+ reg and output is a mem.
9059 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9060 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9061 xmm16+ reg and output is a mem.
9063 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9065 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9066 also for flag_dwarf2_cfi_asm.
9068 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9070 PR rtl-optimization/85342
9071 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9072 a bool scalar var inside of the loop instead. Don't try to update
9073 recog_data.operand after failed apply_change_group.
9075 2018-04-12 Tom de Vries <tom@codesourcery.com>
9078 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9079 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9080 array with flexible array member as array without given dimension.
9081 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9082 argument for undefined param to true.
9084 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9087 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9088 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9089 from PowerPC section.
9090 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9091 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9092 help text that is too long.
9093 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9094 help text that is too long.
9095 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9096 help text that is too long.
9098 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9100 * config/alpha/alpha.md (stack_probe_internal): Rename
9101 from "probe_stack". Update all callers.
9103 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9105 PR rtl-optimization/84566
9106 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9107 sched_macro_fuse_insns.
9109 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9112 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9113 (compute_block_dependences): ... from here.
9115 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9117 PR tree-optimization/85331
9118 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9119 from int to HOST_WIDE_INT.
9121 2018-04-11 Martin Jambor <mjambor@suse.cz>
9124 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9125 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9126 not the same as the source val.
9127 (cgraph_edge_brings_value_p): New parameter.
9128 (gather_edges_for_value): Pass destination value to
9129 cgraph_edge_brings_value_p.
9130 (perhaps_add_new_callers): Likewise.
9131 (get_info_about_necessary_edges): Likewise and exclude values brought
9132 only by self-recursive edges.
9133 (create_specialized_node): Redirect only clones of self-calling edges.
9134 (+self_recursive_pass_through_p): New function.
9135 (find_more_scalar_values_for_callers_subset): Use it.
9136 (find_aggregate_values_for_callers_subset): Likewise.
9137 (known_aggs_to_agg_replacement_list): Removed.
9138 (decide_whether_version_node): Re-calculate known constants for all
9139 remaining context clones.
9141 2018-04-11 Richard Biener <rguenther@suse.de>
9144 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9145 from early DWARF output.
9146 (dwarf2out_early_finish): Output line info unconditionally into
9147 early DWARF and add reference to it.
9149 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9152 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9153 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9154 other than V2DFmode using iptr mode attribute.
9155 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9157 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9159 PR rtl-optimization/84659
9160 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9162 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9165 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9167 (output_loc_list): Pass address of a dummy size variable even in the
9168 locview handling loop.
9169 (index_location_lists): Add comment on why skip_loc_list_entry can't
9172 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9175 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9178 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9181 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9182 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9183 and -mstring-compare-inline-limit.
9185 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9188 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9189 for stack clash protection in a register whenever we need it to be in
9192 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9194 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9195 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9197 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9200 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9202 (mlong-double-): Ditto.
9203 * config/rs6000/sysv4.opt (msdata=): Ditto.
9204 (mtls-size=): Ditto.
9206 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9208 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9209 erroneous entries for
9210 "vector int vec_ldl (int, long int *)", and
9211 "vector unsigned int vec_ldl (int, unsigned long int *)".
9212 Add comments and entries for
9213 "vector bool char vec_ldl (int, bool char *)",
9214 "vector bool short vec_ldl (int, bool short *)",
9215 "vector bool int vec_ldl (int, bool int *)",
9216 "vector bool long long vec_ldl (int, bool long long *)",
9217 "vector pixel vec_ldl (int, pixel *)",
9218 "vector long long vec_ldl (int, long long *)",
9219 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9220 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9221 type tree bool_long_long_type_node and correct definition of
9222 bool_V2DI_type_node to make reference to this new type tree.
9223 (rs6000_mangle_type): Replace erroneous reference to
9224 bool_long_type_node with bool_long_long_type_node.
9225 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9226 comments to emphasize sign distinctions for char and int types and
9227 replace RS6000_BTI_bool_long constant with
9228 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9229 use of RS6000_BTI_pixel.
9230 (bool_long_type_node): Remove this macro definition.
9231 (bool_long_long_type_node): New macro definition
9233 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9235 PR rtl-optimization/85300
9236 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9237 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9238 simplify_unary_operation fails.
9240 2018-04-10 Martin Liska <mliska@suse.cz>
9242 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9243 cgraph_edge and ipa_ref.
9245 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9249 * config/i386/sse.md
9250 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9251 computation of the VEC_MERGE selector from mask.
9252 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9253 Fix decoding of the VEC_MERGE selector into mask.
9255 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9257 PR tree-optimization/85286
9258 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9260 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9262 * final.c (final_1): Set insn_last_address as well as
9263 insn_current_address.
9265 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9268 * explow.c (emit_stack_probe): Call validize_mem on memory location
9269 before passing it to gen_probe_stack. Create address operand and
9270 legitimize it for the probe_stack_address case.
9272 2018-04-09 Jan Hubicka <jh@suse.cz>
9275 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9276 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9277 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9279 (free_lang_data): Rebuild type inheritance graph.
9281 2018-04-09 Martin Sebor <msebor@redhat.com>
9283 * invoke.texi (-finline-small-functions): Mention other optimization
9285 (-findirect-inlining, -fpartial-inlining): Same.
9286 (-finline-functions-called-once): Same.
9287 (-freorder-blocks-and-partition): Same.
9289 2018-04-09 Jan Hubicka <jh@suse.cz>
9292 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9293 jumps; choose last target that matches the criteria (i.e.
9294 no partition changes for non-crossing jumps).
9295 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9296 support for redirecting crossing jumps to non-crossing.
9298 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9300 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9301 also for naked functions.
9303 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9305 * config/arc/arc.md (add_shift): New pattern.
9306 (add_shift2): Likewise.
9307 (sub_shift): Likewise.
9308 (sub_shift_cmp0_noout): Likewise.
9309 (compare_si_ashiftsi): Likewise.
9310 (xbfu_cmp0_noout): New combine pattern.
9311 (xbfu_cmp0"): Likewise.
9312 (movsi_set_cc_insn): Place the predicable variant first.
9313 (commutative_binary_cmp0_noout): Remove clobber.
9314 (commutative_binary_cmp0): New pattern.
9315 (noncommutative_binary_cmp0): Likewise.
9316 (noncommutative_binary_cmp0_noout): Likewise.
9317 (noncommutative_binary_comparison_result_used): Removed.
9318 (rsub_cmp0): New pattern.
9319 (rsub_cmp0_noout): Likewise.
9320 (extzvsi): Changed, keep only meaningful variants.
9321 (SQH, SEZ): New iterators.
9322 (SQH_postfix): New mode attribute.
9323 (SEZ_prefix): New code attribute.
9324 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9325 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9326 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9328 (noncommutative_operator): Check the availability of barrel
9331 2018-04-09 Richard Biener <rguenther@suse.de>
9333 PR tree-optimization/85284
9334 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9335 Only use the niter constraining form of simple_iv when the exit
9338 2018-04-09 Tom de Vries <tom@codesourcery.com>
9341 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9342 (define_expand "*memory_barrier"): New define_expand.
9343 (define_insn "memory_barrier"): New insn.
9345 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9347 PR rtl-optimization/80463
9348 PR rtl-optimization/83972
9349 PR rtl-optimization/83480
9351 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9352 correct producer for the insn.
9353 (tidy_control_flow): Fixup seqnos in case of debug insns.
9355 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9357 PR rtl-optimization/83913
9359 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9360 different sched-times when merging exprs.
9362 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9364 PR rtl-optimization/83962
9366 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9367 tidy_fallthru_edge and tidy_control_flow.
9369 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9371 PR rtl-optimization/83530
9373 * sel-sched.c (force_next_insn): New global variable.
9374 (remove_insn_for_debug): When force_next_insn is true, also leave only
9375 next insn in the ready list.
9376 (sel_sched_region): When the region wasn't scheduled, make another pass
9377 over it with force_next_insn set to 1.
9379 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9381 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9383 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9384 for interrupt control.
9385 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9386 functions for interrupt control.
9387 * config/nds32/nds32-intrinsic.md: Likewise.
9388 * config/nds32/nds32_intrinsic.h: Likewise.
9389 * config/nds32/nds32.h (nds32_builtins): Likewise.
9391 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9393 * config/nds32/nds32.c (nds32_init_machine_status,
9394 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9395 strict_aligned_p field.
9396 (nds32_expand_to_rtl_hook): New function.
9397 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9398 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9400 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9401 Chung-Ju Wu <jasonwucj@gmail.com>
9403 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9404 * config/nds32/nds32-n7.md: New file.
9405 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9406 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9408 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9409 * config/nds32/nds32.md (pipeline_model): Add n7.
9410 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9411 * config/nds32/pipelines.md: Include n7 settings.
9413 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9414 Chung-Ju Wu <jasonwucj@gmail.com>
9416 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9417 * config/nds32/nds32-e8.md: New file.
9418 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9419 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9421 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9422 * config/nds32/nds32.md (pipeline_model): Add e8.
9423 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9424 * config/nds32/pipelines.md: Include e8 settings.
9426 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9427 Chung-Ju Wu <jasonwucj@gmail.com>
9429 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9430 * config/nds32/nds32-n8.md: New file.
9431 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9432 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9434 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9435 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9436 * config/nds32/nds32.md (pipeline_model): Add n8.
9437 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9438 * config/nds32/pipelines.md: Include n8 settings.
9440 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9441 Chung-Ju Wu <jasonwucj@gmail.com>
9443 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9444 * config/nds32/nds32-n9-2r1w.md: New file.
9445 * config/nds32/nds32-n9-3r2w.md: New file.
9446 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9447 nds32_register_ports): New or modify for cpu n9.
9448 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9450 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9451 * config/nds32/nds32-utils.c: New file.
9452 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9453 TARGET_MUL_SLOW): Define.
9454 * config/nds32/nds32.md (pipeline_model): New attribute.
9455 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9456 New options that support cpu n9.
9457 * config/nds32/pipelines.md: Include n9 settings.
9458 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9460 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9462 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9463 information if necessary.
9464 (output_cond_branch_compare_zero): Likewise.
9465 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9466 (nds32_target_alignment): Refine for alignment.
9467 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9468 (FUNCTION_BOUNDARY): Modify.
9469 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9471 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9473 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9475 * config/nds32/constants.md (unspec_volatile_element): Add values for
9476 TLB operation and data prefetch.
9477 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9478 functions for TLB operation and data prefetch.
9479 * config/nds32/nds32-intrinsic.md: Likewise.
9480 * config/nds32/nds32_intrinsic.h: Likewise.
9481 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9482 (nds32_print_operand): Likewise.
9483 * config/nds32/nds32.h (nds32_builtins): Likewise.
9485 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9486 Andrew Pinski <pinsika@gcc.gnu.org>
9489 * match.pd: Use constant_boolean_node of correct type instead of
9490 boolean_true_node or boolean_false_node for simplifying
9491 pointer comparisons to zero.
9493 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9495 PR tree-optimization/80021
9496 * tree.c (verify_type_variant): Make error call in verify_variant_match
9497 translatable and remove final full stop.
9499 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9501 * config/nds32/constants.md (unspec_volatile_element): Add
9502 UNSPEC_VOLATILE_EH_RETURN.
9503 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9504 nds32_output_stack_pop): Support dwarf exception handling process.
9505 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9506 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9507 exception handling process.
9508 (nds32_compute_stack_frame): Likewise.
9509 (nds32_return_addr_rtx): Likewise.
9510 (nds32_initial_elimination_offset): Likewise.
9511 (nds32_expand_prologue): Likewise.
9512 (nds32_expand_epilogue): Likewise.
9513 (nds32_dynamic_chain_address): New function.
9514 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9516 (DYNAMIC_CHAIN_ADDRESS): Define.
9517 (EH_RETURN_DATA_REGNO): Define.
9518 (EH_RETURN_STACKADJ_RTX): Define.
9519 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9520 patterns for dwarf exception handling.
9522 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9524 * config/nds32/nds32.h: Clean up obsolete macros.
9526 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9528 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9529 Add enum values for particular instructions.
9530 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9531 particular intrinsic functions.
9532 * config/nds32/nds32-intrinsic.md: Likewise.
9533 * config/nds32/nds32_intrinsic.h: Likewise.
9534 * config/nds32/nds32.h (nds32_builtins): Likewise.
9535 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9536 (btst, ave): New patterns for particular instructions.
9538 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9540 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9541 Add enum values for atomic load/store and memory sync.
9542 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9544 * config/nds32/nds32-intrinsic.md: Likewise.
9545 * config/nds32/nds32_intrinsic.h: Likewise.
9546 * config/nds32/nds32.h (nds32_builtins): Likewise.
9548 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9550 PR tree-optimization/85257
9551 * fold-const.c (native_encode_vector): If not all elts could fit
9552 and off is -1, return 0 rather than offset.
9553 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9554 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9555 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9556 adjust buffer in native_interpret_expr call.
9558 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9560 * config/nds32/constants.md (unspec_volatile_element): Add cache
9561 control enum values.
9562 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9563 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9564 * config/nds32/nds32.c (nds32_cctl_names): New.
9565 (nds32_print_operand): Handle cache control register names.
9566 * config/nds32/nds32.h (nds32_builtins): New enum values.
9567 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9569 * config/nds32/nds32.md (type): Add mmu.
9570 * config/nds32/pipelines.md (simple_insn): Add mmu.
9572 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9574 * config/nds32/nds32.md (type): Remove call.
9575 * config/nds32/pipelines.md (simple_insn): Likewise.
9577 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9579 * config/nds32/constants.md (unspec_volatile_element): Add
9580 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9581 UNSPEC_VOLATILE_FMFCFG.
9582 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9583 description for fmfcfg and fmfcsr.
9584 (bdesc_1arg): Add fmtcsr.
9585 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9586 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9587 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9588 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9589 unspec_fmfcfg): New patterns.
9590 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9591 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9592 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9593 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9594 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9595 __nds32__fmfcfg): Define.
9597 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9599 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9600 intrinsic register names.
9601 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9602 intrinsic register enum values and macros.
9604 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9606 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9607 for load/store addressing form.
9608 (nds32_print_operand_address): Likewise.
9610 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9613 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9614 based on LABEL_REF. Remove useless assertion.
9615 (pic_address_needs_scratch): Fix formatting.
9616 (sparc_legitimize_pic_address): Minor tweaks.
9617 (sparc_delegitimize_address): Adjust assertion accordingly.
9618 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9619 into symbolic_operand.
9620 (movsi_high_pic_label_ref): Likewise.
9621 (movsi_lo_sum_pic_label_ref): Likewise.
9622 (movdi_pic_label_ref): Likewise.
9623 (movdi_high_pic_label_ref): Likewise.
9624 (movdi_lo_sum_pic_label_ref): Likewise.
9626 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9628 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9629 custom LIB_SPEC setup.
9631 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9632 Kito Cheng <kito.cheng@gmail.com>
9634 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9635 * config/riscv/freebsd.h: New.
9637 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9639 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9640 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9643 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9644 Kito Cheng <kito.cheng@gmail.com>
9646 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9647 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9648 * config/nds32/nds32-protos.h (nds32_output_call,
9649 nds32_output_return): Declare.
9650 * config/nds32/nds32.md: Refine all the call and return patterns.
9652 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9655 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9656 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9658 PR rtl-optimization/84872
9659 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9660 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9663 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9665 * expr.c (copy_blkmode_to_reg): Revert 254862.
9666 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9668 2018-04-06 Richard Biener <rguenther@suse.de>
9671 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9672 after seeing a component reference with an adjacent field. Treat
9673 refs to arrays at struct end of external decls similar to
9674 refs to unconstrained commons.
9676 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9679 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9680 look through SAVE_EXPRs with non-side-effects argument. Adjust
9682 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9685 2018-04-06 Richard Biener <rguenther@suse.de>
9688 * alias.c (find_base_term): New wrapper around find_base_term
9689 unwinding CSELIB_VAL_PTR changes.
9690 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9693 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9695 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9697 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9698 constant definitions.
9699 ("nop"): lr 0,0 -> nopr r0
9700 ("nop_lr0", "nop_lr1"): New insn definitions.
9702 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9704 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9705 NDS32_V3PUSH_AVAILABLE_P macro.
9707 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9708 Chung-Ju Wu <jasonwucj@gmail.com>
9710 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9711 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9712 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9713 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9714 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9715 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9716 * config/nds32/constraints.md: New constraints and checking for hard
9717 float configuration.
9718 * config/nds32/iterators.md: New mode iterator and attribute for hard
9719 float configuration.
9720 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9722 * config/nds32/nds32-fpu.md: New file.
9723 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9724 deal with hard float code generation.
9725 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9727 (abi_type, float_reg_number): New enum type.
9728 * config/nds32/nds32-predicates.c: New predicates for hard float.
9729 * config/nds32/nds32-protos.h: Declare functions for hard float.
9730 * config/nds32/nds32.c: Implementation for hard float configuration.
9731 * config/nds32/nds32.h: Definitions for hard float configuration.
9732 * config/nds32/nds32.md: Include hard float machine description and
9733 modify patterns for hard float configuration.
9734 * config/nds32/nds32.opt: New options for hard float configuration.
9735 * config/nds32/predicates.md: New predicates for hard float
9738 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9740 * common/config/nds32/nds32-common.c
9741 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9743 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9746 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9747 CONSTRUCTOR_ELT (ctor, ...)->value.
9749 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9752 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9754 2018-04-05 Tom de Vries <tom@codesourcery.com>
9757 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9760 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9761 Kito Cheng <kito.cheng@gmail.com>
9763 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9764 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9765 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9766 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9768 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9769 Kito Cheng <kito.cheng@gmail.com>
9771 * config/nds32/constraints.md (Ufe): New memory constraint.
9772 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9773 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9774 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9776 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9777 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9779 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9781 * config/nds32/nds32.md: Use optimize_size in the condition for
9782 alu-shift instructions.
9784 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9786 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9788 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9790 * config/nds32/nds32.md (negsi2): Refine pattern.
9792 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9793 Chung-Ju Wu <jasonwucj@gmail.com>
9795 * config/nds32/iterators.md (shift_rotate): New code iterator.
9796 (shift): New code attribute.
9797 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9798 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9799 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9800 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9801 bit-wise operations.
9802 (andsi3, *andsi3): Ditto.
9803 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9804 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9805 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9806 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9807 nds32_ior_operand, nds32_xor_operand): New predicates.
9809 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9811 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9812 (addsi3, subsi3): ... this.
9814 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9816 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9818 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9820 * config/nds32/nds32.md: Adjust indention.
9822 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9824 * config/nds32/nds32.md (feature): New attribute.
9826 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9828 * config/nds32/nds32.md (subtype): New attribute.
9830 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9833 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9834 expansion to perform a bitwise AND of the argument followed by a
9835 boolean negation of the result.
9837 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9839 PR rtl-optimization/84878
9840 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9841 the basic block. Assert the use reference is not artificial and that
9842 it has an associated insn.
9844 2018-04-04 Michael Matz <matz@suse.de>
9846 * builtins.c (compute_objsize): Pass correct operand
9847 to array_at_struct_end_p.
9849 2018-04-04 Richard Biener <rguenther@suse.de>
9852 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9853 from contexts for DINFO_LEVEL_TERSE and below.
9855 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9857 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9858 resiter_operand condition.
9859 * config/nds32/nds32.md (*move<mode>): Ditto.
9861 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9862 Monk Chiang <sh.chiang04@gmail.com>
9864 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9866 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9868 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9870 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9871 Kito Cheng <kito.cheng@gmail.com>
9873 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9874 nds32_cond_code_str, output_cond_branch,
9875 output_cond_branch_compare_zero, nds32_expand_cbranch,
9876 nds32_expand_cstore, nds32_expand_movcc,
9877 nds32_output_cbranchsi4_equality_zero,
9878 nds32_output_cbranchsi4_equality_reg,
9879 nds32_output_cbranchsi4_equality_reg_or_const_int,
9880 nds32_output_cbranchsi4_greater_less_zero: New functions.
9881 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9882 nds32_expand_cstore, nds32_expand_movcc,
9883 nds32_output_cbranchsi4_equality_zero,
9884 nds32_output_cbranchsi4_equality_reg,
9885 nds32_output_cbranchsi4_equality_reg_or_const_int,
9886 nds32_output_cbranchsi4_greater_less_zero): Declare.
9887 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9888 nds32_rimm11s_operand): New predicates.
9889 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9890 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9893 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9895 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9896 * config/nds32/nds32.md: Ditto.
9897 * config/nds32/pipelines.md: Ditto.
9899 2018-04-04 Richard Biener <rguenther@suse.de>
9901 PR tree-optimization/85168
9902 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9903 propagating abnormals.
9905 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9907 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9909 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9910 Kito Cheng <kito.cheng@gmail.com>
9912 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9913 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9914 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9915 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9916 * config/nds32/nds32.md (sibcall_internal): New.
9917 (sibcall_register): Remove.
9918 (sibcall_immediate): Remove.
9919 (sibcall_value_internal): New.
9920 (sibcall_value_register): Remove.
9921 (sibcall_value_immediate): Remove.
9922 * config/nds32/predicates.md (nds32_general_register_operand): New.
9923 (nds32_call_address_operand): New.
9925 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9927 PR rtl-optimization/85167
9928 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9929 bb_defs if *split_p, instead preinitialize it to NULL.
9931 PR tree-optimization/85156
9932 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9933 evaluating the argument multiple times.
9935 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9937 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9939 (_mm_cvtpd_ps): Likewise.
9940 (_mm_cvttpd_epi32): Likewise.
9941 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9942 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9943 vector, pixel, and bool following altivec.h include.
9945 2018-04-03 Martin Sebor <msebor@redhat.com>
9947 * doc/extend.texi (Common Function Attributes): Clarify.
9948 (const attribute): Likewise.
9949 (pure attribute): Likewise.
9951 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9954 * config/i386/i386.c (ix86_expand_vector_set): Use
9955 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9957 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9959 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9960 instructions when changing rounding bits to preserve precision bits
9961 in the x87 control word.
9963 2018-04-03 Martin Liska <mliska@suse.cz>
9965 PR tree-optimization/82491
9966 * rtl.h (strip_offset_and_add): Replace += suboffset with
9967 poly_uint64 () + suboffset.
9969 2018-03-29 Martin Liska <mliska@suse.cz>
9970 Martin Jambor <mjambor@suse.cz>
9973 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9974 param_type is not an integral or pointer type.
9976 2018-04-03 Richard Biener <rguenther@suse.de>
9978 * sese.h (recompute_all_dominators): Remove.
9980 2018-04-02 Martin Sebor <msebor@redhat.com>
9982 * doc/invoke.texi (-Wrestrict): Fix typos.
9984 2018-04-02 Jim Wilson <jimw@sifive.com>
9986 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9987 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9988 (<optab>di3, <optab>si3_extend): Likewise.
9989 (<optab>si3_mask, <optab>si3_mask_1): New.
9990 (<optab>di3_mask, <optab>di3_mask_1): New.
9991 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9992 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9993 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9995 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9997 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10000 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10002 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10003 (nds32_canonicalize_comparison): New function.
10005 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10006 Kito Cheng <kito.cheng@gmail.com>
10007 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10009 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10010 * config/nds32/constants.md (unspec_volatile_element): Add
10011 UNSPEC_VOLATILE_RELAX_GROUP.
10012 * config/nds32/nds32-relax-opt.c: New file.
10013 * config/nds32/nds32-predicates.c
10014 (nds32_symbol_load_store_p): New function.
10015 * config/nds32/nds32-protos.h
10016 (nds32_symbol_load_store_p): Declare function.
10017 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10018 * config/nds32/nds32.c
10019 (nds32_register_pass): New function to register pass.
10020 (nds32_register_passes): New function to register passes.
10021 * config/nds32/nds32.md (relax_group): New pattern.
10022 * config/nds32/nds32.opt (mrelax-hint): New option.
10023 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10025 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10027 * config/nds32/t-nds32: Modify files dependency.
10029 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10031 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10032 (PROFILE_HOOK): Define its implementation.
10034 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10036 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10037 type and 32-bit size.
10039 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10041 PR middle-end/85090
10042 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10043 (V_128_256): New mode iterator.
10044 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10045 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10046 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10048 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10049 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10051 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10054 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10055 NaN inputs correctly.
10057 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10060 * config/rs6000/vsx.md (??r): New mode attribute.
10061 (*vsx_mov<mode>_64bit): Use it.
10062 (*vsx_mov<mode>_32bit): Likewise.
10064 2018-03-30 Martin Sebor <msebor@redhat.com>
10066 PR tree-optimization/84818
10067 * builtins.c (check_access): Use warning_n.
10069 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10072 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10074 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10077 2018-03-30 Julia Koval <julia.koval@intel.com>
10080 * x86-tune.def (movx, partial_reg_dependency): Enable for
10083 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10085 PR inline-asm/84985
10086 * lra-constraints.c (process_alt_operands): Move setting
10087 this_alternative_matches below.
10089 2018-03-29 Martin Liska <mliska@suse.cz>
10092 * doc/invoke.texi: Document how LTO works with debug info.
10093 Describe auto-load support of binutils. Mention 'x86-64'
10094 as valid option value of -march option.
10096 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10098 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10101 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10102 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10103 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10106 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10109 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10110 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10111 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10112 for RS6000_BTM_POWERPC64.
10113 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10114 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10115 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10118 (DIVDEU): Likewise.
10120 2018-03-28 Carl Love <cel@us.ibm.com>
10123 2017-09-27 Carl Love <cel@us.ibm.com>
10125 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10126 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10127 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10130 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10132 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10133 instead of __vector bool.
10134 (_mm_max_pu8): Likewise.
10135 (_mm_min_pi16): Likewise.
10137 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10140 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10141 (DIVWEUO): Likewise.
10142 (DIVDEO): Likewise.
10143 (DIVDEUO): Likewise.
10144 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10145 DIVWEUO and DIVDEUO.
10146 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10147 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10148 (div_extend): Likewise.
10149 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10151 (__builtin_divweuo): Likewise.
10152 (__builtin_divdeo): Likewise.
10153 (__builtin_divdeuo): Likewise.
10155 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10158 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10159 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10161 PR tree-optimization/82004
10162 * gimple-match-head.c (optimize_pow_to_exp): New function.
10163 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10164 Don't fold to exp if optimize_pow_to_exp is false.
10166 2018-03-28 Martin Liska <mliska@suse.cz>
10169 * calls.c (initialize_argument_information): Fix trailing space.
10170 * common.opt: Fix typo and provide better explanation for
10171 -fsanitize-coverage option.
10172 * config/i386/i386.opt: Fix typo.
10174 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10175 Martin Liska <mliska@suse.cz>
10178 * gimplify.c (asan_poison_variable): Don't do the check for
10179 gimplify_omp_ctxp here.
10180 (gimplify_decl_expr): Do it here.
10181 (gimplify_target_expr): Likewise.
10183 2018-03-28 Martin Liska <mliska@suse.cz>
10186 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10187 chkp_type_bounds_count if MPX is not enabled.
10189 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10191 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10193 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10196 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10197 function to create the function decl for complex long double
10198 multiply and divide for -mabi=ieeelongdouble.
10199 (init_float128_ieee): Call it.
10201 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10204 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10205 -fcf-protection=branch -mibt.
10206 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10208 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10211 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10212 arm_word_relocations.
10214 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10217 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10218 extern array declarations.
10220 2018-03-27 Richard Biener <rguenther@suse.de>
10222 PR middle-end/84067
10223 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10224 explicit single_use checks.
10226 2018-03-27 Richard Biener <rguenther@suse.de>
10228 PR tree-optimization/85082
10229 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10232 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10234 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10235 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10236 Turn on fasynchronous-unwind-tables and funwind-tables.
10238 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10241 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10242 (*bmi_blsr_<mode>_ccz): Ditto.
10244 2018-03-26 Tom de Vries <tom@codesourcery.com>
10246 PR tree-optimization/85063
10247 * omp-general.c (offloading_function_p): New function. Factor out
10249 * omp-offload.c (pass_omp_target_link::gate): ... here.
10250 * omp-general.h (offloading_function_p): Declare.
10251 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10252 with attribute omp declare target for offloading functions.
10254 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10256 PR tree-optimization/84005
10257 * tree-data-ref.h (get_base_for_alignment): Declare.
10258 * tree-data-ref.c (get_base_for_alignment_1): New function.
10259 (get_base_for_alignment): Likewise.
10260 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10261 get_base_for_alignment to find a suitable base object, instead
10262 of always using drb->base_address.
10264 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10266 PR inline-asm/85022
10267 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10268 known size by default.
10270 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10272 PR inline-asm/85030
10273 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10274 and non BLKmode operands.
10276 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10279 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10280 Clean up attributes.
10282 2018-03-23 Richard Biener <rguenther@suse.de>
10285 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10286 we are going to emit early debug for LTO.
10288 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10290 PR inline-asm/85034
10291 * function.c (match_asm_constraints_1): Don't optimize if input
10292 doesn't satisfy general_operand predicate for output's mode.
10294 PR inline-asm/85022
10295 * alias.c (write_dependence_p): Don't require for x_canonicalized
10296 non-VOIDmode if x has VOIDmode.
10299 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10300 just don't try to optimize it rather than assert it never happens.
10302 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10304 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10305 macro expansions for definition of ST_INTERNAL_<mode> and
10306 LD_INTERNAL_<mode> builtins.
10307 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10309 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10311 (altivec_expand_st_builtin): Likewise.
10312 (altivec_expand_builtin): Remove calls to deleted functions.
10313 (rs6000_address_for_altivec): Delete this function.
10314 * config/rs6000/vector.md: Remove expands for
10315 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10317 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10320 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10321 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10322 re-computing once computed.
10323 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10324 (arm_init_machine_status): Initialize
10325 machine->static_chain_stack_bytes.
10327 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10330 * doc/extend.texi: Add four new prototypes for vec_ld.
10331 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10332 definitions for more logical presentation.
10333 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10334 entries for V1TI variants of __builtin_altivec_ld builtin.
10335 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10336 handling of V1TI variant of LVX icode pattern.
10337 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10338 (rs6000_gimple_fold_builtin): Likewise.
10339 (altivec_init_builtins): Add code to define
10340 __builtin_altivec_lvx_v1ti function.
10342 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10344 PR inline-asm/84941
10345 * function.c (match_asm_constraints_1): Don't do the optimization
10346 if input isn't a REG, SUBREG, MEM or constant.
10348 2018-03-22 Tom de Vries <tom@codesourcery.com>
10350 PR tree-optimization/84956
10351 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10352 bb_has_abnormal_pred.
10354 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10357 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10358 DECL_INITIAL (decl) to decl at the end.
10359 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10360 adjust the comment.
10362 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10364 * doc/extend.texi (__builtin_tgmath): Document when complex
10365 integer types are treated as _Complex _Float64.
10367 2018-03-21 Tom de Vries <tom@codesourcery.com>
10369 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10371 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10373 PR tree-optimization/84960
10374 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10375 if it is ENTRY block, move them into single succ of ENTRY in that case.
10377 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10379 PR tree-optimization/84811
10380 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10381 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10382 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10383 (known_subrange_p): Update accordingly. Cast each value involved
10384 in the size comparison, rather than casting the result of the
10387 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10389 PR tree-optimization/84982
10390 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10391 by flipping the least significant bit rather than all bits from
10392 bitpos to bitpos + bitsize - 1.
10394 2018-03-21 Nathan Sidwell <nathan@acm.org>
10396 * doc/extend.texi (Deprecated Features): Remove mention of
10397 long-deleted deprecations.
10399 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10402 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10403 * configure: Regenerate.
10405 2018-03-21 Tom de Vries <tom@codesourcery.com>
10407 PR tree-optimization/83126
10408 * tree-parloops.c (num_phis): New function.
10409 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10411 2018-03-21 Nathan Sidwell <nathan@acm.org>
10413 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10414 mention anon-struct/union members and trailing attributes.
10416 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10418 PR tree-optimization/84969
10419 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10420 builtin memset partitions if they set different rhs values.
10422 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10424 PR rtl-optimization/84989
10425 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10426 VEC_DUPLICATE with scalar result mode.
10428 2018-03-21 Martin Liska <mliska@suse.cz>
10431 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10432 not intended return statement.
10434 2018-03-21 Martin Liska <mliska@suse.cz>
10437 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10438 (chkp_find_bound_slots_1): Limit number of iterations.
10440 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10443 * Minor grammar fixes for x86 options.
10445 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10448 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10449 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10451 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10454 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10455 (altivec_lvx_<mode>_1op): Likewise.
10456 (altivec_stvx_<mode>_2op): Likewise.
10457 (altivec_stvx_<mode>_1op): Likewise.
10458 (altivec_lvx_<VM2:mode>): New define_expand.
10459 (altivec_stvx_<VM2:mode>): Likewise.
10460 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10461 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10462 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10463 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10464 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10465 (rs6000_gen_lvx): Likewise.
10466 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10467 (altivec_expand_stv_builtin): Likewise.
10468 (altivec_expand_builtin): Likewise.
10469 * config/rs6000/vector.md: Likewise.
10471 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10474 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10477 2018-03-20 Richard Biener <rguenther@suse.de>
10480 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10481 sign-conversions as zero, fall back to standard scalar_stmt
10484 2018-03-20 Martin Liska <mliska@suse.cz>
10487 * predict.c (rebuild_frequencies): Handle case when we have
10488 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10490 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10493 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10494 flag_section_anchors.
10495 * varasm.c (use_blocks_for_decl_p): Remove hack for
10496 dw2_force_const_mem.
10499 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10501 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10502 be created, use lowpart_subreg of operands[0] rather than operands[0]
10504 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10505 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10506 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10507 and n constraint instead of aarch64_shift_imm_di and Usd.
10508 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10509 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10511 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10514 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10515 to favor GPR over NEON registers.
10516 (<shift>di3_neon): Likewise.
10518 2018-03-20 Tom de Vries <tom@codesourcery.com>
10521 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10522 (nvptx_process_pars): Emit bar.sync asap and alap.
10524 2018-03-20 Tom de Vries <tom@codesourcery.com>
10527 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10528 seen_label if seen_label is already set.
10530 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10533 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10534 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10535 Use 1U instead of 1. Formatting fixes.
10538 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10539 instead of TREE_TYPE (s1) for the return value.
10541 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10543 PR tree-optimization/84946
10544 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10545 bitsize + bitsize in poly_uint64 rather than poly_int64.
10548 * dwarf2asm.c: Include fold-const.c.
10549 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10550 of decl rather than decl itself.
10552 PR rtl-optimization/84643
10553 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10555 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10558 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10559 calling assemble_variable.
10561 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10564 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10565 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10567 2018-03-19 Jim Wilson <jimw@sifive.com>
10570 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10571 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10572 (riscv_first_stack_step): Likewise.
10573 (riscv_option_override): Use STACK_BOUNDARY instead of
10574 MIN_STACK_BOUNDARY.
10575 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10576 MIN_STACK_BOUNDARY.
10577 (BIGGEST_ALIGNMENT): Set to 128.
10578 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10579 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10582 2018-03-19 Richard Biener <rguenther@suse.de>
10584 PR tree-optimization/84933
10585 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10586 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10588 2018-03-19 Richard Biener <rguenther@suse.de>
10590 PR tree-optimization/84859
10591 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10592 (cond_if_else_store_replacement): Perform sinking operation on
10593 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10594 Generalize what a BB with a single eligible store is.
10596 2018-03-19 Richard Biener <rguenther@suse.de>
10598 PR tree-optimization/84929
10599 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10600 chrec_is_positive against non-chrec arg.
10602 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10605 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10607 2018-03-18 Martin Liska <mliska@suse.cz>
10609 PR rtl-optimization/84635
10610 * regrename.c (build_def_use): Use matches_mode only when
10613 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10615 PR tree-optimization/84913
10616 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10617 vectorize chains of COND_EXPRs.
10619 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10621 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10623 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10625 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10627 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10629 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10631 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10632 Kito Cheng <kito.cheng@gmail.com>
10634 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10635 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10636 (nds32_adjust_reg_alloc_order): New function.
10637 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10639 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10641 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10642 nds32_print_operand, nds32_print_operand_address): Use
10643 HOST_WIDE_INT_PRINT_DEC instead.
10645 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10647 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10649 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10652 * config/i386/i386.c (initial_ix86_tune_features,
10653 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10654 unsigned long long.
10655 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10656 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10657 rather than 1u << ix86_tune. Formatting fix.
10658 (ix86_option_override_internal): Change ix86_arch_mask from
10659 unsigned int to unsigned HOST_WIDE_INT, initialize to
10660 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10661 (ix86_function_specific_restore): Likewise.
10663 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10666 * postreload.c (reload_combine_recognize_pattern): Perform
10667 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10668 truncate_int_for_mode the result for the destination's mode.
10671 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10672 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10674 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10677 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10680 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10681 regno_allocno_class_array and sorted_pseudos.
10682 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10683 insns where regno is used.
10685 2018-03-16 Martin Liska <mliska@suse.cz>
10688 * multiple_target.c (create_dispatcher_calls): Redirect
10689 reference in the symbol table.
10691 2018-03-16 Martin Liska <mliska@suse.cz>
10694 * multiple_target.c (create_dispatcher_calls): Redirect also
10697 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10701 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10702 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10703 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10705 2018-03-16 Julia Koval <julia.koval@intel.com>
10707 * doc/invoke.texi (Skylake Server): Add CLWB.
10708 Cannonlake): Remove CLWB.
10710 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10712 PR tree-optimization/84841
10713 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10715 (FLOAT_ONE_CONST_TYPE): Define.
10716 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10717 (sort_by_operand_rank): Put entries with higher constant_type last
10718 rather than first to match comments.
10720 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10722 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10725 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10728 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10729 check and use address of target always.
10731 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10734 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10735 (indirect_thunk_bnd_needed): Likewise.
10736 (indirect_thunks_used): Likewise.
10737 (indirect_thunks_bnd_used): Likewise.
10738 (indirect_return_needed): New.
10739 (indirect_return_bnd_needed): Likewise.
10740 (output_indirect_thunk_function): Add a bool argument for
10742 (output_indirect_thunk_function): Don't generate alias for
10743 function return thunk.
10744 (ix86_code_end): Call output_indirect_thunk_function to generate
10745 function return thunks.
10746 (ix86_output_function_return): Set indirect_return_bnd_needed
10747 and indirect_return_needed instead of indirect_thunk_bnd_needed
10748 and indirect_thunk_needed.
10750 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10752 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10753 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10754 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10756 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10757 Paul Hua <paul.hua.gm@gmail.com>
10760 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10762 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10764 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10765 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10766 resp. SFmode cases.
10768 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10771 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10772 instead of GET_MODE_SIZE when comparing Units.
10774 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10777 * varasm.c (hash_section): Return an unchangeble hash value
10778 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10779 Return !aarch64_can_use_per_function_literal_pools_p ().
10781 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10784 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10785 rather than address of cmode as last argument to prepare_cmp_insn.
10787 2018-03-15 Julia Koval <julia.koval@intel.com>
10789 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10790 F_AVX512VNNI, F_AVX512BITALG): New.
10792 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10795 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10796 insn for floating-point loads and stores.
10798 2018-03-14 Carl Love <cel@us.ibm.com>
10800 * config/rs6000/rs6000-c.c: Add macro definitions for
10801 ALTIVEC_BUILTIN_VEC_PERMXOR.
10802 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10803 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10804 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10805 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10807 * config/doc/extend.texi: Add prototypes for vec_permxor.
10809 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10812 * diagnostic-show-locus.c (class layout_point): Convert m_line
10813 from int to linenum_type.
10814 (line_span::comparator): Use linenum "compare" function when
10815 comparing line numbers.
10816 (test_line_span): New function.
10817 (layout_range::contains_point): Convert param "row" from int to
10819 (layout_range::intersects_line_p): Likewise.
10820 (layout::will_show_line_p): Likewise.
10821 (layout::print_source_line): Likewise.
10822 (layout::should_print_annotation_line_p): Likewise.
10823 (layout::print_annotation_line): Likewise.
10824 (layout::print_leading_fixits): Likewise.
10825 (layout::annotation_line_showed_range_p): Likewise.
10826 (struct line_corrections): Likewise for field m_row.
10827 (line_corrections::line_corrections): Likewise for param "row".
10828 (layout::print_trailing_fixits): Likewise.
10829 (layout::get_state_at_point): Likewise.
10830 (layout::get_x_bound_for_row): Likewise.
10831 (layout::print_line): Likewise.
10832 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10833 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10834 * input.c (selftest::test_linenum_comparisons): New function.
10835 (selftest::input_c_tests): Call it.
10836 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10837 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10838 * selftest.h (ASSERT_GT): New macro.
10839 (ASSERT_GT_AT): New macro.
10840 (ASSERT_LT): New macro.
10841 (ASSERT_LT_AT): New macro.
10843 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10845 PR rtl-optimization/84780
10846 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10848 2018-03-14 Martin Liska <mliska@suse.cz>
10850 * tree.c (record_node_allocation_statistics): Use
10851 get_stats_node_kind.
10852 (get_stats_node_kind): New function extracted from
10853 record_node_allocation_statistics.
10854 (free_node): Use get_stats_node_kind.
10856 2018-03-14 Richard Biener <rguenther@suse.de>
10858 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10859 that the value-set of ANTIC_IN doesn't grow.
10862 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10864 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10865 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10867 2018-03-14 Julia Koval <julia.koval@intel.com>
10869 * config.gcc (icelake-client, icelake-server): New.
10871 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10872 (initial_ix86_arch_features): Ditto.
10873 (PTA_SKYLAKE): Add SGX.
10874 (PTA_ICELAKE): Remove.
10875 (PTA_ICELAKE_CLIENT): New.
10876 (PTA_ICELAKE_SERVER): New.
10877 (ix86_option_override_internal): Split up icelake on icelake client and
10879 (get_builtin_code_for_version): Ditto.
10880 (fold_builtin_cpu): Ditto.
10881 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10882 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10883 * config/i386/i386.h (processor_type): Ditto.
10884 * doc/invoke.texi: Ditto.
10886 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10889 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10890 INTEGER_CST offset, add it together with bitpos / 8 and
10891 sign extend based on POINTER_SIZE.
10895 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10898 * config/i386/constraints.md (Yc): New register constraint.
10899 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10900 Use Yc constraint for alternative 2 of operand 0. Remove
10901 preferred_for_speed attribute.
10903 2018-03-14 Richard Biener <rguenther@suse.de>
10905 PR tree-optimization/84830
10906 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10907 with the old one to avoid oscillations.
10909 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10912 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10914 (assign_by_spills): Return a flag of reload assignment failure.
10915 Do not process the reload assignment failures. Do not spill other
10916 reload pseudos if they has the same reg class. Update n if
10918 (lra_assign): Add a return arg. Set up from the result of
10919 assign_by_spills call.
10920 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10921 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10922 usage_insns if it is not NULL.
10923 (spill_hard_reg_in_range): New function.
10924 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10925 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10926 function prototypes.
10927 (lra_assign): Change prototype.
10928 * lra.c (lra): Add code to deal with fails by splitting hard reg
10931 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10933 * config/riscv/riscv.opt (mrelax): New option.
10934 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10935 "norelax" when riscv_mrelax is disabled.
10936 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10938 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10941 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10942 reassociation for int modes.
10944 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10946 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10947 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10949 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10950 * config/aarch64/aarch64-sve.md
10951 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10952 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10953 (*extend<mode><Vwide>2): Rename to...
10954 (aarch64_sve_extend<mode><Vwide>2): ...this.
10955 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10956 renaming the old pattern to...
10957 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10959 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10960 define_expand, renaming the old pattern to...
10961 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10962 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10963 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10964 account when deciding which SVE instruction the optab should use.
10965 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10967 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10969 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10970 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10971 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10972 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10973 (tlsdesc_small_<mode>): Turn a define_expand and use
10974 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10975 (tlsdesc_small_advsimd_<mode>): ...this.
10976 (tlsdesc_small_sve_<mode>): New pattern.
10978 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10980 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10981 (UNSPEC_UMUL_HIGHPART): New constants.
10982 (MUL_HIGHPART): New int iteraor.
10983 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10984 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10986 (*<su>mul<mode>3_highpart): New define_insn.
10988 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10991 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10994 2018-03-13 Martin Liska <mliska@suse.cz>
10997 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10999 (sem_item_optimizer::~sem_item_optimizer): Release it.
11000 (sem_item_optimizer::merge_classes): Register variable aliases.
11001 (sem_item_optimizer::fixup_pt_set): New function.
11002 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11003 * ipa-icf.h: Declare new variables and functions.
11005 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11007 PR middle-end/84834
11008 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11009 integer_pow2p@2 and test integer_pow2p in condition.
11010 (A < 0 ? C : 0): Similarly for @1.
11012 PR middle-end/84831
11013 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11014 characters starting at p contain '\0' character, don't look beyond
11018 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11019 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11022 * reg-stack.c (change_stack): Change update_end var from int to
11023 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11024 also call set_block_for_insn on the newly added insns and rescan.
11027 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11028 on the last operand.
11031 * tree.c (stabilize_reference_1): Return save_expr (e) for
11032 STATEMENT_LIST even if it doesn't have side-effects.
11034 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11036 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11038 2018-03-12 Renlin Li <renlin.li@arm.com>
11040 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11041 aarch64_output_scalar_simd_mov_immediate.
11043 2018-03-12 Martin Sebor <msebor@redhat.com>
11045 PR tree-optimization/83456
11046 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11047 for perfectly overlapping calls to memcpy.
11048 (gimple_fold_builtin_memory_chk): Same.
11049 (gimple_fold_builtin_strcpy): Handle no-warning.
11050 (gimple_fold_builtin_stxcpy_chk): Same.
11051 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11053 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11055 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11056 parameter. Use it for SFmode.
11057 (rs6000_function_arg_advance_1): Adjust.
11058 (rs6000_function_arg): Adjust.
11059 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11061 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11063 PR rtl-optimization/84169
11064 PR rtl-optimization/84780
11065 * combine.c (can_combine_p): Check for a 2-insn combination whether
11066 the destination register is used between the two insns, too.
11068 2018-03-12 Richard Biener <rguenther@suse.de>
11070 PR tree-optimization/84803
11071 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11072 for refs DR analysis didn't process.
11074 2018-03-12 Richard Biener <rguenther@suse.de>
11076 PR tree-optimization/84777
11077 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11078 force-vectorize loops ignore whether we are optimizing for size.
11080 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11082 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11083 (TARGET_MD_ASM_ADJUST): Define.
11085 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11086 Kito Cheng <kito.cheng@gmail.com>
11087 Chung-Ju Wu <jasonwucj@gmail.com>
11089 * config/nds32/nds32.c (nds32_compute_stack_frame,
11090 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11091 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11092 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11093 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11094 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11095 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11096 * config/nds32/nds32.md (prologue, epilogue): Use macro
11097 NDS32_V3PUSH_AVAILABLE_P to do checking.
11099 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11102 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11103 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11104 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11105 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11106 addition of most attributes on !orig_type_die or the attribute not
11107 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11109 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11110 Chung-Ju Wu <jasonwucj@gmail.com>
11112 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11113 __NDS32_VH__ macro.
11114 * config/nds32/nds32.opt (mvh): New option.
11116 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11117 Chung-Ju Wu <jasonwucj@gmail.com>
11119 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11121 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11122 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11125 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11126 Chung-Ju Wu <jasonwucj@gmail.com>
11128 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11130 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11131 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11133 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11134 Kito Cheng <kito.cheng@gmail.com>
11135 Chung-Ju Wu <jasonwucj@gmail.com>
11137 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11138 UNSPEC_FFMISM and UNSPEC_FLMISM.
11139 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11140 for ffb, ffmism and flmism.
11141 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11142 (unspec_ffmism): Ditto.
11143 (unspec_flmism): Ditto.
11144 (nds32_expand_builtin_impl): Check if string extension is available.
11145 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11146 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11148 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11151 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11154 * lra-assigns.c (assign_by_spills): Return a flag of reload
11155 assignment failure. Do not process the reload assignment
11156 failures. Do not spill other reload pseudos if they has the same
11158 (lra_assign): Add a return arg. Set up from the result of
11159 assign_by_spills call.
11160 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11161 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11162 usage_insns if it is not NULL.
11163 (spill_hard_reg_in_range): New function.
11164 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11165 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11166 function prototypes.
11167 (lra_assign): Change prototype.
11168 * lra.c (lra): Add code to deal with fails by splitting hard reg
11171 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11174 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11176 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11179 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11180 (dw_val_node): Add val_symbolic_view.
11181 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11182 (symview_upper_bound): New.
11183 (new_line_info_table): Initialize symviews_since_reset.
11184 (dwarf2out_source_line): Count symviews_since_reset and set
11185 symview_upper_bound.
11186 (dw_val_equal_p): Handle symview.
11187 (add_AT_symview): New.
11188 (print_dw_val): Handle symview.
11189 (attr_checksum, attr_checksum_ordered): Likewise.
11190 (same_dw_val_p, size_of_die): Likewise.
11191 (value_format, output_die): Likewise.
11192 (add_high_low_attributes): Use add_AT_symview for entry_view.
11193 (dwarf2out_finish): Reset symview_upper_bound, clear
11196 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11199 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11200 Add strict argument and use it.
11201 (rs6000_split_multireg_move): Update for new strict argument.
11202 (mem_operand_gpr): Disallow all non-offsettable addresses.
11203 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11205 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11208 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11209 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11210 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11213 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11214 decl, use remap_type if we want to use the type.
11216 2018-03-09 Martin Sebor <msebor@redhat.com>
11218 PR tree-optimization/84526
11219 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11221 (builtin_access::generic_overlap): Be prepared to handle non-array
11224 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11226 PR rtl-optimization/84682
11227 * lra-constraints.c (process_address_1): Check is_address flag
11228 for address constraints.
11229 (process_alt_operands): Likewise.
11230 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11231 preprocess_constraints.
11232 * recog.h (preprocess_constraints): Add oploc parameter.
11234 * recog.c (preprocess_constraints): Test address_operand for
11235 CT_ADDRESS constraints.
11237 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11240 * lra-assigns.c (assign_by_spills): Return a flag of reload
11241 assignment failure. Do not process the reload assignment
11242 failures. Do not spill other reload pseudos if they has the same
11244 (lra_assign): Add a return arg. Set up from the result of
11245 assign_by_spills call.
11246 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11247 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11248 usage_insns if it is not NULL.
11249 (spill_hard_reg_in_range): New function.
11250 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11251 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11252 function prototypes.
11253 (lra_assign): Change prototype.
11254 * lra.c (lra): Add code to deal with fails by splitting hard reg
11257 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11260 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11261 Accept complain bool parameter. Only emit errors if it is true.
11262 (arm_parse_cpu_option_name): Likewise.
11263 (arm_target_thumb_only): Adjust callers of the above.
11264 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11265 prototype to take a default true bool parameter.
11266 (arm_parse_arch_option_name): Likewise.
11268 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11269 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11273 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11274 * configure: Regenerate.
11275 * configure.ac ("linker --version-script option"): New.
11276 ("linker soname option"): New.
11278 2018-03-09 Richard Biener <rguenther@suse.de>
11280 PR tree-optimization/84775
11281 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11282 immediate uses of predicate stmts and mark them modified.
11285 PR tree-optimization/84178
11286 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11288 (version_loop_for_if_conversion): Delay update_ssa call.
11289 (tree_if_conversion): Delay update_ssa until after predicate
11292 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11295 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11296 when the function accesses prior frames.
11298 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11301 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11302 gen_llsym, otherwise call maybe_gen_llsym.
11304 PR inline-asm/84742
11305 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11306 has ',' character inside of it.
11308 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11311 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11312 as clobbering CC_REGNUM.
11314 2018-03-08 Richard Biener <rguenther@suse.de>
11316 PR middle-end/84552
11317 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11318 (follow_copies_to_constant): Do not follow SSA names registered
11321 2018-03-08 Richard Biener <rguenther@suse.de>
11323 PR tree-optimization/84178
11324 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11326 (version_loop_for_if_conversion): Delay update_ssa call.
11327 (tree_if_conversion): Delay update_ssa until after predicate
11330 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11332 PR tree-optimization/84178
11333 * tree-if-conv.c (release_bb_predicate): Remove the
11334 the assertion that the stmts have NULL use_ops.
11335 Discard the statements, asserting that they haven't
11336 yet been added to a BB.
11338 2018-03-08 Richard Biener <rguenther@suse.de>
11340 PR tree-optimization/84746
11341 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11342 (phi_translate): Pass in destination ANTIC_OUT set.
11343 (phi_translate_1): Likewise. For a simplified result lookup
11344 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11345 (phi_translate_set): Adjust.
11346 (do_pre_regular_insertion): Likewise.
11347 (do_pre_partial_partial_insertion): Likewise.
11349 2018-03-08 Martin Liska <mliska@suse.cz>
11351 PR gcov-profile/84735
11352 * doc/gcov.texi: Document usage of profile files.
11353 * gcov-io.h: Document changes in the format.
11355 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11359 * dwarf2out.c (struct dw_line_info_table): Update comments for
11361 (FORCE_RESET_NEXT_VIEW): New.
11362 (FORCE_RESETTING_VIEW_P): New.
11363 (RESETTING_VIEW_P): Check for -1 too.
11364 (ZERO_VIEW_P): Likewise.
11365 (new_line_info_table): Force-reset next view.
11366 (dwarf2out_begin_function): Likewise.
11367 (dwarf2out_source_line): Simplify zero_view_p initialization.
11368 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11369 view directly. Omit view when omitting .loc at line 0.
11371 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11373 PR tree-optimization/84740
11374 * tree-switch-conversion.c (process_switch): Call build_constructors
11375 only if info.phi_count is non-zero.
11377 PR tree-optimization/84739
11378 * tree-tailcall.c (find_tail_calls): Check call arguments against
11379 DECL_ARGUMENTS (current_function_decl) rather than
11380 DECL_ARGUMENTS (func) when checking for tail recursion.
11382 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11384 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11385 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11386 Volker Reichelt's entry and add entries for people that perform
11387 GCC fuzzy testing and report numerous bugs.
11389 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11392 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11393 readonly data in sdata, if that is disabled.
11394 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11395 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11396 -mreadonly-in-sdata option.
11398 2018-03-07 Martin Sebor <msebor@redhat.com>
11400 PR tree-optimization/84468
11401 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11402 basic block when looking for nul assignment.
11404 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11407 * except.h (output_function_exception_table): Adjust prototype.
11408 * except.c (output_function_exception_table): Remove FNNAME parameter
11409 and add SECTION parameter. Ouput one part of the table at a time.
11410 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11411 the first part of the exception table and emit unwind directives.
11412 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11413 (i386_pe_seh_cold_init): Likewise.
11414 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11415 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11416 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11417 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11418 * config/i386/winnt.c: Include except.h.
11419 (struct seh_frame_state): Add reg_offset, after_prologue and
11420 in_cold_section fields.
11421 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11422 (i386_pe_seh_cold_init): New function.
11423 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11424 to seh->in_cold_section.
11425 (seh_emit_push): Record the offset of the push.
11426 (seh_emit_save): Record the offet of the save.
11427 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11428 Test seh->after_prologue to disregard the epilogue.
11429 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11430 (i386_pe_end_cold_function): New function.
11432 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11435 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11436 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11439 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11441 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11442 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11444 PR middle-end/84723
11445 * multiple_target.c: Include tree-inline.h and intl.h.
11446 (expand_target_clones): Diagnose and fail if node->definition and
11447 !tree_versionable_function_p (node->decl).
11449 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11451 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11453 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11454 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11455 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11457 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11460 * combine.c (try_combine): Use reg_or_subregno instead of handling
11461 just paradoxical SUBREGs and REGs.
11463 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11465 * config/arc/arc.c (arc_finalize_pic): Remove function.
11466 (arc_must_save_register): We use single base PIC register, remove
11467 checks to save/restore the PIC register.
11468 (arc_expand_prologue): Likewise.
11469 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11471 (arc_verify_short): Likewise.
11472 (arc_attr_type): Likewise.
11473 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11474 (walk_stores): Likewise.
11475 (arc_address_cost): Make it static.
11476 (arc_verify_short): Likewise.
11477 (branch_dest): Likewise.
11478 (arc_attr_type): Likewise.
11479 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11480 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11481 (arc_final_prescan_insn): Remove inserting the nops due to
11482 hardware hazards. It is done in reorg step.
11483 (insn_length_variant_t): Remove.
11484 (insn_length_parameters_t): Likewise.
11485 (arc_insn_length_parameters): Likewise.
11486 (arc_get_insn_variants): Likewise.
11487 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11489 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11491 PR inline-asm/84683
11492 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11495 PR tree-optimization/84687
11496 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11498 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11500 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11502 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11503 Rename to ppc_speculation_barrier.
11504 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11505 __builtin_ppc_speculation_barrier.
11507 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11510 * combine.c (combine_simplify_rtx): Don't try to simplify if
11511 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11514 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11516 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11517 to 32 bytes when compiling for POWER9.
11519 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11522 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11523 regparm >= 3 with no arg reg available also for calls with
11524 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11527 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11529 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11531 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11534 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11536 2018-03-05 Richard Biener <rguenther@suse.de>
11538 PR tree-optimization/84486
11539 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11540 When inserting a __builtin_assume_aligned call set the LHS
11541 SSA name alignment info accordingly.
11543 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11545 PR tree-optimization/84114
11546 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11547 Avoid reassociation of FLOAT_MODE addition.
11549 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11551 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11552 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11553 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11554 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11555 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11556 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11557 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11559 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11560 __builtin_ia32_wbinvd): New builtins.
11561 (SPECIAL_ARGS2): New.
11562 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11563 (SPECIAL_ARGS2): New.
11564 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11565 (ix86_valid_target_attribute_inner_p): Ditto.
11566 (ix86_init_mmx_sse_builtins): Add special_args2.
11567 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11568 TARGET_WBNOINVD_P): New.
11569 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11570 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11571 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11572 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11573 * config/i386/pconfigintrin.h: New file.
11574 * config/i386/wbnoinvdintrin.h: Ditto.
11575 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11577 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11579 2018-03-05 Richard Biener <rguenther@suse.de>
11581 PR tree-optimization/84670
11582 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11584 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11585 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11586 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11587 doesn't grow if all successors (recursively) were visited at least
11590 2018-03-05 Richard Biener <rguenther@suse.de>
11592 PR tree-optimization/84650
11593 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11594 if executed in the loop pipeline.
11596 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11598 * doc/configfiles.texi (Configuration Files): Move info about
11599 conditionalizing $target-protos.h to...
11600 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11601 differs from $target-protos.h.
11603 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11604 Chung-Ju Wu <jasonwucj@gmail.com>
11606 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11607 * config/nds32/nds32-multiple.md (setmemsi): Define.
11608 * config/nds32/nds32-memory-manipulation.c
11609 (nds32_gen_dup_4_byte_to_word_value): New.
11610 (emit_setmem_word_loop): New.
11611 (emit_setmem_byte_loop): New.
11612 (nds32_expand_setmem_loop): New.
11613 (nds32_expand_setmem_loop_v3m): New.
11614 (nds32_expand_setmem_unroll): New.
11615 (nds32_expand_setmem): New.
11617 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11618 Chung-Ju Wu <jasonwucj@gmail.com>
11620 * config/nds32/nds32-memory-manipulation.c
11621 (nds32_emit_load_store): New.
11622 (nds32_emit_post_inc_load_store): New.
11623 (nds32_emit_mem_move): New.
11624 (nds32_emit_mem_move_block): New.
11625 (nds32_expand_movmemsi_loop_unknown_size): New.
11626 (nds32_expand_movmemsi_loop_known_size): New.
11627 (nds32_expand_movmemsi_loop): New.
11628 (nds32_expand_movmemsi_unroll): New.
11629 (nds32_expand_movmemqi): Rename ...
11630 (nds32_expand_movmemsi): ... to this.
11631 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11632 (movmemsi): ... to this.
11633 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11634 (nds32_expand_movmemsi): ... to this.
11636 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11637 Monk Chiang <sh.chiang04@gmail.com>
11638 Chung-Ju Wu <jasonwucj@gmail.com>
11640 * config/nds32/nds32-protos.h
11641 (nds32_expand_load_multiple): New arguments.
11642 (nds32_expand_store_multiple): Ditto.
11643 (nds32_valid_multiple_load_store): Rename ...
11644 (nds32_valid_multiple_load_store_p): ... to this.
11645 * config/nds32/nds32-memory-manipulation.c
11646 (nds32_expand_load_multiple): Refine implementation.
11647 (nds32_expand_store_multiple): Ditto.
11648 * config/nds32/nds32-multiple.md
11649 (load_multiple): Update nds32_expand_load_multiple interface.
11650 (store_multiple): Update nds32_expand_store_multiple interface.
11651 * config/nds32/nds32-predicates.c
11652 (nds32_valid_multiple_load_store): Rename ...
11653 (nds32_valid_multiple_load_store_p): ... to this and refine
11655 * config/nds32/predicates.md
11656 (nds32_load_multiple_and_update_address_operation): New predicate.
11657 (nds32_store_multiple_and_update_address_operation): New predicate.
11659 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11660 Chung-Ju Wu <jasonwucj@gmail.com>
11662 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11663 (combo): New attribute.
11664 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11666 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11668 * config/nds32/nds32.opt: Change -mcmodel= default value.
11670 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11671 Monk Chiang <sh.chiang04@gmail.com>
11672 Chung-Ju Wu <jasonwucj@gmail.com>
11674 * config/nds32/constants.md (unspec_element): New enum.
11675 * config/nds32/constraints.md (Umw): New constraint.
11676 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11677 * config/nds32/nds32-intrinsic.md: Likewise.
11678 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11679 (nds32_valid_smw_lwm_base_p): New.
11680 (nds32_output_smw_single_word): New.
11681 (nds32_output_lmw_single_word): New.
11682 (nds32_expand_unaligned_load): New.
11683 (nds32_expand_unaligned_store): New.
11684 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11685 (nds32_output_smw_single_word): Declare.
11686 (nds32_output_lmw_single_word): Declare.
11687 (nds32_expand_unaligned_load): Declare.
11688 (nds32_expand_unaligned_store): Declare.
11689 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11690 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11691 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11692 NDS32_BUILTIN_UASTORE_DW.
11693 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11696 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11697 Kito Cheng <kito.cheng@gmail.com>
11698 Chung-Ju Wu <jasonwucj@gmail.com>
11700 * config/nds32/nds32-intrinsic.c
11701 (nds32_expand_builtin_null_ftype_reg): Delete.
11702 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11703 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11704 (nds32_read_argument): New.
11705 (nds32_legitimize_target): Ditto.
11706 (nds32_legitimize_argument): Ditto.
11707 (nds32_check_constant_argument): Ditto.
11708 (nds32_expand_unop_builtin): Ditto.
11709 (nds32_expand_unopimm_builtin): Ditto.
11710 (nds32_expand_binop_builtin): Ditto.
11711 (nds32_builtin_decl_impl): Ditto.
11712 (builtin_description): Ditto.
11713 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11714 (nds32_init_builtins_impl): Ditto.
11715 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11716 (nds32_builtin_decl): New.
11717 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11718 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11720 2018-03-02 Jeff Law <law@redhat.com>
11722 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11723 (redundant_insn, fill_simple_delay_slots): Likewise.
11724 (fill_slots_from_thread): Likewise.
11725 * resource.c (mark_referenced_resources): Likewise.
11726 (mark_set_resources, find_dead_or_set_registers): Likewise.
11728 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11730 * substring-locations.h (format_warning_va): Formatting fix for
11731 ATTRIBUTE_GCC_DIAG.
11732 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11734 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11735 * substring-locations.c: Include intl.h.
11736 (format_warning_va): Turned into small wrapper around
11737 format_warning_n_va, renamed to ...
11738 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11739 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11741 (format_warning_at_substring_n): New function.
11742 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11743 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11744 format_warning_at_substring with just a shorter name instead of
11745 const function pointer.
11746 (fmtwarn_n): New function.
11747 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11748 appropriate, get rid of all the fmtstr temporaries, move conditionals
11749 with G_() wrapped string literals directly into fmtwarn arguments,
11750 cast dir.len to (int), formatting fixes.
11752 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11754 * doc/invoke.texi: Remove "Cilk Plus" references.
11756 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11757 Richard Biener <rguenther@suse.de>
11760 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11761 for error or warning attributes if CALL_FROM_THUNK_P is set.
11764 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11767 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11768 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11771 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11772 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11774 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11775 instead of -1U in last predictors element's probability member.
11777 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11780 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11781 arguments if they are comparable.
11783 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11785 PR tree-optimization/84634
11786 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11787 masks and masked_loop_p with a single loop_masks, making sure it's
11788 null for bb vectorization.
11790 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11792 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11793 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11794 loop->force_vectorize to check whether there is no alias.
11796 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11799 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11801 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11803 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11804 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11805 instead of a loop around prev_real_insn.
11806 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11809 PR inline-asm/84625
11810 * config/i386/i386.c (ix86_print_operand): Use conditional
11811 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11814 2018-03-02 Richard Biener <rguenther@suse.de>
11816 PR tree-optimization/84427
11817 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11818 (bitmap_set_subtract_values): Rewrite to handle multiple
11821 (prune_clobbered_mems): Likewise.
11822 (phi_translate): Take edge instead of pred/phiblock.
11823 (phi_translate_1): Likewise.
11824 (phi_translate_set): Likewise. Insert all translated
11825 exprs for a value into the set, keeping possibly multiple
11826 expressions per value.
11827 (compute_antic_aux): Adjust for phi_translate changes.
11828 When intersecting union the expressions and prune those
11829 not in the final value set, keeping possibly multiple
11830 expressions per value. Do not use value-insertion
11831 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11832 all expressions. Add verification that the value-sets
11833 only shrink during iteration.
11834 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11835 (do_pre_regular_insertion): Likewise.
11836 (do_pre_partial_partial_insertion): Likewise.
11838 2018-03-02 Richard Biener <rguenther@suse.de>
11841 * config/darwin.c (saved_debug_info_level): New static global.
11842 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11843 (darwin_asm_lto_end): Restore debug info generation settings.
11845 2018-03-01 Martin Liska <mliska@suse.cz>
11848 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11849 volatile arguments.
11851 2018-03-01 Richard Biener <rguenther@suse.de>
11854 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11855 type annotation with LTO when debug was disabled at compile-time.
11857 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11859 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11861 (mips_final_postscan_insn): Likewise.
11863 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11865 PR rtl-optimization/84528
11866 * alias.c (init_alias_target): Add commentary.
11867 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11868 a unique base value if the frame pointer is not eliminated
11869 to the stack pointer.
11871 2018-03-01 Tom de Vries <tom@codesourcery.com>
11873 PR rtl-optimization/83327
11874 * lra-int.h (hard_regs_spilled_into): Declare.
11875 * lra.c (hard_regs_spilled_into): Define.
11876 (init_reg_info): Init hard_regs_spilled_into.
11877 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11878 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11879 (process_bb_lives): Handle hard_regs_spilled_into.
11880 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11881 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11883 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11885 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11886 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11887 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11888 * config/rs6000/aix72.h: New file.
11890 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11892 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11893 instead of warning_at with conditional singular and plural messages
11897 * stor-layout.c (update_alignment_for_field): For
11898 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11899 && !DECL_PACKED (field), do the alignment update, just use
11900 only desired_align instead of MAX (type_align, desired_align)
11902 (place_field): Don't do known_align < desired_align handling
11903 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11904 is non-NULL, instead do it after rli->prev_field handling and
11905 only if not within a bitfield word. For DECL_PACKED (field)
11906 use type_align of BITS_PER_UNIT.
11908 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11910 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11911 superfluous parentheses and trailing spaces.
11913 2018-02-28 Richard Biener <rguenther@suse.de>
11915 PR tree-optimization/84584
11916 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11917 SCoPs with fake exit edge.
11919 2018-02-28 Martin Liska <mliska@suse.cz>
11922 * timevar.c (timer::print): Fix format to properly print 100%
11925 2018-02-28 Richard Biener <rguenther@suse.de>
11927 PR middle-end/84607
11928 * genmatch.c (capture_info::walk_match): Do not mark
11929 captured expressions without operands as expr_p given
11930 they act more like predicates and should be subject to
11931 "lost tail" side-effect preserving.
11933 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11935 PR rtl-optimization/81611
11936 * auto-inc-dec.c (attempt_change): Move dead note from
11937 mem_insn if it's the next use of regno
11938 (find_address): Take address use of reg holding
11939 non-incremented value. Add parm to limit search to the named
11941 (merge_in_block): Attempt to use a mem insn that is the next
11942 use of the original regno.
11944 2018-02-27 Martin Sebor <msebor@redhat.com>
11947 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11948 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11950 2018-02-27 Martin Sebor <msebor@redhat.com>
11952 PR translation/84207
11953 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11954 n argument to unsigned HOST_WIDE_INT.
11955 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11956 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11957 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11958 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11960 2018-02-27 Richard Biener <rguenther@suse.de>
11962 PR tree-optimization/84512
11963 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11964 Do not use the estimate returned from record_stmt_cost for
11965 the scalar iteration cost but sum properly using add_stmt_cost.
11967 2018-02-27 Richard Biener <rguenther@suse.de>
11969 PR tree-optimization/84466
11970 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11971 Adjust last change to less strictly validate use operands.
11973 2018-02-27 Martin Liska <mliska@suse.cz>
11975 PR gcov-profile/84548
11976 * gcov.c (process_file): Allow partial overlap and consider it
11977 also as group functions.
11978 (output_lines): Properly calculate range of lines for a group.
11980 2018-02-27 Martin Liska <mliska@suse.cz>
11982 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11983 'ggc' suffixes. Change first column width.
11984 (timer::print): Fix formatting of the column.
11986 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11988 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11989 preserve inline entry blocks for the sake of debug inline
11990 entry point markers alone.
11991 (remove_unused_locals): Suggest in comments a better place to
11992 force the preservation of inline entry blocks that are
11993 otherwise unused, but do not preserve them.
11995 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11997 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11999 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12002 * config/i386/constraints.md (Bs): Replace
12003 ix86_indirect_branch_register with
12004 TARGET_INDIRECT_BRANCH_REGISTER.
12006 * config/i386/i386.md (indirect_jump): Likewise.
12007 (tablejump): Likewise.
12008 (*sibcall_memory): Likewise.
12009 (*sibcall_value_memory): Likewise.
12010 Peepholes of indirect call and jump via memory: Likewise.
12011 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12012 (*sibcall_value_GOT_32): Likewise.
12013 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12014 (GOT_memory_operand): Likewise.
12015 (call_insn_operand): Likewise.
12016 (sibcall_insn_operand): Likewise.
12017 (GOT32_symbol_operand): Likewise.
12018 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12020 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12022 PR rtl-optimization/83496
12023 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12024 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12025 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12026 redundant insn, if any.
12027 (relax_delay_slots): Likewise.
12028 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12030 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12032 PR tree-optimization/83965
12033 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12034 that grouped statements are part of a reduction chain. Return
12035 true if the statement is not marked as a reduction itself but
12036 is part of a group.
12037 (vect_recog_dot_prod_pattern): Don't check whether the statement
12038 is part of a group here.
12039 (vect_recog_sad_pattern): Likewise.
12040 (vect_recog_widen_sum_pattern): Likewise.
12042 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12045 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12047 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12050 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12052 (ix86_output_indirect_function_return): New prototype.
12053 (ix86_split_simple_return_pop_internal): Likewise.
12054 * config/i386/i386.c (indirect_return_via_cx): New.
12055 (indirect_return_via_cx_bnd): Likewise.
12056 (indirect_thunk_name): Handle return va CX_REG.
12057 (output_indirect_thunk_function): Create alias for
12058 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12059 (ix86_output_indirect_jmp): Remove the bool argument.
12060 (ix86_output_indirect_function_return): New function.
12061 (ix86_split_simple_return_pop_internal): Likewise.
12062 * config/i386/i386.md (*indirect_jump): Don't pass false
12063 to ix86_output_indirect_jmp.
12064 (*tablejump_1): Likewise.
12065 (simple_return_pop_internal): Change it to define_insn_and_split.
12066 Call ix86_split_simple_return_pop_internal to split it for
12067 -mfunction-return=.
12068 (simple_return_indirect_internal): Call
12069 ix86_output_indirect_function_return instead of
12070 ix86_output_indirect_jmp.
12072 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12075 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12076 memset and value initialization afterwards.
12078 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12080 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12082 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12085 * common/config/aarch64/aarch64-common.c
12086 (aarch_option_optimization_table[]): Switch
12087 off fomit-frame-pointer
12089 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12090 Chung-Ju Wu <jasonwucj@gmail.com>
12092 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12094 (store_multiple): Ditto.
12096 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12098 * config.gcc: Add --with-cpu support for nds32 target.
12099 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12100 * config/nds32/nds32.opt: Add -mcpu= option.
12102 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12104 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12105 isel=yes): Warn for these deprecated options.
12107 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12109 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12110 ISA_2_5_MASKS_EMBEDDED.
12112 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12114 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12115 p->max as pointers rather than using iterative_hash_expr.
12117 2018-02-23 Carl Love <cel@us.ibm.com>
12119 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12120 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12122 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12123 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12124 P8V_BUILTIN_VEC_VUNSIGNED2.
12126 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12129 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12130 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12132 * lra-constraints.c (curr_insn_transform): Set up
12133 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12135 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12136 * lra-eliminations.c (spill_pseudos): Ditto.
12137 (process_insn_for_elimination): Ditto.
12138 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12139 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12140 LRA_NON_CLOBBERED_ALT.
12142 2018-02-22 Martin Sebor <msebor@redhat.com>
12144 PR tree-optimization/84480
12145 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12146 to maybe_diag_stxncpy_trunc. Call it.
12147 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12148 from gimple_fold_builtin_strcpy. Print inlining stack.
12149 (handle_builtin_stxncpy): Print inlining stack.
12150 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12152 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12155 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12156 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12157 and -fcheck-pointer-bounds are used together.
12158 (indirect_thunk_prefix): New enum.
12159 (indirect_thunk_need_prefix): New function.
12160 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12161 "_nt" instead of "_bnd" for NOTRACK prefix.
12162 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12163 (output_indirect_thunk_function): Likewise.
12165 (ix86_code_end): Update output_indirect_thunk_function calls.
12166 (ix86_output_indirect_branch_via_reg): Replace
12167 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12168 (ix86_output_indirect_branch_via_push): Likewise.
12169 (ix86_output_function_return): Likewise.
12170 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12171 incompatible with -fcf-protection=branch and
12172 -fcheck-pointer-bounds.
12174 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12177 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12178 Change gcc_assert call to output_operand_lossage.
12180 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12182 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12184 2018-02-22 DJ Delorie <dj@redhat.com>
12185 Sebastian Perta <sebastian.perta@renesas.com>
12186 Oleg Endo <olegendo@gcc.gnu.org>
12188 * config/rx/rx.c (rx_rtx_costs): New function.
12189 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12191 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12193 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12195 2018-02-22 Martin Liska <mliska@suse.cz>
12198 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12199 Add "native" as a possible value.
12201 2018-02-22 Martin Liska <mliska@suse.cz>
12204 * config/i386/i386.c (ix86_option_override_internal):
12205 Add "native" as a possible value for -march and -mtune.
12207 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12210 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12211 to all type variants.
12213 PR tree-optimization/84503
12214 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12215 width as info->bitpos + info->bitsize - start.
12216 (merged_store_group::merge_overlapping): Simplify width computation.
12217 (check_no_overlap): New function.
12218 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12219 start + width and last_order of the group, fail if check_no_overlap
12221 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12222 to group if check_no_overlap fails.
12224 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12226 * config/rs6000/altivec.md: Delete contraint arguments to
12227 define_expand, define_split, and define_peephole2, and in
12228 define_insn_and_split if always unused.
12229 * config/rs6000/darwin.md: Ditto.
12230 * config/rs6000/dfp.md: Ditto.
12231 * config/rs6000/rs6000.md: Ditto.
12232 * config/rs6000/sync.md: Ditto.
12233 * config/rs6000/vector.md: Ditto.
12234 * config/rs6000/vsx.md: Ditto.
12236 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12238 * config/rs6000/altivec.md: Write output control strings as braced
12239 blocks instead of double-quoted strings.
12240 * config/rs6000/darwin.md: Ditto.
12241 * config/rs6000/rs6000.md: Ditto.
12242 * config/rs6000/vector.md: Ditto.
12243 * config/rs6000/vsx.md: Ditto.
12245 2018-02-21 Jason Merrill <jason@redhat.com>
12247 PR c++/84314 - ICE with templates and fastcall attribute.
12248 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12250 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12252 * ipa-cp.c (determine_versionability): Fix comment typos.
12254 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12257 * ipa-cp.c (determine_versionability): Do not version functions caling
12260 2018-02-21 Martin Liska <mliska@suse.cz>
12263 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12264 Add "native" as a possible value.
12265 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12266 the macro when native cpu detection is available.
12268 2018-02-21 Martin Liska <mliska@suse.cz>
12271 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12272 Add "native" as a possible value.
12273 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12274 when native cpu detection is available.
12276 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12277 Martin Sebor <msebor@redhat.com>
12279 PR tree-optimization/84478
12280 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12282 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12283 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12284 support which is conservatively correct, for 2 only stay conservative
12285 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12286 argument to the 2 argument get_range_strlen, adjust 6 arg
12287 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12289 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12290 (gimple_fold_builtin_strlen): Pass true as last argument to
12293 2018-02-20 Martin Sebor <msebor@redhat.com>
12295 PR middle-end/84095
12296 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12297 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12298 (builtin_memref::builtin_memref): Factor out parts into
12299 set_base_and_offset and call it.
12301 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12303 PR middle-end/84406
12304 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12305 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12306 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12307 search at the associated MODE_INT.
12309 2018-02-20 Jeff Law <law@redhat.com>
12311 PR middle-end/82123
12312 PR tree-optimization/81592
12313 PR middle-end/79257
12314 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12315 for range data rather than using global data.
12316 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12317 range data rather than using global data.
12318 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12319 pass it to children as needed.
12320 (struct directive::fmtresult): Similarly.
12321 (struct directive::set_width): Similarly.
12322 (struct directive::set_precision): Similarly.
12323 (format_integer, format_directive, parse_directive): Similarly.
12324 (format_none): Accept unnamed vr_values parameter.
12325 (format_percent, format_floating, format_character): Similarly.
12326 (format_string, format_plain): Similarly.
12327 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12328 the EVRP range analyzer for range data rather than using global data.
12329 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12330 gimple-ssa-evrp-analyze.h
12331 (class sprintf_dom_walker): Add after_dom_children member function.
12332 Add evrp_range_analyzer member.
12333 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12334 range analyzer as needed.
12335 (sprintf_dom_walker::after_dom_children): New member function.
12336 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12338 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12339 (evrp_range_analyzer::pop_to_marker): Likewise.
12341 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12343 PR tree-optimization/84419
12344 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12345 with the required type if its current type is compatible but
12348 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12350 PR middle-end/82004
12351 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12352 after vectorization.
12354 2018-02-20 Martin Liska <mliska@suse.cz>
12357 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12358 possible values if we don't have a hint.
12360 2018-02-20 Martin Liska <mliska@suse.cz>
12364 * final.c (shorten_branches): Build align_tab array with one
12366 * opts.c (finish_options): Add alignment option limit check.
12367 (MAX_CODE_ALIGN): Likewise.
12368 (MAX_CODE_ALIGN_VALUE): Likewise.
12369 * doc/invoke.texi: Document maximum allowed option value for
12370 all -falign-* options.
12372 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12375 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12376 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12377 * var-tracking.c (emit_note_insn_var_location): Remove all references
12378 to NOTE_INSN_CALL_ARG_LOCATION.
12379 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12380 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12381 Use copy_rtx_if_shared.
12382 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12383 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12384 (dwarf2out_var_location): Remove handling of
12385 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12387 * final.c (final_scan_insn): Remove all references to
12388 NOTE_INSN_CALL_ARG_LOCATION.
12389 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12390 before dumping final insns.
12391 * except.c (emit_note_eh_region_end): Remove all references to
12392 NOTE_INSN_CALL_ARG_LOCATION.
12393 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12394 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12395 * config/arc/arc.c (hwloop_optimize): Likewise.
12396 * config/arm/arm.c (create_fix_barrier): Likewise.
12397 * config/s390/s390.c (s390_chunkify_start): Likewise.
12398 * config/sh/sh.c (find_barrier): Likewise.
12399 * config/i386/i386.c (rest_of_insert_endbranch,
12400 ix86_seh_fixup_eh_fallthru): Likewise.
12401 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12402 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12403 * config/frv/frv.c (frv_function_prologue): Likewise.
12404 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12406 (note_outside_basic_block_p): Remove all references to
12407 NOTE_INSN_CALL_ARG_LOCATION.
12408 * gengtype.c (adjust_field_rtx_def): Likewise.
12409 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12411 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12412 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12415 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12418 PR tree-optimization/84452
12419 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12420 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12423 2018-02-19 Martin Liska <mliska@suse.cz>
12426 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12428 2018-02-19 Martin Liska <mliska@suse.cz>
12429 Richard Sandiford <richard.sandiford@linaro.org>
12431 PR tree-optimization/82491
12432 * gimple-fold.c (get_base_constructor): Make earlier bail out
12435 2018-02-19 Carl Love <cel@us.ibm.com>
12437 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12438 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12440 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12441 P8V_BUILTIN_VEC_NEG.
12443 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12445 * config/rl78/rl78.md (movdf): New define expand.
12447 2018-02-19 Martin Liska <mliska@suse.cz>
12450 * doc/invoke.texi: Fix typo.
12451 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12453 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12455 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12456 handle rs6000_single_float and rs6000_double_float specially for
12459 2018-02-16 Jeff Law <law@redhat.com>
12461 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12462 (pop_regs): Use it.
12464 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12467 * ipa-inline.c (inline_small_functions): Fix a typo.
12469 2018-02-16 Nathan Sidwell <nathan@acm.org>
12471 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12473 2018-02-16 Carl Love <cel@us.ibm.com>
12475 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12476 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12477 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12478 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12479 expansion to P8V_BUILTIN_VEC_FLOAT2.
12481 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12483 PR rtl-optimization/70023
12484 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12485 src_regno into account.
12487 2018-02-16 Carl Love <cel@us.ibm.com>
12489 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12490 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12491 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12492 * config/rs6000/rs6000.c: Remove case statements for
12493 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12494 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12495 and P9V_BUILTIN_VEC_VINSERT4B.
12496 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12497 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12498 * config/rs6000/vsx.md:
12499 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12502 2018-02-16 Carl Love <cel@us.ibm.com>
12504 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12506 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12508 * config/rs6000/rs6000-c.c: Add the definitions for
12509 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12510 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12511 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12512 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12513 definition for insert4b and define insn *insert3b_internal.
12514 * doc/extend.texi: Add documentation for vec_extract4b.
12516 2018-02-16 Nathan Sidwell <nathan@acm.org>
12518 * doc/extend.texi (Backwards Compatibility): Mention friend
12519 injection. Note for-scope is deprecated.
12520 * doc/invoke.texi (-ffriend-injection): Deprecate.
12522 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12524 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12525 that moved to I2, also allow destinations that are a paradoxical
12526 subreg (instead of a normal reg).
12528 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12531 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12534 2018-02-16 Richard Biener <rguenther@suse.de>
12536 PR tree-optimization/84037
12537 PR tree-optimization/84016
12539 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12540 Adjust vec_construct for the fact we need additional higher latency
12541 128bit inserts for AVX256 and AVX512 vector builds.
12542 (ix86_add_stmt_cost): Scale vector construction cost for
12545 2018-02-16 Richard Biener <rguenther@suse.de>
12547 PR tree-optimization/84417
12548 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12549 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12550 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12552 2018-02-16 Martin Liska <mliska@suse.cz>
12555 * internal-fn.def (ASAN_CHECK): Set proper flags.
12556 (ASAN_MARK): Likewise.
12558 2018-02-16 Julia Koval <julia.koval@intel.com>
12560 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12561 from PTA_CANNONLAKE.
12563 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12566 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12567 Use ++iter rather than iter++ for std::list iterators.
12568 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12569 defer deleting them until all nodes in the forest are processed. Do
12570 free even leaf nodes. Change to_process into auto_vec.
12573 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12574 * vec.h (vec_default_construct): Use memset instead of placement new
12575 if BROKEN_VALUE_INITIALIZATION is defined.
12576 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12577 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12580 PR rtl-optimization/83723
12581 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12582 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12583 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12585 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12587 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12589 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12591 PR rtl-optimization/81443
12592 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12593 from inner REGs to paradoxical SUBREGs.
12595 2018-02-16 Richard Biener <rguenther@suse.de>
12597 PR tree-optimization/84399
12598 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12599 For operands we can analyze at their definition make sure we can
12600 analyze them at each use as well.
12602 2018-02-16 Richard Biener <rguenther@suse.de>
12604 PR tree-optimization/84190
12605 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12606 volatile accesses if the decl isn't volatile.
12608 2018-02-15 Jason Merrill <jason@redhat.com>
12610 PR c++/84314 - ICE with templates and fastcall attribute.
12611 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12612 TYPE_CANONICAL on an existing type.
12614 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12616 PR tree-optimization/84383
12617 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12618 dstoff nor call operand_equal_p if dstbase is NULL.
12620 PR tree-optimization/84334
12621 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12622 also a CONSTANT_CLASS_P, punt.
12624 2018-02-14 Jim Wilson <jimw@sifive.com>
12626 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12627 first SMALL_OPERAND check. New local min_second_step. Move assert
12628 to where locals are set. Add TARGET_RVC support.
12629 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12631 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12633 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12635 2018-02-14 Martin Sebor <msebor@redhat.com>
12637 PR tree-optimization/83698
12638 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12639 arrays constrain the offset range to their bounds.
12640 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12641 (builtin_access::overlap): Avoid setting the size of overlap if it's
12643 (maybe_diag_overlap): Also consider arrays when deciding what values
12644 of offsets to include in diagnostics.
12646 2018-02-14 Martin Sebor <msebor@redhat.com>
12649 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12650 that correspond to the kind of a declaration.
12652 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12655 * config/pa/pa.md: Load address of PIC label using the linkage table
12656 if the label is nonlocal.
12658 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12660 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12661 warning message if user requests -maltivec=be.
12662 * doc/invoke.texi: Document deprecation of -maltivec=be.
12664 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12667 * config/rs6000/rs6000-c.c: Update definitions for
12668 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12669 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12671 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12674 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12675 add _get_ssp intrinsics. Remove argument from
12676 __builtin_ia32_rdssp[d|q].
12677 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12678 * config/i386/i386-builtin.def: Remove argument from
12679 __builtin_ia32_rdssp[d|q].
12680 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12681 ix86_expand_special_args_builtin for _rdssp[d|q].
12682 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12683 Clear register before usage.
12684 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12685 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12687 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12689 PR tree-optimization/84357
12690 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12691 operand 1 of an ARRAY_REF too.
12693 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12696 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12697 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12699 (set_of_reg): New struct.
12700 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12701 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12702 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12704 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12705 Split into bitclr, bitset, bitinvert patterns if appropriate.
12706 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12707 use rx_fuse_in_memory_bitop.
12708 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12709 to named insn, correct maximum insn length.
12711 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12714 * machmode.def: Define a complex mode for PARTIAL_INT.
12715 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12717 * doc/rtl.texi: Document CSPImode.
12718 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12720 (msp430_hard_regno_nregs_with_padding): Likewise.
12722 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12725 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12727 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12729 PR rtl-optimization/84169
12730 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12731 we generated a parallel as new i3 and we split that to new i2 and i3
12732 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12733 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12734 those to i2, not i1. Partially rewrite this scan code.
12736 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12739 * stor-layout.c (place_field): For variable length fields, adjust
12740 offset_align afterwards not just based on the field's alignment,
12741 but also on the size.
12743 PR middle-end/84309
12744 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12745 of exps and logs in the use_exp2 case.
12747 2018-02-13 Jeff Law <law@redhat.com>
12749 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12750 entry for "vector".
12752 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12755 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12759 * common.opt (gas-loc-support, gas-locview-support): New.
12760 (ginline-points, ginternal-reset-location-views): New.
12761 * doc/invoke.texi: Document them. Use @itemx where intended.
12762 (gvariable-location-views): Adjust.
12763 * target.def (reset_location_view): New.
12764 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12765 (TARGET_RESET_LOCATION_VIEW): New.
12766 * doc/tm.texi: Rebuilt.
12767 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12768 (dwarf2out_default_as_locview_support): New.
12769 (output_asm_line_debug_info): Use option variables.
12770 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12771 (output_loc_list): Likewise.
12772 (add_high_low_attributes): Check option variables.
12773 Don't output entry view attribute in strict mode.
12774 (gen_inlined_subroutine_die): Check option variables.
12775 (dwarf2out_inline_entry): Likewise.
12776 (init_sections_and_labels): Likewise.
12777 (dwarf2out_early_finish): Likewise.
12778 (maybe_reset_location_view): New, from...
12779 (dwarf2out_var_location): ... here. Call it.
12780 * debug.h (dwarf2out_default_as_loc_support): Declare.
12781 (dwarf2out_default_as_locview_support): Declare.
12782 * hooks.c (hook_int_rtx_insn_0): New.
12783 * hooks.h (hook_int_rtx_insn_0): Declare.
12784 * toplev.c (process_options): Take -gas-loc-support and
12785 -gas-locview-support from dwarf2out. Enable
12786 -gvariable-location-views by default only with locview
12787 assembler support. Enable -ginternal-reset-location-views by
12788 default only if the target defines the corresponding hook.
12789 Enable -ginline-points by default if location views are
12790 enabled; force it disabled if statement frontiers are
12792 * tree-inline.c (expand_call_inline): Check option variables.
12793 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12795 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12797 PR tree-optimization/84321
12798 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12799 handling. Also check whether the anti-range contains any values
12800 that satisfy the mask; switch to a VR_RANGE if not.
12802 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12805 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12807 2018-02-13 Martin Jambor <mjambor@suse.cz>
12810 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12811 of call statements, also set location of a load to a temporary.
12813 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12815 * config/rl78/rl78.c (add_vector_labels): New function.
12816 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12817 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12818 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12819 which checks that no arguments are passed.
12820 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12821 * doc/extend.texi: Documentation for the new attribute.
12823 2018-02-13 Andreas Schwab <schwab@suse.de>
12825 * config/riscv/linux.h (CPP_SPEC): Define.
12827 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12830 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12831 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12832 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12833 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12834 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12835 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12836 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12837 temporarily for AES and PCLMUL builtins.
12839 PR tree-optimization/84339
12840 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12841 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12844 PR middle-end/84309
12845 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12846 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12847 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12849 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12851 * omp-simd-clone.h: New file.
12852 * omp-simd-clone.c: Include omp-simd-clone.h.
12853 (expand_simd_clones): No longer static.
12854 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12855 cgraph.h and omp-simd-clone.h.
12856 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12857 (vect_recog_widen_shift_pattern): Formatting fix.
12858 (vect_pattern_recog_1): Don't check optab for calls.
12861 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12862 operands[2] into a REG before using gen_lowpart on it.
12864 2018-02-12 Jeff Law <law@redhat.com>
12867 * config/sh/sh.c (find_barrier): Consider a sibling call
12870 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12871 successfully back substituting a reg.
12873 2018-02-12 Richard Biener <rguenther@suse.de>
12875 PR tree-optimization/84037
12876 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12877 parameter, move visited init to caller.
12878 (vect_slp_analyze_operations): Separate cost from validity
12879 check, initialize visited once for all instances.
12880 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12882 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12884 (vect_model_promotion_demotion_cost): Likewise.
12885 (vectorizable_bswap): Guard cost modeling with !slp_node
12886 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12888 (vectorizable_call): Likewise.
12889 (vectorizable_conversion): Likewise.
12890 (vectorizable_assignment): Likewise.
12891 (vectorizable_shift): Likewise.
12892 (vectorizable_operation): Likewise.
12893 (vectorizable_store): Likewise.
12894 (vectorizable_load): Likewise.
12895 (vectorizable_condition): Likewise.
12896 (vectorizable_comparison): Likewise.
12898 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12901 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12902 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12903 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12905 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12907 PR middle-end/83665
12908 * params.def (inline-min-speedup): Increase from 8 to 15.
12909 (max-inline-insns-auto): Decrease from 40 to 30.
12910 * ipa-split.c (consider_split): Add some buffer for function to
12911 be considered inlining candidate.
12912 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12915 2018-02-12 Richard Biener <rguenther@suse.de>
12917 PR tree-optimization/84037
12918 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12919 matched stmts if we cannot swap the non-matched ones.
12921 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12923 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12924 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12925 _mm_maskz_scalef_round_ss): New intrinsics.
12926 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12927 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12928 __builtin_ia32_scalefss_round): Remove.
12929 (__builtin_ia32_scalefsd_mask_round,
12930 __builtin_ia32_scalefss_mask_round): New intrinsics.
12931 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12932 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12933 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12934 "<round_constraint>")): Changed to ...
12935 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12936 "<round_scalar_constraint>")): ... this.
12937 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12938 %0, %1, %2<round_op3>}"): Changed to ...
12939 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12940 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12941 %2<round_scalar_mask_op3>}"): ... this.
12942 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12944 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12946 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12947 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12948 (_mm_maskz_sqrt_round_ss): New intrinsics.
12949 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12950 (__builtin_ia32_sqrtsd_mask_round)
12951 (__builtin_ia32_sqrtss_mask_round): New builtins.
12952 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12953 (__builtin_ia32_sqrtss_round): Remove.
12954 (__builtin_ia32_sqrtsd_mask_round)
12955 (__builtin_ia32_sqrtss_mask_round): New builtins.
12956 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12957 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12958 ((match_operand:VF_128 1 "vector_operand"
12959 "xBm,<round_constraint>")): Changed to ...
12960 ((match_operand:VF_128 1 "vector_operand"
12961 "xBm,<round_scalar_constraint>")): ... this.
12962 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12963 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12964 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12965 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12966 %<iptr>1<round_scalar_mask_op3>}): ... this.
12967 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12968 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12970 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12973 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12974 Cast vec_cmpeq result to correct type.
12975 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12976 Cast vec_cmpgt result to correct type.
12978 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12980 * final.c (final_scan_insn_1): Renamed from...
12981 (final_scan_insn): ... this. New wrapper, to recover
12982 seen from the outermost call in recursive ones.
12983 * config/sparc/sparc.c (output_return): Drop seen from call.
12984 (output_sibcall): Likewise.
12985 * config/visium/visium.c (output_branch): Likewise.
12987 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12989 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12992 2018-02-10 Alan Modra <amodra@gmail.com>
12995 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12996 Specify LR as an input.
12998 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13001 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13002 remove_member_access_dummy_vars): New functions.
13003 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13004 lower_omp_1, execute_lower_omp): Use them.
13006 PR rtl-optimization/84308
13007 * shrink-wrap.c (spread_components): Release todo vector.
13009 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13011 PR rtl-optimization/57193
13012 * ira-color.c (struct allocno_color_data): Add member
13013 conflict_allocno_hard_prefs.
13014 (update_conflict_allocno_hard_prefs): New.
13015 (bucket_allocno_compare_func): Add a preference based on
13016 conflict_allocno_hard_prefs.
13017 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13018 (color_allocnos): Remove a dead code. Initiate
13019 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13021 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13024 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13025 constraint from =wa to wa. Avoid a subreg on the output operand,
13026 instead use a pseudo and subreg it in a move.
13027 (p9_xxbrd_<mode>): Changed to ...
13028 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13029 (p9_xxbrd_v2df): New expander.
13030 (p9_xxbrw_<mode>): Changed to ...
13031 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13032 (p9_xxbrw_v4sf): New expander.
13034 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13036 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13038 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13041 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13042 multiply in 32-bit mode.
13043 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13044 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13047 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13049 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13050 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13051 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13052 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13054 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13057 * dwarf2out.c (is_trivial_indirect_ref): New function.
13058 (dwarf2out_late_global_decl): Do not generate a location
13059 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13060 and that are not defined in the current unit.
13062 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13064 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13065 instead of a libcall for UNORDERED.
13067 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13070 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13071 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13073 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13076 * config/s390/s390.c (s390_set_current_function): Invoke
13077 s390_indirect_branch_settings also if fndecl didn't change.
13079 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13081 * config/rs6000/rs6000.md (blockage): Set length to zero.
13083 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13085 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13087 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13090 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13091 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13095 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13096 PARALLEL incoming that failed vt_get_decl_and_offset check.
13098 PR middle-end/84237
13099 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13100 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13102 (get_variable_section): For decls in named .bss* sections pass true as
13103 second argument to bss_initializer_p.
13105 2018-02-09 Marek Polacek <polacek@redhat.com>
13106 Jakub Jelinek <jakub@redhat.com>
13109 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13110 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13111 Sync some changes from cxx_fold_indirect_ref.
13113 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13115 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13117 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13118 (BLOCK_INLINE_ENTRY_LABEL): New.
13119 (dwarf2out_var_location): Disregard inline entry markers.
13120 (inline_entry_data): New struct.
13121 (inline_entry_data_hasher): New hashtable type.
13122 (inline_entry_data_hasher::hash): New.
13123 (inline_entry_data_hasher::equal): New.
13124 (inline_entry_data_table): New variable.
13125 (add_high_low_attributes): Add DW_AT_entry_pc and
13126 DW_AT_GNU_entry_view attributes if a pending entry is found
13127 in inline_entry_data_table. Add old entry_pc attribute only
13128 if debug nonbinding markers are disabled.
13129 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13130 markers are enabled.
13131 (block_within_block_p, dwarf2out_inline_entry): New.
13132 (dwarf2out_finish): Check that no entries remained in
13133 inline_entry_data_table.
13134 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13135 (final_scan_insn, notice_source_line): Likewise.
13136 (rest_of_clean_state): Skip inline entry markers.
13137 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13139 * gimple.c (gimple_build_debug_inline_entry): New.
13140 * gimple.h (enum gimple_debug_subcode): Add
13141 GIMPLE_DEBUG_INLINE_ENTRY.
13142 (gimple_build_debug_inline_entry): Declare.
13143 (gimple_debug_inline_entry_p): New.
13144 (gimple_debug_nonbind_marker_p): Adjust.
13145 * insn-notes.def (INLINE_ENTRY): New.
13146 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13147 inline entry marker notes.
13148 (print_insn): Likewise.
13149 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13150 (INSN_DEBUG_MARKER_KIND): Likewise.
13151 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13152 * tree-inline.c (expand_call_inline): Build and insert
13153 debug_inline_entry stmt.
13154 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13155 inline entry blocks early, if nonbind markers are enabled.
13156 (dump_scope_block): Dump fragment info.
13157 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13158 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13159 (gimple_build_debug_inline_entry): New.
13160 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13161 Enable/disable inline entry points too.
13162 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13163 (DEBUG_INSN): Describe inline entry markers.
13165 * common.opt (gvariable-location-views): New.
13166 (gvariable-location-views=incompat5): New.
13167 * config.in: Rebuilt.
13168 * configure: Rebuilt.
13169 * configure.ac: Test assembler for view support.
13170 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13171 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13172 * dwarf2out.c (var_loc_view): New typedef.
13173 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13174 (dwarf2out_locviews_in_attribute): New.
13175 (dwarf2out_locviews_in_loclist): New.
13176 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13177 (enum dw_line_info_opcode): Add LI_adv_address.
13178 (struct dw_line_info_table): Add view.
13179 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13180 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13181 (zero_view_p): New variable.
13182 (ZERO_VIEW_P): New macro.
13183 (output_asm_line_debug_info): New.
13184 (struct var_loc_node): Add view.
13185 (add_AT_view_list, AT_loc_list): New.
13186 (add_var_loc_to_decl): Add view param. Test it against last.
13187 (new_loc_list): Add view params. Record them.
13188 (AT_loc_list_ptr): Handle loc and view lists.
13189 (view_list_to_loc_list_val_node): New.
13190 (print_dw_val): Handle dw_val_class_view_list.
13191 (size_of_die): Likewise.
13192 (value_format): Likewise.
13193 (loc_list_has_views): New.
13194 (gen_llsym): Set vl_symbol too.
13195 (maybe_gen_llsym, skip_loc_list_entry): New.
13196 (dwarf2out_maybe_output_loclist_view_pair): New.
13197 (output_loc_list): Output view list or entries too.
13198 (output_view_list_offset): New.
13199 (output_die): Handle dw_val_class_view_list.
13200 (output_dwarf_version): New.
13201 (output_compilation_unit_header): Use it.
13202 (output_skeleton_debug_sections): Likewise.
13203 (output_rnglists, output_line_info): Likewise.
13204 (output_pubnames, output_aranges): Update version comments.
13205 (output_one_line_info_table): Output view numbers in asm comments.
13206 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13207 Call maybe_gen_llsym.
13208 (loc_list_from_tree_1): Adjust.
13209 (add_AT_location_description): Create view list attribute if
13210 needed, check it's absent otherwise.
13211 (convert_cfa_to_fb_loc_list): Adjust.
13212 (maybe_emit_file): Call output_asm_line_debug_info for test.
13213 (dwarf2out_var_location): Reset views as needed. Precompute
13214 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13215 attribute. Set view.
13216 (new_line_info_table): Reset next view.
13217 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13218 (dwarf2out_source_line): Likewise. Output view resets and labels to
13219 the assembler, or select appropriate line info opcodes.
13220 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13221 (optimize_string_length): Catch it. Adjust.
13222 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13223 dw_val_class_view_list, and remove it if no longer needed.
13224 (hash_loc_list): Hash view numbers.
13225 (loc_list_hasher::equal): Compare them.
13226 (optimize_location_lists): Check whether a view list symbol is
13227 needed, and whether the locview attribute is present, and
13228 whether they match. Remove the locview attribute if no longer
13230 (index_location_lists): Call skip_loc_list_entry for test.
13231 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13232 Use output_dwarf_version.
13233 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13234 (struct dw_val_node): Add val_view_list.
13235 * final.c (SEEN_NEXT_VIEW): New.
13236 (set_next_view_needed): New.
13237 (clear_next_view_needed): New.
13238 (maybe_output_next_view): New.
13239 (final_start_function): Rename to...
13240 (final_start_function_1): ... this. Take pointer to FIRST,
13241 add SEEN parameter. Emit param bindings in the initial view.
13242 (final_start_function): Reintroduce SEEN-less interface.
13243 (final): Rename to...
13244 (final_1): ... this. Take SEEN parameter. Output final pending
13245 next view at the end.
13246 (final): Reintroduce seen-less interface.
13247 (final_scan_insn): Output pending next view before switching
13248 sections or ending a block. Mark the next view as needed when
13249 outputting variable locations. Notify debug backend of section
13250 changes, and of location view changes.
13251 (rest_of_handle_final): Adjust.
13252 * toplev.c (process_options): Autodetect value for debug variable
13253 location views option. Warn on incompat5 without -gdwarf-5.
13254 * doc/invoke.texi (gvariable-location-views): New.
13255 (gvariable-location-views=incompat5): New.
13256 (gno-variable-location-views): New.
13258 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13260 PR tree-optimization/84136
13261 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13262 that the result of find_edge is non-NULL.
13264 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13267 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13268 storing integer register in SImode. Fix cost of 256 and 512
13269 byte aligned SSE register store.
13271 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13273 * config/i386/i386.c (ix86_multiplication_cost): Fix
13274 multiplication cost for TARGET_AVX512DQ.
13276 2018-02-08 Marek Polacek <polacek@redhat.com>
13278 PR tree-optimization/84238
13279 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13282 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13284 PR tree-optimization/84265
13285 * tree-vect-stmts.c (vectorizable_store): Don't treat
13286 VMAT_CONTIGUOUS accesses as grouped.
13287 (vectorizable_load): Likewise.
13289 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13291 PR tree-optimization/81635
13292 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13293 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13294 (test_round_for_mask): New functions.
13295 (wide_int_cc_tests): Call test_round_for_mask.
13296 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13297 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13298 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13299 range returned by get_range_info.
13301 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13304 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13305 * symtab.c: Include builtins.h
13306 (symtab_node::output_to_lto_symbol_table_p): Move here
13307 from lto-streamer-out.c:output_symbol_p.
13308 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13309 (output_symbol_p): Move all logic to symtab.c
13310 (produce_symtab): Update.
13312 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13314 * config/s390/s390-opts.h (enum indirect_branch): Define.
13315 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13316 (s390_indirect_branch_via_thunk)
13317 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13318 (enum s390_indirect_branch_type): Define.
13319 * config/s390/s390.c (struct s390_frame_layout, struct
13320 machine_function): Remove.
13321 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13322 (indirect_branch_table_label_no, indirect_branch_table_name):
13324 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13325 (enum s390_indirect_branch_option): Define.
13326 (s390_return_addr_from_memory): New function.
13327 (s390_handle_string_attribute): New function.
13328 (s390_attribute_table): Add new attribute handler.
13329 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13330 (s390_indirect_branch_via_thunk): New function.
13331 (s390_indirect_branch_via_inline_thunk): New function.
13332 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13333 sibling call optimization for non z10 compiles.
13334 (s390_emit_call): Force indirect branch target to be a single
13335 register. Add r1 clobber for non-z10 compiles.
13336 (s390_emit_epilogue): Emit return jump via return_use expander.
13337 (s390_reorg): Handle JUMP_INSNs as execute targets.
13338 (s390_option_override_internal): Perform validity checks for the
13339 new command line options.
13340 (s390_indirect_branch_attrvalue): New function.
13341 (s390_indirect_branch_settings): New function.
13342 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13343 (s390_output_indirect_thunk_function): New function.
13344 (s390_code_end): Implement target hook.
13345 (s390_case_values_threshold): Implement target hook.
13346 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13348 * config/s390/s390.h (struct s390_frame_layout)
13349 (struct machine_function): Move here from s390.c.
13350 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13351 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13352 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13353 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13354 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13355 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13356 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13357 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13358 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13359 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13360 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13361 (mnemonic attribute): Add values which aren't recognized
13363 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13364 pattern for branch conversion. Fix mnemonic attribute.
13365 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13366 indirect branch via thunk if requested.
13367 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13368 ("*indirect_jump"): Disable for branch conversion using out of
13370 ("indirect_jump_via_thunk<mode>_z10")
13371 ("indirect_jump_via_thunk<mode>")
13372 ("indirect_jump_via_inlinethunk<mode>_z10")
13373 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13374 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13375 ("casesi_jump_via_inlinethunk<mode>_z10")
13376 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13377 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13378 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13379 ("*indirect2_jump"): Disable for branch conversion.
13380 ("casesi_jump"): Turn into expander and expand patterns for branch
13382 ("return_use"): New expander.
13383 ("*return"): Emit return via thunk and rename it to ...
13384 ("*return<mode>"): ... this one.
13385 * config/s390/s390.opt: Add new options and and enum for the
13388 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13390 * lra-constraints.c (match_reload): Unconditionally use
13391 gen_lowpart_SUBREG, rather than selecting between that
13392 and equivalent gen_rtx_SUBREG code.
13394 2018-02-08 Richard Biener <rguenther@suse.de>
13396 PR tree-optimization/84233
13397 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13398 changed flag instead of boguously re-using phi_inserted.
13400 2018-02-08 Martin Jambor <mjambor@suse.cz>
13402 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13403 static local variables.
13405 2018-02-08 Richard Biener <rguenther@suse.de>
13407 PR tree-optimization/84278
13408 * tree-vect-stmts.c (vectorizable_store): When looking for
13409 smaller vector types to perform grouped strided loads/stores
13410 make sure the mode is supported by the target.
13411 (vectorizable_load): Likewise.
13413 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13415 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13416 Increase LDP/STP opportunities by adding adjacent callee-saves.
13418 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13420 PR rtl-optimization/84068
13421 PR rtl-optimization/83459
13422 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13424 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13426 PR tree-optimization/84224
13427 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13428 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13429 non-zero arguments.
13431 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13434 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13435 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13436 count, remove one USE.
13438 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13440 * doc/install.texi (Configuration): Document the
13441 --with-long-double-format={ibm,ieee} PowerPC configuration
13445 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13446 Convert from define_expand to be define_insn_and_split. Rework
13447 float/double/_Float128 conversions to QI/HI/SImode to work with
13448 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13449 conversions to QI/HImode types did a store and then a load to
13450 truncate the value. For conversions to VSX registers, don't split
13451 the insn, instead emit the code directly. Use the code iterator
13452 any_fix to combine signed and unsigned conversions.
13453 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13454 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13455 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13456 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13457 (fix_<mode>di2_hw): Likewise.
13458 (fixuns_<mode>di2_hw): Likewise.
13459 (fix_<mode>si2_hw): Likewise.
13460 (fixuns_<mode>si2_hw): Likewise.
13461 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13462 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13463 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13464 fix<uns>_trunc<SFDF:mode>si2_p8.
13465 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13466 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13467 (fix<uns>_<mode>_mem): Likewise.
13468 (fctiw<u>z_<mode>_mem): Likewise.
13469 (fix<uns>_<mode>_mem): Likewise.
13470 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13471 the register allocator from doing a direct move to the GPRs to do
13472 a store, and instead use the ISA 3.0 store byte/half-word from
13473 vector register instruction. For IEEE 128-bit floating point,
13474 also optimize stores of 32-bit ints.
13475 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13477 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13479 * genextract.c (push_pathstr_operand): New function to support
13481 (walk_rtx): Call push_pathstr_operand.
13482 (print_path): Support [a-zA-Z].
13484 2018-02-07 Richard Biener <rguenther@suse.de>
13486 PR tree-optimization/84037
13487 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13488 (cse_and_gimplify_to_preheader): Declare.
13489 (vect_get_place_in_interleaving_chain): Likewise.
13490 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13492 (_loop_vec_info::~_loop_vec_info): Delete it.
13493 (cse_and_gimplify_to_preheader): New function.
13494 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13495 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13496 (vectorizable_load): Likewise. For grouped stores always base
13497 the IV on the first element.
13498 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13499 condition before gimplifying.
13501 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13503 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13504 *DIV_EXPR and *MOD_EXPR.
13506 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13509 * config/i386/i386.c (ix86_option_override_internal): Mask out
13510 the CF_SET bit when checking -fcf-protection.
13512 2018-02-07 Tom de Vries <tom@codesourcery.com>
13515 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13518 2018-02-07 Richard Biener <rguenther@suse.de>
13520 PR tree-optimization/84204
13521 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13524 PR tree-optimization/84205
13525 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13526 special-case isl_ast_op_zdiv_r.
13528 PR tree-optimization/84223
13529 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13530 Only add conditions from within the region.
13531 (gather_bbs::after_dom_children): Adjust.
13533 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13536 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13537 * config/avr/avr.md: Only post-reload split REG-REG moves if
13538 either register is GENERAL_REG_P.
13540 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13542 PR tree-optimization/84235
13543 * tree-ssa-scopedtables.c
13544 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13545 if the subtraction is performed in floating point type where NaNs are
13546 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13547 build 1. Formatting fix.
13549 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13552 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13553 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13554 and skip it regardless of bb boundaries. Use CALL_P macro,
13555 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13556 unnecessarily, formatting fix.
13558 2018-02-06 Michael Collison <michael.collison@arm.com>
13560 * config/arm/thumb2.md:
13561 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13562 (*thumb_mov_notscc): Ditto.
13564 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13567 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13568 unsigned_fix, not "s".
13570 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13572 * configure.ac (gcc_fn_eh_frame_ro): New function.
13573 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13574 correct .eh_frame permissions.
13575 * configure: Regenerate.
13577 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13579 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13580 irrelevant options.
13582 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13584 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13585 Display warning message for -mno-speculate-indirect-jumps.
13587 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13589 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13591 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13593 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13595 PR tree-optimization/84225
13596 * tree-eh.c (find_trapping_overflow): Only call
13597 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13599 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13602 * config/i386/i386.c: Reimplement the check of possible options
13603 -mibt/-mshstk conbination. Change error messages.
13604 * doc/invoke.texi: Fix a typo: remove extra '='.
13606 2018-02-06 Marek Polacek <polacek@redhat.com>
13608 PR tree-optimization/84228
13609 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13611 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13614 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13615 emitted arch directives.
13616 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13617 __ARM_FEATURE_COPROC before changing architectures.
13619 2018-02-06 Richard Biener <rguenther@suse.de>
13621 * config/i386/i386.c (print_reg): Fix typo.
13622 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13624 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13626 * configure: Regenerate.
13628 2018-02-05 Martin Sebor <msebor@redhat.com>
13630 PR tree-optimization/83369
13631 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13634 2018-02-05 Martin Liska <mliska@suse.cz>
13636 * doc/invoke.texi: Cherry-pick upstream r323995.
13638 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13640 * ira.c (ira_init_register_move_cost): Adjust comment.
13642 2018-02-05 Martin Liska <mliska@suse.cz>
13644 PR gcov-profile/84137
13645 * doc/gcov.texi: Fix typo in documentation.
13647 2018-02-05 Martin Liska <mliska@suse.cz>
13649 PR gcov-profile/83879
13650 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13651 using dlopen functionality.
13653 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13655 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13656 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13657 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13658 _mm_maskz_range_round_ss): New intrinsics.
13659 (__builtin_ia32_rangesd128_round)
13660 (__builtin_ia32_rangess128_round): Remove.
13661 (__builtin_ia32_rangesd128_mask_round,
13662 __builtin_ia32_rangess128_mask_round): New builtins.
13663 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13664 __builtin_ia32_rangess128_round): Remove.
13665 (__builtin_ia32_rangesd128_mask_round,
13666 __builtin_ia32_rangess128_mask_round): New builtins.
13667 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13668 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13669 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13670 "<round_saeonly_constraint>")): Changed to ...
13671 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13672 "<round_saeonly_scalar_constraint>")): ... this.
13673 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13674 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13675 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13676 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13677 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13679 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13681 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13683 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13684 Remove all values except native, 8540 and 8548.
13686 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13688 * config/i386/i386.c (ix86_output_function_return): Pass
13689 INVALID_REGNUM, instead of -1, as invalid register number to
13690 indirect_thunk_name and output_indirect_thunk.
13692 2018-02-02 Julia Koval <julia.koval@intel.com>
13694 * config.gcc: Add -march=icelake.
13695 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13696 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13697 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13698 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13699 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13700 (processor_target_table): Add icelake.
13701 (ix86_option_override_internal): Handle new PTAs.
13702 (get_builtin_code_for_version): Handle icelake.
13703 (M_INTEL_COREI7_ICELAKE): New.
13704 (fold_builtin_cpu): Handle icelake.
13705 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13706 * doc/invoke.texi: Add -march=icelake.
13708 2018-02-02 Julia Koval <julia.koval@intel.com>
13710 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13711 to wide_int_bitmask.
13712 * wide-int-bitmask.h: New.
13714 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13717 * config/i386/i386.md: Replace Pmode with word_mode in
13718 builtin_setjmp_setup and builtin_longjmp to support x32.
13720 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13724 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13726 (rs6000_supported_cpu_names): New static variable.
13727 (linux_cpu_translation_table): Likewise.
13728 (elf_platform) <cpu>: Define new static variable and use it.
13729 Translate kernel AT_PLATFORM name to canonical name if needed.
13730 Error if platform name is unknown.
13732 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13735 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13737 2018-02-01 Jeff Law <law@redhat.com>
13740 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13741 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13742 the scratch if RELEASE_VIA_POP is false.
13743 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13744 If we have to save a temporary register, decrement SIZE appropriately.
13745 Pass new arguments to release_scratch_register_on_entry.
13746 (ix86_adjust_stack_and_probe): Likewise.
13747 (ix86_emit_probe_stack_range): Pass new arguments to
13748 release_scratch_register_on_entry.
13750 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13752 PR rtl-optimization/84157
13753 * combine.c (change_zero_ext): Use REG_P predicate in
13754 front of HARD_REGISTER_P predicate.
13756 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13758 * config/avr/avr.c (avr_option_override): Move disabling of
13759 -fdelete-null-pointer-checks to...
13760 * common/config/avr/avr-common.c (avr_option_optimization_table):
13763 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13765 PR tree-optimization/81635
13766 * tree-data-ref.c (split_constant_offset_1): For types that
13767 wrap on overflow, try to use range info to prove that wrapping
13770 2018-02-01 Renlin Li <renlin.li@arm.com>
13773 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13774 TAILCALL_ADDR_REGS.
13775 (aarch64_register_move_cost): Likewise.
13776 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13777 TAILCALL_ADDR_REGS.
13778 (REG_CLASS_NAMES): Likewise.
13779 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13780 TAILCALL_ADDR_REGS. Remove IP registers.
13781 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13783 2018-02-01 Richard Biener <rguenther@suse.de>
13785 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13786 for specifying RPO order and allow NULL for that.
13787 * domwalk.c (dom_walker::dom_walker): Likewise.
13788 (dom_walker::walk): Handle NULL RPO order.
13789 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13791 (rewrite_update_dom_walker): Likewise.
13792 (mark_def_dom_walker): Likewise.
13794 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13796 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13797 (aarch64_maybe_expand_sve_subreg_move): Declare.
13798 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13799 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13801 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13802 that are semantically a reverse operation.
13803 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13804 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13805 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13807 (aarch64_can_change_mode_class): For big-endian, forbid changes
13808 between two SVE modes if they have different element sizes.
13810 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13812 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13813 the TImode handling for big-endian targets.
13815 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13817 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13818 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13820 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13821 Remove BSWAP handing for big-endian targets and use the form of
13822 LD1RQ appropariate for the mode.
13824 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13826 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13827 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13828 duplicated element.
13830 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13833 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13834 check for operands that need to go through aarch64_sve_reload_be.
13836 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13838 PR tree-optimization/81661
13839 PR tree-optimization/84117
13840 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13841 * tree-eh.c: Include gimplify.h.
13842 (find_trapping_overflow, replace_trapping_overflow,
13843 rewrite_to_non_trapping_overflow): New functions.
13844 * tree-vect-loop.c: Include tree-eh.h.
13845 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13846 * tree-data-ref.c: Include tree-eh.h.
13847 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13849 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13851 PR rtl-optimization/84123
13852 * combine.c (change_zero_ext): Check if hard register satisfies
13853 can_change_dest_mode before calling gen_lowpart_SUBREG.
13855 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13858 * ira.c (ira_init_register_move_cost): Remove assert.
13860 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13862 PR rtl-optimization/84071
13863 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13864 * doc/tm.texi: Regenerate.
13866 2018-01-31 Richard Biener <rguenther@suse.de>
13868 PR tree-optimization/84132
13869 * tree-data-ref.c (analyze_miv_subscript): Properly
13870 check whether evolution_function_is_affine_multivariate_p
13871 before calling gcd_of_steps_may_divide_p.
13873 2018-01-31 Julia Koval <julia.koval@intel.com>
13876 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13877 * config/i386/i386.md (rdpid_rex64) New.
13878 (rdpid): Make 32bit only.
13880 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13883 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13884 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13886 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13889 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13891 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13893 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13895 PR rtl-optimization/84071
13896 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13897 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13899 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13901 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13902 (arc_attribute_table): Add 'aux' attribute.
13903 (arc_in_small_data_p): Consider aux like variables.
13904 (arc_is_aux_reg_p): New function.
13905 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13906 (arc_get_aux_arg): New function.
13907 (prepare_move_operands): Handle aux-register access.
13908 (arc_handle_aux_attribute): New function.
13909 * doc/extend.texi (ARC Variable attributes): Add subsection.
13911 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13913 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13914 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13915 (arc_attribute_table): Add 'uncached' attribute.
13916 (arc_print_operand): Print '.di' flag for uncached memory
13918 (arc_in_small_data_p): Do not consider for small data the uncached
13920 (arc_is_uncached_mem_p): New function.
13921 * config/arc/predicates.md (compact_store_memory_operand): Check
13922 for uncached memory accesses.
13923 (nonvol_nonimm_operand): Likewise.
13924 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13926 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13929 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13930 falign-loops=): Add Optimization flag.
13932 2018-01-30 Jeff Law <law@redhat.com>
13935 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13936 INT_REGISTERS_SAVED. Check it prior to calling
13937 get_scratch_register_on_entry.
13938 (ix86_adjust_stack_and_probe): Similarly.
13939 (ix86_emit_probe_stack_range): Similarly.
13940 (ix86_expand_prologue): Corresponding changes.
13942 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13945 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13946 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13948 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13951 * lra-constraints.c (curr_insn_transform): Process AND in the
13954 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13956 PR rtl-optimization/83986
13957 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13958 dependence against last_pending_memory_flush in addition to
13959 pending_jump_insns.
13961 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13963 PR tree-optimization/81611
13964 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13967 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13970 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13973 2018-01-30 Richard Biener <rguenther@suse.de>
13974 Jakub Jelinek <jakub@redhat.com>
13976 PR tree-optimization/84111
13977 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13978 inner loops added during recursion, as they don't have up-to-date
13981 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13984 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13985 (can_inline_edge_by_limits_p): ... here.
13986 (can_early_inline_edge_p, check_callers,
13987 update_caller_keys, update_callee_keys, recursive_inlining,
13988 add_new_edges_to_heap, speculation_useful_p,
13989 inline_small_functions,
13990 inline_small_functions, flatten_function,
13991 inline_to_all_callers_1): Update.
13993 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13995 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13998 2018-01-30 Richard Biener <rguenther@suse.de>
14000 PR tree-optimization/83008
14001 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14002 invariant and constant vector uses in stmts when they need
14003 more than one stmt.
14005 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14008 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14009 * configure: Regenerate.
14011 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14013 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14015 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14016 Use gen_rtx_REG rather than gen_lowpart.
14018 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14020 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14021 rather than 0 when creating partial subregs.
14023 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14025 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14028 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14031 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14032 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14033 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14034 flags. This restores the settings used before the 2017-07-24.
14035 Turning off pre increment/decrement/modify allows IVOPTS to
14036 optimize DF/SF loops where the index is an int.
14038 2018-01-29 Richard Biener <rguenther@suse.de>
14039 Kelvin Nilsen <kelvin@gcc.gnu.org>
14042 * tree-vect-stmts.c (vectorizable_call): Don't call
14043 targetm.vectorize_builtin_md_vectorized_function if callee is
14046 2018-01-22 Carl Love <cel@us.ibm.com>
14048 * doc/extend.tex: Fix typo in second arg in
14049 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14051 2018-01-29 Richard Biener <rguenther@suse.de>
14053 PR tree-optimization/84086
14054 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14055 (flush_ssaname_freelist): When SSA names were released reset
14056 the SCEV hash table.
14058 2018-01-29 Richard Biener <rguenther@suse.de>
14060 PR tree-optimization/84057
14061 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14062 removed paths when removing edges.
14064 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14066 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14067 -mfunction-return=@var{choice}.
14069 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14071 PR diagnostic/84034
14072 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14073 Handle CR like TAB.
14074 (layout::print_source_line): Likewise.
14075 (test_get_line_width_without_trailing_whitespace): Add test cases.
14077 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14079 PR middle-end/84040
14080 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14083 2018-01-26 Jim Wilson <jimw@sifive.com>
14085 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14087 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14090 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14092 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14093 and CMP + SUB-immediate -> SUBS.
14095 2018-01-26 Martin Sebor <msebor@redhat.com>
14097 PR tree-optimization/83896
14098 * tree-ssa-strlen.c (get_string_len): Rename...
14099 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14100 Avoid assuming length is constant.
14101 (handle_char_store): Use HOST_WIDE_INT for string length.
14103 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14106 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14107 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14109 2018-01-26 Richard Biener <rguenther@suse.de>
14111 PR rtl-optimization/84003
14112 * dse.c (record_store): Only record redundant stores when
14113 the earlier store aliases at least all accesses the later one does.
14115 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14117 PR rtl-optimization/83985
14118 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14119 REG_CFA_RESTORE insns.
14120 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14121 REG_CFA_RESTORE insns here.
14124 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14125 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14127 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14129 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14130 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14131 (arc_init): Likewise.
14132 (arc_override_options): Likewise.
14133 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14135 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14137 * config/arc/arc.h (TARGET_DBNZ): Define.
14138 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14139 properly set the tune attribute.
14140 (dbnz): Use TARGET_DBNZ guard.
14141 * config/arc/arc.opt (mtune): Add core3 option.
14143 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14145 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14146 recognize new pic like addresses.
14147 (arc_delegitimize_address): Clean up.
14149 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14151 * config/arc/arc-arches.def: Option mrf16 valid for all
14153 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14154 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14155 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14156 * config/arc/arc-tables.opt: Regenerate.
14157 * config/arc/arc.c (arc_conditional_register_usage): Handle
14158 reduced register file case.
14159 (arc_file_start): Set must have build attributes.
14160 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14161 mrf16 option value.
14162 * config/arc/arc.opt (mrf16): Add new option.
14163 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14164 * config/arc/genmultilib.awk: Handle new mrf16 option.
14165 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14166 * config/arc/t-multilib: Regenerate.
14167 * doc/invoke.texi (ARC Options): Document mrf16 option.
14169 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14171 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14172 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14173 (arc_attribute_table): Add 'secure_call' attribute.
14174 (arc_print_operand): Print secure call operand.
14175 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14177 (arc_is_secure_call_p): New function. * config/arc/arc.md
14178 (call_i): Add support for sjli instruction.
14179 (call_value_i): Likewise.
14180 * config/arc/constraints.md (Csc): New constraint.
14182 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14183 John Eric Martin <John.Martin@emmicro-us.com>
14185 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14186 * config/arc/arc.c (_arc_jli_section): New struct.
14187 (arc_jli_section): New type.
14188 (rc_jli_sections): New static variable.
14189 (arc_handle_jli_attribute): New function.
14190 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14191 (arc_file_end): New function.
14192 (TARGET_ASM_FILE_END): Define.
14193 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14194 (arc_add_jli_section): New function.
14195 (jli_call_scan): Likewise.
14196 (arc_reorg): Call jli_call_scan.
14197 (arc_output_addsi): Remove 'S' from printing asm operand.
14198 (arc_is_jli_call_p): New function.
14199 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14201 (movhi_insn): Likewise.
14202 (movsi_insn): Likewise.
14203 (movsi_set_cc_insn): Likewise.
14204 (loadqi_update): Likewise.
14205 (load_zeroextendqisi_update): Likewise.
14206 (load_signextendqisi_update): Likewise.
14207 (loadhi_update): Likewise.
14208 (load_zeroextendhisi_update): Likewise.
14209 (load_signextendhisi_update): Likewise.
14210 (loadsi_update): Likewise.
14211 (loadsf_update): Likewise.
14212 (movsicc_insn): Likewise.
14213 (bset_insn): Likewise.
14214 (bxor_insn): Likewise.
14215 (bclr_insn): Likewise.
14216 (bmsk_insn): Likewise.
14217 (bicsi3_insn): Likewise.
14218 (cmpsi_cc_c_insn): Likewise.
14219 (movsi_ne): Likewise.
14220 (movsi_cond_exec): Likewise.
14221 (clrsbsi2): Likewise.
14222 (norm_f): Likewise.
14232 (call_i): Remove 'S' asm letter, add jli instruction.
14233 (call_value_i): Likewise.
14234 * config/arc/arc.op (mjli-always): New option.
14235 * config/arc/constraints.md (Cji): New constraint.
14236 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14238 (subsf3_fpx): Likewise.
14239 (mulsf3_fpx): Likewise.
14240 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14242 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14243 function attrbutes.
14244 * doc/invoke.texi (ARC): Document mjli-always option.
14246 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14248 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14249 avoid addition with 0 and use incw and decw where possible.
14251 2018-01-26 Richard Biener <rguenther@suse.de>
14253 PR tree-optimization/81082
14254 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14255 association if it requires casting to unsigned.
14256 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14257 from fold_plusminus_mult_expr to catch important cases late when
14258 range info is available.
14260 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14262 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14263 * configure.ac (hidden_linkonce): New test.
14264 * configure: Regenerate.
14265 * config.in: Regenerate.
14267 2018-01-26 Julia Koval <julia.koval@intel.com>
14269 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14270 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14271 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14272 _mm_mask_bitshuffle_epi64_mask): Fix type.
14273 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14274 USI_FTYPE_V4DI_V4DI_USI): Remove.
14275 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14276 __builtin_ia32_vpshufbitqmb256_mask,
14277 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14278 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14279 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14281 2018-01-26 Alan Modra <amodra@gmail.com>
14284 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14285 UNSPEC_VBPERMQ. Sort other unspecs.
14287 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14289 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14291 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14293 PR middle-end/83055
14294 * predict.c (drop_profile): Do not push/pop cfun; update also
14296 (handle_missing_profiles): Fix logic looking for zero profiles.
14298 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14300 PR middle-end/83977
14301 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14302 on functions with #pragma omp declare simd or functions with simd
14304 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14305 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14306 Remove trailing \n from warning_at calls.
14308 2018-01-25 Tom de Vries <tom@codesourcery.com>
14311 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14312 for neutered workers.
14314 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14317 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14318 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14320 2018-01-24 Jeff Law <law@redhat.com>
14323 * i386.c (get_probe_interval): Move to earlier point.
14324 (ix86_compute_frame_layout): If -fstack-clash-protection and
14325 the frame is larger than the probe interval, then use pushes
14326 to save registers rather than reg->mem moves.
14327 (ix86_expand_prologue): Remove conditional for int_registers_saved
14330 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14333 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14334 min/max for never referenced object.
14336 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14338 PR middle-end/83977
14339 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14341 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14342 attributes from DECL_ATTRIBUTES (decl) without affecting
14343 DECL_ATTRIBUTES (current_function_decl).
14344 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14345 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14347 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14349 PR tree-optimization/83979
14350 * fold-const.c (fold_comparison): Use constant_boolean_node
14351 instead of boolean_{true,false}_node.
14353 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14355 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14358 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14360 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14361 Simplify the clause that sets the length attribute.
14362 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14363 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14364 clause that sets the length attribute.
14365 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14367 2018-01-24 Tom de Vries <tom@codesourcery.com>
14370 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14371 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14372 Add strict parameter.
14373 (prevent_branch_around_nothing): Insert dummy insn between branch to
14374 label and label with no ptx insn inbetween.
14375 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14377 2018-01-24 Tom de Vries <tom@codesourcery.com>
14380 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14381 for neutered threads in warp.
14382 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14384 2018-01-24 Richard Biener <rguenther@suse.de>
14386 PR tree-optimization/83176
14387 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14390 2018-01-24 Richard Biener <rguenther@suse.de>
14392 PR tree-optimization/82819
14393 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14394 code generating pluses that are no-ops in the target precision.
14396 2018-01-24 Richard Biener <rguenther@suse.de>
14398 PR middle-end/84000
14399 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14401 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14403 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14404 to merge probabilities.
14405 * predict.c (probably_never_executed): Also mark as cold functions
14406 with global 0 profile and guessed local profile.
14407 * profile-count.c (profile_probability::combine_with_count): New
14409 * profile-count.h (profile_probability::operator*,
14410 profile_probability::operator*=, profile_probability::operator/,
14411 profile_probability::operator/=): Reduce precision to adjusted
14412 and set value to guessed on contradictory divisions.
14413 (profile_probability::combine_with_freq): Remove.
14414 (profile_probability::combine_wiht_count): Declare.
14415 (profile_count::force_nonzero):: Set to adjusted.
14416 (profile_count::probability_in):: Set quality to adjusted.
14417 * tree-ssa-tail-merge.c (replace_block_by): Use
14418 combine_with_count.
14420 2018-01-23 Andrew Waterman <andrew@sifive.com>
14421 Jim Wilson <jimw@sifive.com>
14423 * config/riscv/riscv.c (riscv_stack_boundary): New.
14424 (riscv_option_override): Set riscv_stack_boundary. Handle
14425 riscv_preferred_stack_boundary_arg.
14426 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14427 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14428 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14429 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14430 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14431 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14433 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14436 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14437 of struct ix86_frame.
14438 (ix86_expand_epilogue): Likewise. Add a local variable for
14439 the reg_save_offset field in struct ix86_frame.
14441 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14443 PR tree-optimization/82604
14444 * tree-loop-distribution.c (enum partition_kind): New enum item
14445 PKIND_PARTIAL_MEMSET.
14446 (partition_builtin_p): Support above new enum item.
14447 (generate_code_for_partition): Ditto.
14448 (compute_access_range): Differentiate cases that equality can be
14449 proven at all loops, the innermost loops or no loops.
14450 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14451 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14452 (finalize_partitions, distribute_loop): Don't fuse partition of
14453 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14454 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14455 parloop is enabled.
14457 2018-01-23 Martin Liska <mliska@suse.cz>
14459 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14460 order to ignore the predictor.
14461 (PRED_POLYMORPHIC_CALL): Likewise.
14462 (PRED_RECURSIVE_CALL): Likewise.
14464 2018-01-23 Martin Liska <mliska@suse.cz>
14466 * tree-profile.c (tree_profiling): Print function header to
14467 aware reader which function we are working on.
14468 * value-prof.c (gimple_find_values_to_profile): Do not print
14469 not interesting value histograms.
14471 2018-01-23 Martin Liska <mliska@suse.cz>
14473 * profile-count.h (enum profile_quality): Add
14474 profile_uninitialized as the first value. Do not number values
14475 as they are zero based.
14476 (profile_count::verify): Update sanity check.
14477 (profile_probability::verify): Likewise.
14479 2018-01-23 Nathan Sidwell <nathan@acm.org>
14481 * doc/invoke.texi (ffor-scope): Deprecate.
14483 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14485 PR tree-optimization/83510
14486 * domwalk.c (set_all_edges_as_executable): New function.
14487 (dom_walker::dom_walker): Convert bool param
14488 "skip_unreachable_blocks" to enum reachability. Move setup of
14489 edge flags to set_all_edges_as_executable and only do it when
14490 reachability is REACHABLE_BLOCKS.
14491 * domwalk.h (enum dom_walker::reachability): New enum.
14492 (dom_walker::dom_walker): Convert bool param
14493 "skip_unreachable_blocks" to enum reachability.
14494 (set_all_edges_as_executable): New decl.
14495 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14496 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14498 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14499 but converting true to REACHABLE_BLOCKS.
14500 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14502 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14503 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14504 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14506 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14507 if check_all_array_refs will be called.
14509 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14511 * tree.c (selftest::test_location_wrappers): Add more test
14514 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14516 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14517 (selftest::test_bit_in_range): Likewise.
14519 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14522 * doc/sourcebuild.texi (vect_float): Say that the selector
14523 only describes the situation when -funsafe-math-optimizations is on.
14524 (vect_float_strict): Document.
14526 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14528 PR tree-optimization/83965
14529 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14530 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14531 instead of checking only for a reduction.
14532 (vect_recog_widen_sum_pattern): Likewise.
14534 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14536 * predict.c (probably_never_executed): Only use precise profile info.
14537 (compute_function_frequency): Skip after inlining hack since we now
14538 have quality checking.
14540 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14542 * profile-count.h (profile_probability::very_unlikely,
14543 profile_probability::unlikely, profile_probability::even): Set
14544 precision to guessed.
14546 2018-01-23 Richard Biener <rguenther@suse.de>
14548 PR tree-optimization/83963
14549 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14550 Properly terminate dominator walk when crossing the exit edge not
14551 when visiting its source block.
14553 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14556 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14557 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14559 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14561 PR tree-optimization/83957
14562 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14563 semicolon after for body surrounded by braces.
14565 PR tree-optimization/83081
14566 * profile-count.h (profile_probability::split): New method.
14567 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14568 Use profile_probability::split.
14569 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14570 when splitting a single conditional jump into 2.
14572 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14574 PR tree-optimization/69452
14575 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14578 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14580 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14581 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14582 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14584 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14586 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14588 * config/rl78/rl78.md (movdi): New define_expand.
14589 * config/rl78/rl78.c (rl78_split_movdi): New function.
14591 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14594 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14596 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14597 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14598 128-bit to produce an UNSPEC move to get the double word with the
14599 signbit and then a shift directly to do signbit.
14600 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14601 implementation with a new version that just does either a direct
14602 move or a regular move. Move memory interface to separate insns.
14603 Move insns so they are next to the expander.
14604 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14605 with signbit move. Split big and little endian case.
14606 (signbit<mode>2_dm_mem_le): Likewise.
14607 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14608 (signbit<mode>2_dm2): Likewise.
14610 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14612 * config/rl78/rl78.md (anddi3): New define_expand.
14614 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14616 * config/rl78/rl78.md (umindi3): New define_expand.
14618 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14620 * config/rl78/rl78.md (smindi3): New define_expand.
14622 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14624 * config/rl78/rl78.md (smaxdi3): New define_expand.
14626 2018-01-22 Carl Love <cel@us.ibm.com>
14628 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14629 LVX_V1TI): Add macro expansion.
14630 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14631 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14632 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14633 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14634 Change check to determine if the instruction is a byte reversing
14635 entry. Fix typo in comment.
14636 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14637 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14638 Add def_builtin calls for new builtins.
14639 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14640 Add define_insn expansion.
14642 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14644 * config/rl78/rl78.md (umaxdi3): New define_expand.
14646 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14648 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14649 for non-QImode registers.
14651 2018-01-22 Richard Biener <rguenther@suse.de>
14653 PR tree-optimization/83963
14654 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14655 including the loop exit block.
14656 (scop_detection::merge_sese): Likewise.
14657 (scop_detection::add_scop): Do it here instead.
14659 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14661 * doc/sourcebuild.texi (arm_softfloat): Document.
14663 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14666 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14667 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14668 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14670 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14671 David Edelsohn <dje.gcc@gmail.com>
14674 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14675 Change "crset eq" to "crset 2".
14676 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14677 (*call_indirect_aix<mode>_nospec): Likewise.
14678 (*call_value_indirect_aix<mode>_nospec): Likewise.
14679 (*call_indirect_elfv2<mode>_nospec): Likewise.
14680 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14681 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14682 change assembly output from . to $.
14683 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14684 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14685 (*tablejump<mode>_internal1_nospec): Likewise.
14687 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14690 * config/sh/sh_optimize_sett_clrt.cc:
14691 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14693 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14695 PR tree-optimization/83940
14696 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14697 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14698 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14699 use it to pass back the definition type.
14700 (vect_check_store_rhs): Likewise rhs_dt_out.
14701 (vect_build_gather_load_calls): Add a mask_dt argument and use
14702 it instead of a call to vect_is_simple_use.
14703 (vectorizable_store): Update calls to vect_check_load_store_mask
14704 and vect_check_store_rhs. Use the dt returned by the latter instead
14705 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14706 instead of calls to vect_is_simple_use. Pass the scalar rather
14707 than the vector operand to vect_is_simple_use when handling
14708 second and subsequent copies of an rhs value.
14709 (vectorizable_load): Update calls to vect_check_load_store_mask
14710 and vect_build_gather_load_calls. Use the cached mask_dt and
14711 gs_info.offset_dt instead of calls to vect_is_simple_use.
14713 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14715 PR middle-end/83945
14716 * tree-emutls.c: Include gimplify.h.
14717 (lower_emutls_2): New function.
14718 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14719 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14720 it before further processing.
14723 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14724 UINTVAL (trueop1) instead of INTVAL (op1).
14726 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14730 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14731 INCOMING_FRAME_SP_OFFSET if not defined.
14732 (scan_trace): Add ENTRY argument. If true and
14733 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14734 emit a note to adjust the CFA offset.
14735 (create_cfi_notes): Adjust scan_trace callers.
14736 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14737 INCOMING_FRAME_SP_OFFSET in the CIE.
14738 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14739 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14741 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14742 * doc/tm.texi: Regenerated.
14744 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14746 PR rtl-optimization/83147
14747 * lra-constraints.c (remove_inheritance_pseudos): Use
14748 lra_substitute_pseudo_within_insn.
14750 2018-01-19 Tom de Vries <tom@codesourcery.com>
14751 Cesar Philippidis <cesar@codesourcery.com>
14754 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14756 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14759 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14760 spaces for function labels.
14762 2018-01-19 Martin Liska <mliska@suse.cz>
14764 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14765 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14766 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14767 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14768 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14769 (PRED_CONST_RETURN): Change from 69 to 65.
14770 (PRED_NULL_RETURN): Change from 91 to 71.
14771 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14772 (PRED_LOOP_GUARD): Change from 66 to 73.
14774 2018-01-19 Martin Liska <mliska@suse.cz>
14776 * predict.c (predict_insn_def): Add new assert.
14777 (struct branch_predictor): Change type to signed integer.
14778 (test_prediction_value_range): Amend test to cover
14779 PROB_UNINITIALIZED.
14780 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14781 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14782 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14783 (PRED_LOOP_IV_COMPARE): Likewise.
14784 * predict.h (PROB_UNINITIALIZED): Define new constant.
14786 2018-01-19 Martin Liska <mliska@suse.cz>
14788 * predict.c (dump_prediction): Add new format for
14789 analyze_brprob.py script which is enabled with -details
14791 * profile-count.h (precise_p): New function.
14793 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14795 PR tree-optimization/83922
14796 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14797 there are no statements that need masking.
14798 (vect_active_double_reduction_p): New function.
14799 (vect_analyze_loop_operations): Use it when handling phis that
14800 are not in the loop header.
14802 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14804 PR tree-optimization/83914
14805 * tree-vect-loop.c (vectorizable_induction): Don't convert
14806 init_expr or apply the peeling adjustment for inductions
14807 that are nested within the vectorized loop.
14809 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14811 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14814 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14818 * function.h (gimplify_parameters): Add gimple_seq * argument.
14819 * function.c: Include gimple.h and options.h.
14820 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14821 for the added local temporaries if needed.
14822 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14823 if there are any parameter cleanups, wrap whole body into a
14824 try/finally with the cleanups.
14826 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14829 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14830 Use GET_MODE_CLASS for scalar floating point.
14832 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14836 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14837 Fix call of call_cgraph_insertion_hooks.
14839 2018-01-18 Martin Sebor <msebor@redhat.com>
14841 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14843 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14846 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14849 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14852 * common.opt: (-ffile-prefix-map): New option.
14853 * opts.c (common_handle_option): Defer it.
14854 * opts-global.c (handle_common_deferred_options): Handle it.
14855 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14856 * file-prefix-map.h: New file.
14857 (remap_debug_filename, add_debug_prefix_map): ...here.
14858 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14859 * final.c (debug_prefix_map, add_debug_prefix_map
14860 remap_debug_filename): Move to...
14861 * file-prefix-map.c: New file.
14862 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14863 generalize, get rid of alloca(), use strrchr() instead of strchr().
14864 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14865 Implement in terms of add_prefix_map().
14866 (remap_macro_filename, remap_debug_filename): Implement in term of
14868 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14869 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14870 * dbxout.c: Include file-prefix-map.h.
14871 * varasm.c: Likewise.
14872 * vmsdbgout.c: Likewise.
14873 * xcoffout.c: Likewise.
14874 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14875 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14876 * doc/invoke.texi (-ffile-prefix-map): Document.
14877 (-fdebug-prefix-map): Update description.
14879 2018-01-18 Martin Liska <mliska@suse.cz>
14881 * config/i386/i386.c (indirect_thunk_name): Document that also
14883 (output_indirect_thunk): Document why both instructions
14884 (pause and lfence) are generated.
14886 2018-01-18 Richard Biener <rguenther@suse.de>
14888 PR tree-optimization/83887
14889 * graphite-scop-detection.c
14890 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14891 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14892 (scop_detection::merge_sese): Re-implement with a flood-fill
14893 algorithm that properly finds a SESE region if it exists.
14895 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14898 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14899 pointer_diff optimizations use view_convert instead of convert.
14901 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14903 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14904 Generate different code for -mno-speculate-indirect-jumps.
14905 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14906 (*call_indirect_aix<mode>): Disable for
14907 -mno-speculate-indirect-jumps.
14908 (*call_indirect_aix<mode>_nospec): New define_insn.
14909 (*call_value_indirect_aix<mode>): Disable for
14910 -mno-speculate-indirect-jumps.
14911 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14912 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14913 -mno-speculate-indirect-jumps.
14914 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14916 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14918 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14919 long double type, set the flags for noting the default long double
14920 type, even if we don't pass or return a long double type.
14922 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14925 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14928 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14930 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14931 support for merge[hl].
14932 (fold_mergehl_helper): New helper function.
14933 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14934 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14935 (altivec_vmrglw_direct): Add xxmrglw insn.
14937 2018-01-17 Andrew Waterman <andrew@sifive.com>
14939 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14940 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14942 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14945 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14946 call the lto_location_cache before reading the
14947 DECL_SOURCE_LOCATION of the types.
14949 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14950 Richard Sandiford <richard.sandiford@linaro.org>
14952 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14953 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14954 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14955 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14956 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14958 * config/aarch64/constraints.md (aarch64_movti_operand):
14960 * config/aarch64/predicates.md (Uti): Add new constraint.
14962 2018-01-17 Carl Love <cel@us.ibm.com>
14964 * config/rs6000/vsx.md (define_expand xl_len_r,
14965 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14966 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14968 (define_expand, define_insn): Move the shift left from the
14969 define_insn to the define_expand for lxvl and stxvl instructions.
14970 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14971 and XL_LEN_R definitions to PURE.
14973 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14975 * config/i386/i386.c (indirect_thunk_name): Declare regno
14976 as unsigned int. Compare regno with INVALID_REGNUM.
14977 (output_indirect_thunk): Ditto.
14978 (output_indirect_thunk_function): Ditto.
14979 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14980 in the call to output_indirect_thunk_function.
14982 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14984 PR middle-end/83884
14985 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14986 rather than the size of inner_type to determine the stack slot size
14987 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14989 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14992 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14995 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14997 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14998 endian Linux systems to optionally enable multilibs for selecting
14999 the long double type if the user configured an explicit type.
15000 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15001 have no long double multilibs if not defined.
15002 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15003 warn if the user used -mabi={ieee,ibm}longdouble and we built
15004 multilibs for long double.
15005 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15006 appropriate multilib option.
15007 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15009 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15010 for building long double multilibs.
15011 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15013 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15015 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15018 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15020 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15023 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15026 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15029 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15031 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15032 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15034 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15036 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15037 different rtl trees depending on TARGET_64BIT.
15038 (rs6000_gen_lvx): Likewise.
15040 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15042 * config/visium/visium.md (nop): Tweak comment.
15043 (hazard_nop): Likewise.
15045 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15047 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15048 -mspeculate-indirect-jumps.
15049 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15050 for -mno-speculate-indirect-jumps.
15051 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15052 (*call_value_indirect_elfv2<mode>): Disable for
15053 -mno-speculate-indirect-jumps.
15054 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15055 (indirect_jump): Emit different RTL for
15056 -mno-speculate-indirect-jumps.
15057 (*indirect_jump<mode>): Disable for
15058 -mno-speculate-indirect-jumps.
15059 (*indirect_jump<mode>_nospec): New define_insn.
15060 (tablejump): Emit different RTL for
15061 -mno-speculate-indirect-jumps.
15062 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15063 (tablejumpsi_nospec): New define_expand.
15064 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15065 (tablejumpdi_nospec): New define_expand.
15066 (*tablejump<mode>_internal1): Disable for
15067 -mno-speculate-indirect-jumps.
15068 (*tablejump<mode>_internal1_nospec): New define_insn.
15069 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15072 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15074 * caller-save.c (insert_save): Drop unnecessary parameter. All
15077 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15078 Richard Biener <rguenth@suse.de>
15081 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15082 return early, inline manually is_gimple_sizepos. Make sure if we
15083 call gimplify_expr we don't end up with a gimple constant.
15084 * tree.c (variably_modified_type_p): Don't return true for
15085 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15086 * gimplify.h (is_gimple_sizepos): Remove.
15088 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15090 PR tree-optimization/83857
15091 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15092 vectorizable_live_operation for pure SLP statements.
15093 (vectorizable_live_operation): Handle PHIs.
15095 2018-01-16 Richard Biener <rguenther@suse.de>
15097 PR tree-optimization/83867
15098 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15099 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15101 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15104 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15105 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15106 If off is not INTEGER_CST, issue a may not be aligned warning
15107 rather than isn't aligned. Use isn%'t rather than isn't.
15108 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15110 <case MULT_EXPR>: Improve the case when bottom and one of the
15111 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15112 operand, in that case check if the other operand is multiple of
15113 bottom divided by the INTEGER_CST operand.
15115 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15118 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15119 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15120 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15121 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15122 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15123 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15124 (pa_function_arg_size): New function.
15126 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15128 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15129 in a separate statement.
15131 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15133 PR tree-optimization/83847
15134 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15135 group gathers and scatters.
15137 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15139 PR rtl-optimization/86620
15140 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15142 PR rtl-optimization/83213
15143 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15144 to last if both are JUMP_INSNs.
15146 PR tree-optimization/83843
15147 * gimple-ssa-store-merging.c
15148 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15149 store_immediate_info for bswap/nop orig_stores.
15151 2018-01-15 Andrew Waterman <andrew@sifive.com>
15153 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15155 <UDIV>: Increase cost if !TARGET_DIV.
15157 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15159 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15160 (define_attr "cr_logical_3op"): New.
15161 (cceq_ior_compare): Adjust.
15162 (cceq_ior_compare_complement): Adjust.
15163 (*cceq_rev_compare): Adjust.
15164 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15165 (is_cracked_insn): Adjust.
15166 (insn_must_be_first_in_group): Adjust.
15167 * config/rs6000/40x.md: Adjust.
15168 * config/rs6000/440.md: Adjust.
15169 * config/rs6000/476.md: Adjust.
15170 * config/rs6000/601.md: Adjust.
15171 * config/rs6000/603.md: Adjust.
15172 * config/rs6000/6xx.md: Adjust.
15173 * config/rs6000/7450.md: Adjust.
15174 * config/rs6000/7xx.md: Adjust.
15175 * config/rs6000/8540.md: Adjust.
15176 * config/rs6000/cell.md: Adjust.
15177 * config/rs6000/e300c2c3.md: Adjust.
15178 * config/rs6000/e500mc.md: Adjust.
15179 * config/rs6000/e500mc64.md: Adjust.
15180 * config/rs6000/e5500.md: Adjust.
15181 * config/rs6000/e6500.md: Adjust.
15182 * config/rs6000/mpc.md: Adjust.
15183 * config/rs6000/power4.md: Adjust.
15184 * config/rs6000/power5.md: Adjust.
15185 * config/rs6000/power6.md: Adjust.
15186 * config/rs6000/power7.md: Adjust.
15187 * config/rs6000/power8.md: Adjust.
15188 * config/rs6000/power9.md: Adjust.
15189 * config/rs6000/rs64.md: Adjust.
15190 * config/rs6000/titan.md: Adjust.
15192 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15194 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15195 ix86_indirect_branch_register logic.
15197 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15199 * config/i386/constraints.md (Bs): Update
15200 ix86_indirect_branch_register check. Don't check
15201 ix86_indirect_branch_register with GOT_memory_operand.
15203 * config/i386/predicates.md (GOT_memory_operand): Don't check
15204 ix86_indirect_branch_register here.
15205 (GOT32_symbol_operand): Likewise.
15207 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15209 * config/i386/predicates.md (constant_call_address_operand):
15210 Rewrite ix86_indirect_branch_register logic.
15211 (sibcall_insn_operand): Likewise.
15213 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15215 * config/i386/constraints.md (Bs): Replace
15216 ix86_indirect_branch_thunk_register with
15217 ix86_indirect_branch_register.
15219 * config/i386/i386.md (indirect_jump): Likewise.
15220 (tablejump): Likewise.
15221 (*sibcall_memory): Likewise.
15222 (*sibcall_value_memory): Likewise.
15223 Peepholes of indirect call and jump via memory: Likewise.
15224 * config/i386/i386.opt: Likewise.
15225 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15226 (GOT_memory_operand): Likewise.
15227 (call_insn_operand): Likewise.
15228 (sibcall_insn_operand): Likewise.
15229 (GOT32_symbol_operand): Likewise.
15231 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15233 PR middle-end/83837
15234 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15235 type rather than type addr's type points to.
15236 (expand_omp_atomic_mutex): Likewise.
15237 (expand_omp_atomic): Likewise.
15239 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15242 * config/i386/i386.c (output_indirect_thunk_function): Use
15243 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15244 for __x86_return_thunk.
15246 2018-01-15 Richard Biener <rguenther@suse.de>
15248 PR middle-end/83850
15249 * expmed.c (extract_bit_field_1): Fix typo.
15251 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15254 * config/arm/iterators.md (VF): New mode iterator.
15255 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15256 Remove integer-related logic from pattern.
15257 (neon_vabd<mode>_3): Likewise.
15259 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15261 PR middle-end/82694
15262 * common.opt (fstrict-overflow): No longer an alias.
15263 (fwrapv-pointer): New option.
15264 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15265 also for pointer types based on flag_wrapv_pointer.
15266 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15267 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15268 opts->x_flag_wrapv got set.
15269 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15270 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15271 POINTER_TYPE_OVERFLOW_UNDEFINED.
15272 * match.pd: Likewise in address comparison pattern.
15273 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15275 2018-01-15 Richard Biener <rguenther@suse.de>
15278 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15279 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15280 Reset type names to their identifier if their TYPE_DECL doesn't
15281 have linkage (and thus is used for ODR and devirt).
15282 (save_debug_info_for_decl): Remove.
15283 (save_debug_info_for_type): Likewise.
15284 (add_tree_to_fld_list): Adjust.
15285 * tree-pretty-print.c (dump_generic_node): Make dumping of
15286 type names more robust.
15288 2018-01-15 Richard Biener <rguenther@suse.de>
15290 * BASE-VER: Bump to 8.0.1.
15292 2018-01-14 Martin Sebor <msebor@redhat.com>
15295 * builtins.c (check_access): Avoid warning when the no-warning bit
15298 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15300 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15301 * ira-color (allocno_hard_regs_compare): Likewise.
15303 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15306 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15307 Use .pushsection/.popsection.
15309 2018-01-14 Martin Sebor <msebor@redhat.com>
15312 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15314 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15316 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15317 entry from extra_headers.
15318 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15319 extra_headers, make the list bitwise identical to the i?86-*-* one.
15321 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15323 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15324 -mcmodel=large with -mindirect-branch=thunk,
15325 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15326 -mfunction-return=thunk-extern.
15327 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15328 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15329 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15331 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15333 * config/i386/i386.c (print_reg): Print the name of the full
15334 integer register without '%'.
15335 (ix86_print_operand): Handle 'V'.
15336 * doc/extend.texi: Document 'V' modifier.
15338 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15340 * config/i386/constraints.md (Bs): Disallow memory operand for
15341 -mindirect-branch-register.
15343 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15344 (GOT_memory_operand): Likewise.
15345 (call_insn_operand): Likewise.
15346 (sibcall_insn_operand): Likewise.
15347 (GOT32_symbol_operand): Likewise.
15348 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15349 for -mindirect-branch-register.
15350 (tablejump): Likewise.
15351 (*sibcall_memory): Likewise.
15352 (*sibcall_value_memory): Likewise.
15353 Disallow peepholes of indirect call and jump via memory for
15354 -mindirect-branch-register.
15355 (*call_pop): Replace m with Bw.
15356 (*call_value_pop): Likewise.
15357 (*sibcall_pop_memory): Replace m with Bs.
15358 * config/i386/i386.opt (mindirect-branch-register): New option.
15359 * doc/invoke.texi: Document -mindirect-branch-register option.
15361 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15363 * config/i386/i386-protos.h (ix86_output_function_return): New.
15364 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15365 set function_return_type.
15366 (indirect_thunk_name): Add ret_p to indicate thunk for function
15368 (output_indirect_thunk_function): Pass false to
15369 indirect_thunk_name.
15370 (ix86_output_indirect_branch_via_reg): Likewise.
15371 (ix86_output_indirect_branch_via_push): Likewise.
15372 (output_indirect_thunk_function): Create alias for function
15373 return thunk if regno < 0.
15374 (ix86_output_function_return): New function.
15375 (ix86_handle_fndecl_attribute): Handle function_return.
15376 (ix86_attribute_table): Add function_return.
15377 * config/i386/i386.h (machine_function): Add
15378 function_return_type.
15379 * config/i386/i386.md (simple_return_internal): Use
15380 ix86_output_function_return.
15381 (simple_return_internal_long): Likewise.
15382 * config/i386/i386.opt (mfunction-return=): New option.
15383 (indirect_branch): Mention -mfunction-return=.
15384 * doc/extend.texi: Document function_return function attribute.
15385 * doc/invoke.texi: Document -mfunction-return= option.
15387 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15389 * config/i386/i386-opts.h (indirect_branch): New.
15390 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15391 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15392 with local indirect jump when converting indirect call and jump.
15393 (ix86_set_indirect_branch_type): New.
15394 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15395 (indirectlabelno): New.
15396 (indirect_thunk_needed): Likewise.
15397 (indirect_thunk_bnd_needed): Likewise.
15398 (indirect_thunks_used): Likewise.
15399 (indirect_thunks_bnd_used): Likewise.
15400 (INDIRECT_LABEL): Likewise.
15401 (indirect_thunk_name): Likewise.
15402 (output_indirect_thunk): Likewise.
15403 (output_indirect_thunk_function): Likewise.
15404 (ix86_output_indirect_branch_via_reg): Likewise.
15405 (ix86_output_indirect_branch_via_push): Likewise.
15406 (ix86_output_indirect_branch): Likewise.
15407 (ix86_output_indirect_jmp): Likewise.
15408 (ix86_code_end): Call output_indirect_thunk_function if needed.
15409 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15411 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15412 (ix86_attribute_table): Add indirect_branch.
15413 * config/i386/i386.h (machine_function): Add indirect_branch_type
15414 and has_local_indirect_jump.
15415 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15417 (tablejump): Likewise.
15418 (*indirect_jump): Use ix86_output_indirect_jmp.
15419 (*tablejump_1): Likewise.
15420 (simple_return_indirect_internal): Likewise.
15421 * config/i386/i386.opt (mindirect-branch=): New option.
15422 (indirect_branch): New.
15425 (thunk-inline): Likewise.
15426 (thunk-extern): Likewise.
15427 * doc/extend.texi: Document indirect_branch function attribute.
15428 * doc/invoke.texi: Document -mindirect-branch= option.
15430 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15433 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15435 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15437 * ipa-inline.c (want_inline_small_function_p): Return false if
15438 inlining has already failed with CIF_FINAL_ERROR.
15439 (update_caller_keys): Call want_inline_small_function_p before
15441 (update_callee_keys): Likewise.
15443 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15445 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15447 (rs6000_quadword_masked_address_p): Likewise.
15448 (quad_aligned_load_p): Likewise.
15449 (quad_aligned_store_p): Likewise.
15450 (const_load_sequence_p): Add comment to describe the outer-most loop.
15451 (mimic_memory_attributes_and_flags): New function.
15452 (rs6000_gen_stvx): Likewise.
15453 (replace_swapped_aligned_store): Likewise.
15454 (rs6000_gen_lvx): Likewise.
15455 (replace_swapped_aligned_load): Likewise.
15456 (replace_swapped_load_constant): Capitalize argument name in
15457 comment describing this function.
15458 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15459 and stores that access quad-word aligned addresses and replace
15460 with stvx or lvx instructions when appropriate.
15461 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15462 New function prototype.
15463 (rs6000_quadword_masked_address_p): Likewise.
15464 (rs6000_gen_lvx): Likewise.
15465 (rs6000_gen_stvx): Likewise.
15466 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15467 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15468 when memory address is aligned.
15469 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15470 this split to select lvx instruction when memory address is aligned.
15471 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15472 instruction when memory address is aligned.
15473 (*vsx_le_perm_load_v16qi): Likewise.
15474 (four unnamed splitters): Modify to select the stvx instruction
15475 when memory is aligned.
15477 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15479 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15480 which appears in the queue multiple times.
15482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15483 Alan Hayward <alan.hayward@arm.com>
15484 David Sherwood <david.sherwood@arm.com>
15486 * tree-vectorizer.h (vec_lower_bound): New structure.
15487 (_loop_vec_info): Add check_nonzero and lower_bounds.
15488 (LOOP_VINFO_CHECK_NONZERO): New macro.
15489 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15490 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15491 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15492 fields. Make seg_len the distance travelled, not including the
15494 (dr_direction_indicator): Declare.
15495 (dr_zero_step_indicator): Likewise.
15496 (dr_known_forward_stride_p): Likewise.
15497 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15499 (runtime_alias_check_p): Allow runtime alias checks with
15501 (operator ==): Compare access_size and align.
15502 (prune_runtime_alias_test_list): Rework for new distinction between
15503 the access_size and seg_len.
15504 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15506 (get_segment_min_max): New function.
15507 (create_intersect_range_checks): Use it.
15508 (dr_step_indicator): New function.
15509 (dr_direction_indicator): Likewise.
15510 (dr_zero_step_indicator): Likewise.
15511 (dr_known_forward_stride_p): Likewise.
15512 * tree-loop-distribution.c (data_ref_segment_size): Return
15513 DR_STEP * (niters - 1).
15514 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15516 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15517 (vect_preserves_scalar_order_p): New function, split out from...
15518 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15519 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15520 (vect_vfa_access_size): New function.
15521 (vect_vfa_align): Likewise.
15522 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15523 (dump_lower_bound): New function.
15524 (vect_check_lower_bound): Likewise.
15525 (vect_small_gap_p): Likewise.
15526 (vectorizable_with_step_bound_p): Likewise.
15527 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15528 depencies if the vectorization factor is 1. Convert the checks
15529 for nonzero steps into checks on the bounds of DR_STEP. Try using
15530 a bunds check for variable steps if the minimum required step is
15531 relatively small. Update calls to the dr_with_seg_len
15532 constructor and to vect_compile_time_alias.
15533 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15535 (vect_loop_versioning): Call it.
15536 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15538 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15541 Alan Hayward <alan.hayward@arm.com>
15542 David Sherwood <david.sherwood@arm.com>
15544 * doc/sourcebuild.texi (vect_scatter_store): Document.
15545 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15547 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15549 * genopinit.c (main): Add supports_vec_scatter_store and
15550 supports_vec_scatter_store_cached to target_optabs.
15551 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15552 IFN_MASK_SCATTER_STORE.
15553 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15555 * internal-fn.h (internal_store_fn_p): Declare.
15556 (internal_fn_stored_value_index): Likewise.
15557 * internal-fn.c (scatter_store_direct): New macro.
15558 (expand_scatter_store_optab_fn): New function.
15559 (direct_scatter_store_optab_supported_p): New macro.
15560 (internal_store_fn_p): New function.
15561 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15562 IFN_MASK_SCATTER_STORE.
15563 (internal_fn_mask_index): Likewise.
15564 (internal_fn_stored_value_index): New function.
15565 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15566 for scatter stores.
15567 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15568 * optabs-query.c (supports_vec_scatter_store_p): New function.
15569 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15570 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15571 true for scatter stores.
15572 (vect_gather_scatter_fn_p): Handle scatter stores too.
15573 (vect_check_gather_scatter): Consider using scatter stores if
15574 supports_vec_scatter_store_p.
15575 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15576 scatter stores too.
15577 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15578 internal_fn_stored_value_index.
15579 (check_load_store_masking): Handle scatter stores too.
15580 (vect_get_store_rhs): Make public.
15581 (vectorizable_call): Use internal_store_fn_p.
15582 (vectorizable_store): Handle scatter store internal functions.
15583 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15584 when deciding whether the end of the group has been reached.
15585 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15586 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15587 (mask_scatter_store<mode>): New insns.
15589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15590 Alan Hayward <alan.hayward@arm.com>
15591 David Sherwood <david.sherwood@arm.com>
15593 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15594 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15595 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15597 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15598 Use vect_truncate_gather_scatter_offset if we can't treat the
15599 operation as a normal gather load or scatter store.
15600 (get_group_load_store_type): Take the gather_scatter_info
15601 as argument. Try using a gather load or scatter store for
15602 single-element groups.
15603 (get_load_store_type): Update calls to get_group_load_store_type
15604 and vect_use_strided_gather_scatters_p.
15606 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15607 Alan Hayward <alan.hayward@arm.com>
15608 David Sherwood <david.sherwood@arm.com>
15610 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15611 optional tree argument.
15612 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15614 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15615 but continue to use the current value as a fallback.
15616 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15617 to compare the updates.
15618 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15619 (get_load_store_type): Use it when handling a strided access.
15620 (vect_get_strided_load_store_ops): New function.
15621 (vect_get_data_ptr_increment): Likewise.
15622 (vectorizable_load): Handle strided gather loads. Always pass
15623 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15626 Alan Hayward <alan.hayward@arm.com>
15627 David Sherwood <david.sherwood@arm.com>
15629 * doc/md.texi (gather_load@var{m}): Document.
15630 (mask_gather_load@var{m}): Likewise.
15631 * genopinit.c (main): Add supports_vec_gather_load and
15632 supports_vec_gather_load_cached to target_optabs.
15633 * optabs-tree.c (init_tree_optimization_optabs): Use
15634 ggc_cleared_alloc to allocate target_optabs.
15635 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15636 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15638 * internal-fn.h (internal_load_fn_p): Declare.
15639 (internal_gather_scatter_fn_p): Likewise.
15640 (internal_fn_mask_index): Likewise.
15641 (internal_gather_scatter_fn_supported_p): Likewise.
15642 * internal-fn.c (gather_load_direct): New macro.
15643 (expand_gather_load_optab_fn): New function.
15644 (direct_gather_load_optab_supported_p): New macro.
15645 (direct_internal_fn_optab): New function.
15646 (internal_load_fn_p): Likewise.
15647 (internal_gather_scatter_fn_p): Likewise.
15648 (internal_fn_mask_index): Likewise.
15649 (internal_gather_scatter_fn_supported_p): Likewise.
15650 * optabs-query.c (supports_at_least_one_mode_p): New function.
15651 (supports_vec_gather_load_p): Likewise.
15652 * optabs-query.h (supports_vec_gather_load_p): Declare.
15653 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15654 and memory_type field.
15655 (NUM_PATTERNS): Bump to 15.
15656 * tree-vect-data-refs.c: Include internal-fn.h.
15657 (vect_gather_scatter_fn_p): New function.
15658 (vect_describe_gather_scatter_call): Likewise.
15659 (vect_check_gather_scatter): Try using internal functions for
15660 gather loads. Recognize existing calls to a gather load function.
15661 (vect_analyze_data_refs): Consider using gather loads if
15662 supports_vec_gather_load_p.
15663 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15664 (vect_get_gather_scatter_offset_type): Likewise.
15665 (vect_convert_mask_for_vectype): Likewise.
15666 (vect_add_conversion_to_patterm): Likewise.
15667 (vect_try_gather_scatter_pattern): Likewise.
15668 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15669 (vect_vect_recog_func_ptrs): Add it.
15670 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15671 internal_fn_mask_index and internal_gather_scatter_fn_p.
15672 (check_load_store_masking): Take the gather_scatter_info as an
15673 argument and handle gather loads.
15674 (vect_get_gather_scatter_ops): New function.
15675 (vectorizable_call): Check internal_load_fn_p.
15676 (vectorizable_load): Likewise. Handle gather load internal
15678 (vectorizable_store): Update call to check_load_store_masking.
15679 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15680 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15681 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15682 (aarch64_gather_scale_operand_d): New predicates.
15683 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15684 (mask_gather_load<mode>): New insns.
15686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15687 Alan Hayward <alan.hayward@arm.com>
15688 David Sherwood <david.sherwood@arm.com>
15690 * optabs.def (fold_left_plus_optab): New optab.
15691 * doc/md.texi (fold_left_plus_@var{m}): Document.
15692 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15693 * internal-fn.c (fold_left_direct): Define.
15694 (expand_fold_left_optab_fn): Likewise.
15695 (direct_fold_left_optab_supported_p): Likewise.
15696 * fold-const-call.c (fold_const_fold_left): New function.
15697 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15698 * tree-parloops.c (valid_reduction_p): New function.
15699 (gather_scalar_reductions): Use it.
15700 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15701 (vect_finish_replace_stmt): Declare.
15702 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15703 (needs_fold_left_reduction_p): New function, split out from...
15704 (vect_is_simple_reduction): ...here. Accept reductions that
15705 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15706 (vect_force_simple_reduction): Also store the reduction type in
15707 the assignment's STMT_VINFO_REDUC_TYPE.
15708 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15709 (merge_with_identity): New function.
15710 (vect_expand_fold_left): Likewise.
15711 (vectorize_fold_left_reduction): Likewise.
15712 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15713 scalar phi in place for it. Check for target support and reject
15714 cases that would reassociate the operation. Defer the transform
15715 phase to vectorize_fold_left_reduction.
15716 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15717 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15718 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15722 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15723 call to ifc_temp_var.
15725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15726 Alan Hayward <alan.hayward@arm.com>
15727 David Sherwood <david.sherwood@arm.com>
15729 * target.def (legitimize_address_displacement): Take the original
15730 offset as a poly_int.
15731 * targhooks.h (default_legitimize_address_displacement): Update
15733 * targhooks.c (default_legitimize_address_displacement): Likewise.
15734 * doc/tm.texi: Regenerate.
15735 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15736 as an argument, moving assert of ad->disp == ad->disp_term to...
15737 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15738 Try calling targetm.legitimize_address_displacement before expanding
15739 the address rather than afterwards, and adjust for the new interface.
15740 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15741 Match the new hook interface. Handle SVE addresses.
15742 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15743 new hook interface.
15745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15747 * Makefile.in (OBJS): Add early-remat.o.
15748 * target.def (select_early_remat_modes): New hook.
15749 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15750 * doc/tm.texi: Regenerate.
15751 * targhooks.h (default_select_early_remat_modes): Declare.
15752 * targhooks.c (default_select_early_remat_modes): New function.
15753 * timevar.def (TV_EARLY_REMAT): New timevar.
15754 * passes.def (pass_early_remat): New pass.
15755 * tree-pass.h (make_pass_early_remat): Declare.
15756 * early-remat.c: New file.
15757 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15759 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15762 Alan Hayward <alan.hayward@arm.com>
15763 David Sherwood <david.sherwood@arm.com>
15765 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15766 vfm1 with a bound_epilog parameter.
15767 (vect_do_peeling): Update calls accordingly, and move the prologue
15768 call earlier in the function. Treat the base bound_epilog as 0 for
15769 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15770 this base when peeling for gaps.
15771 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15772 with fully-masked loops.
15773 (vect_estimate_min_profitable_iters): Handle the single peeled
15774 iteration in that case.
15776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15777 Alan Hayward <alan.hayward@arm.com>
15778 David Sherwood <david.sherwood@arm.com>
15780 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15781 single-element interleaving even if the size is not a power of 2.
15782 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15783 accesses for single-element interleaving if the group size is
15786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15787 Alan Hayward <alan.hayward@arm.com>
15788 David Sherwood <david.sherwood@arm.com>
15790 * doc/md.texi (fold_extract_last_@var{m}): Document.
15791 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15792 * optabs.def (fold_extract_last_optab): New optab.
15793 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15794 * internal-fn.c (fold_extract_direct): New macro.
15795 (expand_fold_extract_optab_fn): Likewise.
15796 (direct_fold_extract_optab_supported_p): Likewise.
15797 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15798 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15799 EXTRACT_LAST_REDUCTION.
15800 (get_initial_def_for_reduction): Do not create an initial vector
15801 for EXTRACT_LAST_REDUCTION reductions.
15802 (vectorizable_reduction): Leave the scalar phi in place for
15803 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15804 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15805 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15806 transform phase to vectorizable_condition.
15807 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15809 (vect_finish_stmt_generation): ...here.
15810 (vect_finish_replace_stmt): New function.
15811 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15812 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15814 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15817 Alan Hayward <alan.hayward@arm.com>
15818 David Sherwood <david.sherwood@arm.com>
15820 * doc/md.texi (extract_last_@var{m}): Document.
15821 * optabs.def (extract_last_optab): New optab.
15822 * internal-fn.def (EXTRACT_LAST): New internal function.
15823 * internal-fn.c (cond_unary_direct): New macro.
15824 (expand_cond_unary_optab_fn): Likewise.
15825 (direct_cond_unary_optab_supported_p): Likewise.
15826 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15827 loops using EXTRACT_LAST.
15828 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15829 (extract_last_<mode>): ...this optab.
15830 (vec_extract<mode><Vel>): Update accordingly.
15832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15833 Alan Hayward <alan.hayward@arm.com>
15834 David Sherwood <david.sherwood@arm.com>
15836 * target.def (empty_mask_is_expensive): New hook.
15837 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15838 * doc/tm.texi: Regenerate.
15839 * targhooks.h (default_empty_mask_is_expensive): Declare.
15840 * targhooks.c (default_empty_mask_is_expensive): New function.
15841 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15842 if the target says that empty masks are expensive.
15843 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15845 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15848 Alan Hayward <alan.hayward@arm.com>
15849 David Sherwood <david.sherwood@arm.com>
15851 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15852 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15853 (vect_use_loop_mask_for_alignment_p): New function.
15854 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15855 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15856 niters_skip argument. Make sure that the first niters_skip elements
15857 of the first iteration are inactive.
15858 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15859 Update call to vect_set_loop_masks_directly.
15860 (get_misalign_in_elems): New function, split out from...
15861 (vect_gen_prolog_loop_niters): ...here.
15862 (vect_update_init_of_dr): Take a code argument that specifies whether
15863 the adjustment should be added or subtracted.
15864 (vect_update_init_of_drs): Likewise.
15865 (vect_prepare_for_masked_peels): New function.
15866 (vect_do_peeling): Skip prologue peeling if we're using a mask
15867 instead. Update call to vect_update_inits_of_drs.
15868 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15870 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15871 alignment. Do not include the number of peeled iterations in
15872 the minimum threshold in that case.
15873 (vectorizable_induction): Adjust the start value down by
15874 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15875 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15876 Take the number of skipped iterations into account when calculating
15878 * tree-vect-stmts.c (vect_gen_while_not): New function.
15880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15881 Alan Hayward <alan.hayward@arm.com>
15882 David Sherwood <david.sherwood@arm.com>
15884 * doc/sourcebuild.texi (vect_fully_masked): Document.
15885 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15886 default value to 0.
15887 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15889 (vect_analyze_loop_2): ...here. Don't check the vectorization
15890 factor against the number of loop iterations if the loop is
15893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15894 Alan Hayward <alan.hayward@arm.com>
15895 David Sherwood <david.sherwood@arm.com>
15897 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15898 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15899 (dump_groups): Update accordingly.
15900 (iv_use::mem_type): New member variable.
15901 (address_p): New function.
15902 (record_use): Add a mem_type argument and initialize the new
15904 (record_group_use): Add a mem_type argument. Use address_p.
15905 Remove obsolete null checks of base_object. Update call to record_use.
15906 (find_interesting_uses_op): Update call to record_group_use.
15907 (find_interesting_uses_cond): Likewise.
15908 (find_interesting_uses_address): Likewise.
15909 (get_mem_type_for_internal_fn): New function.
15910 (find_address_like_use): Likewise.
15911 (find_interesting_uses_stmt): Try find_address_like_use before
15912 calling find_interesting_uses_op.
15913 (addr_offset_valid_p): Use the iv mem_type field as the type
15914 of the addressed memory.
15915 (add_autoinc_candidates): Likewise.
15916 (get_address_cost): Likewise.
15917 (split_small_address_groups_p): Use address_p.
15918 (split_address_groups): Likewise.
15919 (add_iv_candidate_for_use): Likewise.
15920 (autoinc_possible_for_pair): Likewise.
15921 (rewrite_groups): Likewise.
15922 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15923 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15924 (get_alias_ptr_type_for_ptr_address): New function.
15925 (rewrite_use_address): Rewrite address uses in calls that were
15926 identified by find_address_like_use.
15928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15929 Alan Hayward <alan.hayward@arm.com>
15930 David Sherwood <david.sherwood@arm.com>
15932 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15934 * gimple-expr.h (is_gimple_addressable: Likewise.
15935 * gimple-expr.c (is_gimple_address): Likewise.
15936 * internal-fn.c (expand_call_mem_ref): New function.
15937 (expand_mask_load_optab_fn): Use it.
15938 (expand_mask_store_optab_fn): Likewise.
15940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15941 Alan Hayward <alan.hayward@arm.com>
15942 David Sherwood <david.sherwood@arm.com>
15944 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15945 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15946 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15947 (cond_umax@var{mode}): Document.
15948 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15949 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15950 (cond_umin_optab, cond_umax_optab): New optabs.
15951 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15952 (COND_IOR, COND_XOR): New internal functions.
15953 * internal-fn.h (get_conditional_internal_fn): Declare.
15954 * internal-fn.c (cond_binary_direct): New macro.
15955 (expand_cond_binary_optab_fn): Likewise.
15956 (direct_cond_binary_optab_supported_p): Likewise.
15957 (get_conditional_internal_fn): New function.
15958 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15959 Cope with reduction statements that are vectorized as calls rather
15961 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15962 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15963 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15964 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15965 (UNSPEC_COND_EOR): New unspecs.
15966 (optab): Add mappings for them.
15967 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15968 (sve_int_op, sve_fp_op): New int attributes.
15970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15971 Alan Hayward <alan.hayward@arm.com>
15972 David Sherwood <david.sherwood@arm.com>
15974 * optabs.def (while_ult_optab): New optab.
15975 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15976 * internal-fn.def (WHILE_ULT): New internal function.
15977 * internal-fn.h (direct_internal_fn_supported_p): New override
15978 that takes two types as argument.
15979 * internal-fn.c (while_direct): New macro.
15980 (expand_while_optab_fn): New function.
15981 (convert_optab_supported_p): Likewise.
15982 (direct_while_optab_supported_p): New macro.
15983 * wide-int.h (wi::udiv_ceil): New function.
15984 * tree-vectorizer.h (rgroup_masks): New structure.
15985 (vec_loop_masks): New typedef.
15986 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15987 and fully_masked_p.
15988 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15989 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15990 (vect_max_vf): New function.
15991 (slpeel_make_loop_iterate_ntimes): Delete.
15992 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15993 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15994 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15995 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15996 internal-fn.h, stor-layout.h and optabs-query.h.
15997 (vect_set_loop_mask): New function.
15998 (add_preheader_seq): Likewise.
15999 (add_header_seq): Likewise.
16000 (interleave_supported_p): Likewise.
16001 (vect_maybe_permute_loop_masks): Likewise.
16002 (vect_set_loop_masks_directly): Likewise.
16003 (vect_set_loop_condition_masked): Likewise.
16004 (vect_set_loop_condition_unmasked): New function, split out from
16005 slpeel_make_loop_iterate_ntimes.
16006 (slpeel_make_loop_iterate_ntimes): Rename to..
16007 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16008 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16009 (vect_do_peeling): Update call accordingly.
16010 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16012 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16013 mask_compare_type, can_fully_mask_p and fully_masked_p.
16014 (release_vec_loop_masks): New function.
16015 (_loop_vec_info): Use it to free the loop masks.
16016 (can_produce_all_loop_masks_p): New function.
16017 (vect_get_max_nscalars_per_iter): Likewise.
16018 (vect_verify_full_masking): Likewise.
16019 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16020 retries, and free the mask rgroups before retrying. Check loop-wide
16021 reasons for disallowing fully-masked loops. Make the final decision
16022 about whether use a fully-masked loop or not.
16023 (vect_estimate_min_profitable_iters): Do not assume that peeling
16024 for the number of iterations will be needed for fully-masked loops.
16025 (vectorizable_reduction): Disable fully-masked loops.
16026 (vectorizable_live_operation): Likewise.
16027 (vect_halve_mask_nunits): New function.
16028 (vect_double_mask_nunits): Likewise.
16029 (vect_record_loop_mask): Likewise.
16030 (vect_get_loop_mask): Likewise.
16031 (vect_transform_loop): Handle the case in which the final loop
16032 iteration might handle a partial vector. Call vect_set_loop_condition
16033 instead of slpeel_make_loop_iterate_ntimes.
16034 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16035 (check_load_store_masking): New function.
16036 (prepare_load_store_mask): Likewise.
16037 (vectorizable_store): Handle fully-masked loops.
16038 (vectorizable_load): Likewise.
16039 (supportable_widening_operation): Use vect_halve_mask_nunits for
16041 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16042 (vect_gen_while): New function.
16043 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16044 (aarch64_uqdec<mode>): New insn.
16046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16047 Alan Hayward <alan.hayward@arm.com>
16048 David Sherwood <david.sherwood@arm.com>
16050 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16051 (reduc_xor_scal_optab): New optabs.
16052 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16053 (reduc_xor_scal_@var{m}): Document.
16054 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16055 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16056 internal functions.
16057 * fold-const-call.c (fold_const_call): Handle them.
16058 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16059 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16060 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16061 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16062 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16063 (UNSPEC_XORV): New unspecs.
16064 (optab): Add entries for them.
16065 (BITWISEV): New int iterator.
16066 (bit_reduc_op): New int attributes.
16068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16069 Alan Hayward <alan.hayward@arm.com>
16070 David Sherwood <david.sherwood@arm.com>
16072 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16073 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16074 * optabs.def (vec_shl_insert_optab): New optab.
16075 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16076 (duplicate_and_interleave): Likewise.
16077 * tree-vect-loop.c: Include internal-fn.h.
16078 (neutral_op_for_slp_reduction): New function, split out from
16079 get_initial_defs_for_reduction.
16080 (get_initial_def_for_reduction): Handle option 2 for variable-length
16081 vectors by loading the neutral value into a vector and then shifting
16082 the initial value into element 0.
16083 (get_initial_defs_for_reduction): Replace the code argument with
16084 the neutral value calculated by neutral_op_for_slp_reduction.
16085 Use gimple_build_vector for constant-length vectors.
16086 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16087 but the first group_size elements have a neutral value.
16088 Use duplicate_and_interleave otherwise.
16089 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16090 Update call to get_initial_defs_for_reduction. Handle SLP
16091 reductions for variable-length vectors by creating one vector
16092 result for each scalar result, with the elements associated
16093 with other scalar results stubbed out with the neutral value.
16094 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16095 Require IFN_VEC_SHL_INSERT for double reductions on
16096 variable-length vectors, or SLP reductions that have
16097 a neutral value. Require can_duplicate_and_interleave_p
16098 support for variable-length unchained SLP reductions if there
16099 is no neutral value, such as for MIN/MAX reductions. Also require
16100 the number of vector elements to be a multiple of the number of
16101 SLP statements when doing variable-length unchained SLP reductions.
16102 Update call to vect_create_epilog_for_reduction.
16103 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16104 and remove initial values.
16105 (duplicate_and_interleave): Make public.
16106 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16107 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16110 Alan Hayward <alan.hayward@arm.com>
16111 David Sherwood <david.sherwood@arm.com>
16113 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16114 (can_duplicate_and_interleave_p): New function.
16115 (vect_get_and_check_slp_defs): Take the vector of statements
16116 rather than just the current one. Remove excess parentheses.
16117 Restriction rejectinon of vect_constant_def and vect_external_def
16118 for variable-length vectors to boolean types, or types for which
16119 can_duplicate_and_interleave_p is false.
16120 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16121 (duplicate_and_interleave): New function.
16122 (vect_get_constant_vectors): Use gimple_build_vector for
16123 constant-length vectors and suitable variable-length constant
16124 vectors. Use duplicate_and_interleave for other variable-length
16125 vectors. Don't defer the update when inserting new statements.
16127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16128 Alan Hayward <alan.hayward@arm.com>
16129 David Sherwood <david.sherwood@arm.com>
16131 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16132 min_profitable_iters doesn't go negative.
16134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16135 Alan Hayward <alan.hayward@arm.com>
16136 David Sherwood <david.sherwood@arm.com>
16138 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16139 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16140 * optabs.def (vec_mask_load_lanes_optab): New optab.
16141 (vec_mask_store_lanes_optab): Likewise.
16142 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16143 (MASK_STORE_LANES): Likewise.
16144 * internal-fn.c (mask_load_lanes_direct): New macro.
16145 (mask_store_lanes_direct): Likewise.
16146 (expand_mask_load_optab_fn): Handle masked operations.
16147 (expand_mask_load_lanes_optab_fn): New macro.
16148 (expand_mask_store_optab_fn): Handle masked operations.
16149 (expand_mask_store_lanes_optab_fn): New macro.
16150 (direct_mask_load_lanes_optab_supported_p): Likewise.
16151 (direct_mask_store_lanes_optab_supported_p): Likewise.
16152 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16154 (vect_load_lanes_supported): Likewise.
16155 * tree-vect-data-refs.c (strip_conversion): New function.
16156 (can_group_stmts_p): Likewise.
16157 (vect_analyze_data_ref_accesses): Use it instead of checking
16158 for a pair of assignments.
16159 (vect_store_lanes_supported): Take a masked_p parameter.
16160 (vect_load_lanes_supported): Likewise.
16161 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16162 vect_store_lanes_supported and vect_load_lanes_supported.
16163 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16164 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16165 parameter. Don't allow gaps for masked accesses.
16166 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16167 and vect_load_lanes_supported.
16168 (get_load_store_type): Take a masked_p parameter and update
16169 call to get_group_load_store_type.
16170 (vectorizable_store): Update call to get_load_store_type.
16171 Handle IFN_MASK_STORE_LANES.
16172 (vectorizable_load): Update call to get_load_store_type.
16173 Handle IFN_MASK_LOAD_LANES.
16175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16176 Alan Hayward <alan.hayward@arm.com>
16177 David Sherwood <david.sherwood@arm.com>
16179 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16181 * config/aarch64/aarch64-protos.h
16182 (aarch64_sve_struct_memory_operand_p): Declare.
16183 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16184 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16185 (VPRED, vpred): Handle SVE structure modes.
16186 * config/aarch64/constraints.md (Utx): New constraint.
16187 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16188 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16189 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16190 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16191 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16192 structure modes. Split into pieces after RA.
16193 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16194 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16196 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16197 SVE structure modes.
16198 (aarch64_classify_address): Likewise.
16199 (sizetochar): Move earlier in file.
16200 (aarch64_print_operand): Handle SVE register lists.
16201 (aarch64_array_mode): New function.
16202 (aarch64_sve_struct_memory_operand_p): Likewise.
16203 (TARGET_ARRAY_MODE): Redefine.
16205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16206 Alan Hayward <alan.hayward@arm.com>
16207 David Sherwood <david.sherwood@arm.com>
16209 * target.def (array_mode): New target hook.
16210 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16211 * doc/tm.texi: Regenerate.
16212 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16213 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16214 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16215 targetm.array_mode.
16216 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16220 Alan Hayward <alan.hayward@arm.com>
16221 David Sherwood <david.sherwood@arm.com>
16223 * fold-const.c (fold_binary_loc): Check the argument types
16224 rather than the result type when testing for a vector operation.
16226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16228 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16229 * doc/tm.texi: Regenerate.
16231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16232 Alan Hayward <alan.hayward@arm.com>
16233 David Sherwood <david.sherwood@arm.com>
16235 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16236 (sve): Document new AArch64 extension.
16237 * doc/md.texi (w): Extend the description of the AArch64
16238 constraint to include SVE vectors.
16239 (Upl, Upa): Document new AArch64 predicate constraints.
16240 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16242 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16243 (msve-vector-bits=): New option.
16244 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16245 SVE when these are disabled.
16246 (sve): New extension.
16247 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16248 modes. Adjust their number of units based on aarch64_sve_vg.
16249 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16250 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16251 aarch64_addr_query_type.
16252 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16253 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16254 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16255 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16256 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16257 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16258 (aarch64_simd_imm_zero_p): Delete.
16259 (aarch64_check_zero_based_sve_index_immediate): Declare.
16260 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16261 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16262 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16263 (aarch64_sve_float_mul_immediate_p): Likewise.
16264 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16265 rather than an rtx.
16266 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16267 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16268 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16269 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16270 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16271 (aarch64_regmode_natural_size): Likewise.
16272 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16273 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16275 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16276 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16277 for VG and the SVE predicate registers.
16278 (V_ALIASES): Add a "z"-prefixed alias.
16279 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16280 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16281 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16282 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16283 (REG_CLASS_NAMES): Add entries for them.
16284 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16285 and the predicate registers.
16286 (aarch64_sve_vg): Declare.
16287 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16288 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16289 (REGMODE_NATURAL_SIZE): Define.
16290 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16292 * config/aarch64/aarch64.c: Include cfgrtl.h.
16293 (simd_immediate_info): Add a constructor for series vectors,
16294 and an associated step field.
16295 (aarch64_sve_vg): New variable.
16296 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16297 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16298 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16299 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16300 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16301 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16302 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16303 (aarch64_get_mask_mode): New functions.
16304 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16305 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16306 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16307 predicate modes and predicate registers. Explicitly restrict
16308 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16309 to store a vector mode if it is recognized by
16310 aarch64_classify_vector_mode.
16311 (aarch64_regmode_natural_size): New function.
16312 (aarch64_hard_regno_caller_save_mode): Return the original mode
16314 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16315 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16316 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16317 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16319 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16320 does not overlap dest if the function is frame-related. Handle
16322 (aarch64_split_add_offset): New function.
16323 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16324 them aarch64_add_offset.
16325 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16326 and update call to aarch64_sub_sp.
16327 (aarch64_add_cfa_expression): New function.
16328 (aarch64_expand_prologue): Pass extra temporary registers to the
16329 functions above. Handle the case in which we need to emit new
16330 DW_CFA_expressions for registers that were originally saved
16331 relative to the stack pointer, but now have to be expressed
16332 relative to the frame pointer.
16333 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16335 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16336 IP0 and IP1 values for SVE frames.
16337 (aarch64_expand_vec_series): New function.
16338 (aarch64_expand_sve_widened_duplicate): Likewise.
16339 (aarch64_expand_sve_const_vector): Likewise.
16340 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16341 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16342 into the register, rather than emitting a SET directly.
16343 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16344 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16345 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16346 (offset_9bit_signed_scaled_p): New functions.
16347 (aarch64_replicate_bitmask_imm): New function.
16348 (aarch64_bitmask_imm): Use it.
16349 (aarch64_cannot_force_const_mem): Reject expressions involving
16350 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16351 (aarch64_classify_index): Handle SVE indices, by requiring
16352 a plain register index with a scale that matches the element size.
16353 (aarch64_classify_address): Handle SVE addresses. Assert that
16354 the mode of the address is VOIDmode or an integer mode.
16355 Update call to aarch64_classify_symbol.
16356 (aarch64_classify_symbolic_expression): Update call to
16357 aarch64_classify_symbol.
16358 (aarch64_const_vec_all_in_range_p): New function.
16359 (aarch64_print_vector_float_operand): Likewise.
16360 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16361 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16362 and the FP immediates 1.0 and 0.5.
16363 (aarch64_print_address_internal): Handle SVE addresses.
16364 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16365 (aarch64_regno_regclass): Handle predicate registers.
16366 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16368 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16369 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16370 (aarch64_convert_sve_vector_bits): New function.
16371 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16372 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16373 rather than an rtx.
16374 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16375 Handle SVE vector and predicate modes. Accept VL-based constants
16376 that need only one temporary register, and VL offsets that require
16377 no temporary registers.
16378 (aarch64_conditional_register_usage): Mark the predicate registers
16379 as fixed if SVE isn't available.
16380 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16381 Return true for SVE vector and predicate modes.
16382 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16383 rather than an unsigned int. Handle SVE modes.
16384 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16386 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16388 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16389 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16390 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16391 (aarch64_sve_float_mul_immediate_p): New functions.
16392 (aarch64_sve_valid_immediate): New function.
16393 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16394 Explicitly reject structure modes. Check for INDEX constants.
16395 Handle PTRUE and PFALSE constants.
16396 (aarch64_check_zero_based_sve_index_immediate): New function.
16397 (aarch64_simd_imm_zero_p): Delete.
16398 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16399 vector modes. Accept constants in the range of CNT[BHWD].
16400 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16401 ask for an Advanced SIMD mode.
16402 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16403 (aarch64_simd_vector_alignment): Handle SVE predicates.
16404 (aarch64_vectorize_preferred_vector_alignment): New function.
16405 (aarch64_simd_vector_alignment_reachable): Use it instead of
16407 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16408 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16410 (MAX_VECT_LEN): Delete.
16411 (expand_vec_perm_d): Add a vec_flags field.
16412 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16413 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16414 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16416 (aarch64_evpc_rev): Rename to...
16417 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16418 (aarch64_evpc_rev_global): New function.
16419 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16420 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16422 (aarch64_evpc_sve_tbl): New function.
16423 (aarch64_expand_vec_perm_const_1): Update after rename of
16424 aarch64_evpc_rev. Handle SVE permutes too, trying
16425 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16426 than aarch64_evpc_tbl.
16427 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16428 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16429 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16430 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16431 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16432 (aarch64_expand_sve_vcond): New functions.
16433 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16434 of aarch64_vector_mode_p.
16435 (aarch64_dwarf_poly_indeterminate_value): New function.
16436 (aarch64_compute_pressure_classes): Likewise.
16437 (aarch64_can_change_mode_class): Likewise.
16438 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16439 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16440 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16441 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16442 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16443 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16444 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16445 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16447 (Dn, Dl, Dr): Accept const as well as const_vector.
16448 (Dz): Likewise. Compare against CONST0_RTX.
16449 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16450 of "vector" where appropriate.
16451 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16452 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16453 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16454 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16455 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16456 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16457 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16458 (v_int_equiv): Extend to SVE modes.
16459 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16461 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16462 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16463 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16464 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16465 (SVE_COND_FP_CMP): New int iterators.
16466 (perm_hilo): Handle the new unpack unspecs.
16467 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16469 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16470 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16471 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16472 (aarch64_equality_operator, aarch64_constant_vector_operand)
16473 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16474 (aarch64_sve_nonimmediate_operand): Likewise.
16475 (aarch64_sve_general_operand): Likewise.
16476 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16477 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16478 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16479 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16480 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16481 (aarch64_sve_float_arith_immediate): Likewise.
16482 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16483 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16484 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16485 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16486 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16487 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16488 (aarch64_sve_float_arith_operand): Likewise.
16489 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16490 (aarch64_sve_float_mul_operand): Likewise.
16491 (aarch64_sve_vec_perm_operand): Likewise.
16492 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16493 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16494 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16495 as well as const_vector.
16496 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16497 in file. Use CONST0_RTX and CONSTM1_RTX.
16498 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16499 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16500 Use aarch64_simd_imm_zero.
16501 * config/aarch64/aarch64-sve.md: New file.
16502 * config/aarch64/aarch64.md: Include it.
16503 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16504 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16505 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16506 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16507 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16508 (sve): New attribute.
16509 (enabled): Disable instructions with the sve attribute unless
16511 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16512 aarch64_expand_mov_immediate.
16513 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16514 CNT[BHSD] immediates.
16515 (movti): Split CONST_POLY_INT moves into two halves.
16516 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16517 Split additions that need a temporary here if the destination
16518 is the stack pointer.
16519 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16520 (*add<mode>3_poly_1): New instruction.
16521 (set_clobber_cc): New expander.
16523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16525 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16526 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16527 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16528 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16529 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16530 Change innermode from fixed_mode_size to machine_mode.
16531 (simplify_subreg): Update call accordingly. Handle a constant-sized
16532 subreg of a variable-length CONST_VECTOR.
16534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16535 Alan Hayward <alan.hayward@arm.com>
16536 David Sherwood <david.sherwood@arm.com>
16538 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16539 (add_offset_to_base): New function, split out from...
16540 (create_mem_ref): ...here. When handling a scale other than 1,
16541 check first whether the address is valid without the offset.
16542 Add it into the base if so, leaving the index and scale as-is.
16544 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16547 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16548 fold_for_warn before checking if arg2 is INTEGER_CST.
16550 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16552 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16553 (store_multiple_operation): Delete.
16554 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16555 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16556 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16557 guarded by TARGET_STRING.
16558 (rs6000_output_load_multiple): Delete.
16559 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16560 OPTION_MASK_STRING / TARGET_STRING handling.
16561 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16562 (const rs6000_opt_masks) <"string">: Change mask to 0.
16563 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16564 (MASK_STRING): Delete.
16565 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16567 (load_multiple): Delete.
16574 (store_multiple): Delete.
16581 (movmemsi_8reg): Delete.
16582 (corresponding unnamed define_insn): Delete.
16583 (movmemsi_6reg): Delete.
16584 (corresponding unnamed define_insn): Delete.
16585 (movmemsi_4reg): Delete.
16586 (corresponding unnamed define_insn): Delete.
16587 (movmemsi_2reg): Delete.
16588 (corresponding unnamed define_insn): Delete.
16589 (movmemsi_1reg): Delete.
16590 (corresponding unnamed define_insn): Delete.
16591 * config/rs6000/rs6000.opt (mno-string): New.
16592 (mstring): Replace by deprecation warning stub.
16593 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16595 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16597 * regrename.c (regrename_do_replace): If replacing the same
16598 reg multiple times, try to reuse last created gen_raw_REG.
16601 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16602 main to workaround a bug in GDB.
16604 2018-01-12 Tom de Vries <tom@codesourcery.com>
16607 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16609 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16611 PR rtl-optimization/80481
16612 * ira-color.c (get_cap_member): New function.
16613 (allocnos_conflict_by_live_ranges_p): Use it.
16614 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16615 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16617 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16620 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16621 (*saddl_se_1): Ditto.
16622 (*ssubsi_1): Ditto.
16623 (*ssubl_se_1): Ditto.
16625 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16627 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16628 rather than wi::to_widest for DR_INITs.
16629 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16630 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16631 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16633 (vect_analyze_group_access_1): Note that here.
16635 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16637 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16638 polynomial type sizes.
16640 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16642 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16643 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16644 (gimple_add_tmp_var): Likewise.
16646 2018-01-12 Martin Liska <mliska@suse.cz>
16648 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16649 (gimple_alloc_sizes): Likewise.
16650 (dump_gimple_statistics): Use PRIu64 in printf format.
16651 * gimple.h: Change uint64_t to int.
16653 2018-01-12 Martin Liska <mliska@suse.cz>
16655 * tree-core.h: Use uint64_t instead of int.
16656 * tree.c (tree_node_counts): Likewise.
16657 (tree_node_sizes): Likewise.
16658 (dump_tree_statistics): Use PRIu64 in printf format.
16660 2018-01-12 Martin Liska <mliska@suse.cz>
16662 * Makefile.in: As qsort_chk is implemented in vec.c, add
16663 vec.o to linkage of gencfn-macros.
16664 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16665 passing the info to record_node_allocation_statistics.
16666 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16668 * ggc-common.c (struct ggc_usage): Add operator== and use
16669 it in operator< and compare function.
16670 * mem-stats.h (struct mem_usage): Likewise.
16671 * vec.c (struct vec_usage): Remove operator< and compare
16672 function. Can be simply inherited.
16674 2018-01-12 Martin Jambor <mjambor@suse.cz>
16677 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16678 * tree-ssa-math-opts.c: Include domwalk.h.
16679 (convert_mult_to_fma_1): New function.
16680 (fma_transformation_info): New type.
16681 (fma_deferring_state): Likewise.
16682 (cancel_fma_deferring): New function.
16683 (result_of_phi): Likewise.
16684 (last_fma_candidate_feeds_initial_phi): Likewise.
16685 (convert_mult_to_fma): Added deferring logic, split actual
16686 transformation to convert_mult_to_fma_1.
16687 (math_opts_dom_walker): New type.
16688 (math_opts_dom_walker::after_dom_children): New method, body moved
16689 here from pass_optimize_widening_mul::execute, added deferring logic
16691 (pass_optimize_widening_mul::execute): Moved most of code to
16692 math_opts_dom_walker::after_dom_children.
16693 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16694 * config/i386/i386.c (ix86_option_override_internal): Added
16695 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16697 2018-01-12 Richard Biener <rguenther@suse.de>
16700 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16701 inline instance vars.
16703 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16706 * config/rx/rx.c (rx_is_restricted_memory_address):
16707 Handle SUBREG case.
16709 2018-01-12 Richard Biener <rguenther@suse.de>
16711 PR tree-optimization/80846
16712 * target.def (split_reduction): New target hook.
16713 * targhooks.c (default_split_reduction): New function.
16714 * targhooks.h (default_split_reduction): Declare.
16715 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16716 target requests first reduce vectors by combining low and high
16718 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16719 (get_vectype_for_scalar_type_and_size): Export.
16720 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16721 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16722 * doc/tm.texi: Regenerate.
16723 * config/i386/i386.c (ix86_split_reduction): Implement
16724 TARGET_VECTORIZE_SPLIT_REDUCTION.
16726 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16729 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16730 in PIC mode except for TARGET_VXWORKS_RTP.
16731 * config/sparc/sparc.c: Include cfgrtl.h.
16732 (TARGET_INIT_PIC_REG): Define.
16733 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16734 (sparc_pic_register_p): New predicate.
16735 (sparc_legitimate_address_p): Use it.
16736 (sparc_legitimize_pic_address): Likewise.
16737 (sparc_delegitimize_address): Likewise.
16738 (sparc_mode_dependent_address_p): Likewise.
16739 (gen_load_pcrel_sym): Remove 4th parameter.
16740 (load_got_register): Adjust call to above. Remove obsolete stuff.
16741 (sparc_expand_prologue): Do not call load_got_register here.
16742 (sparc_flat_expand_prologue): Likewise.
16743 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16744 (sparc_use_pseudo_pic_reg): New function.
16745 (sparc_init_pic_reg): Likewise.
16746 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16747 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16749 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16751 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16752 Add item for branch_cost.
16754 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16756 PR rtl-optimization/83565
16757 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16758 not extend the result to a larger mode for rotate operations.
16759 (num_sign_bit_copies1): Likewise.
16761 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16764 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16766 Use values-Xc.o for -pedantic.
16767 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16769 2018-01-12 Martin Liska <mliska@suse.cz>
16772 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16774 (possible_polymorphic_call_targets): Use it.
16775 (ipa_devirt): Likewise.
16777 2018-01-12 Martin Liska <mliska@suse.cz>
16779 * profile-count.h (enum profile_quality): Use 0 as invalid
16780 enum value of profile_quality.
16782 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16784 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16785 -mext-string options.
16787 2018-01-12 Richard Biener <rguenther@suse.de>
16789 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16790 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16791 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16793 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16795 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16797 * configure.ac (--with-long-double-format): Add support for the
16798 configuration option to change the default long double format on
16800 * config.gcc (powerpc*-linux*-*): Likewise.
16801 * configure: Regenerate.
16802 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16803 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16804 used without modification.
16806 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16808 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16809 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16810 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16811 MISC_BUILTIN_SPEC_BARRIER.
16812 (rs6000_init_builtins): Likewise.
16813 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16815 (speculation_barrier): New define_insn.
16816 * doc/extend.texi: Document __builtin_speculation_barrier.
16818 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16821 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16822 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16823 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16825 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16826 integral modes instead of "ss" and "sd".
16827 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16828 vectors with 32-bit and 64-bit elements.
16829 (vecdupssescalarmodesuffix): New mode attribute.
16830 (vec_dup<mode>): Use it.
16832 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16835 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16836 frame if argument is passed on stack.
16838 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16841 * ree.c (combine_reaching_defs): Optimize also
16842 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16843 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16845 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16847 PR middle-end/83189
16848 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16850 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16852 PR middle-end/83718
16853 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16854 after they are computed.
16856 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16858 PR tree-optimization/83695
16859 * gimple-loop-linterchange.cc
16860 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16861 reset cached scev information after interchange.
16862 (pass_linterchange::execute): Remove call to scev_reset_htab.
16864 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16866 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16867 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16868 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16869 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16870 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16871 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16872 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16873 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16874 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16875 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16876 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16877 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16878 (V_lane_reg): Likewise.
16879 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16881 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16882 (vfmal_lane_low<mode>_intrinsic,
16883 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16884 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16885 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16886 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16887 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16888 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16890 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16892 * config/arm/arm-cpus.in (fp16fml): New feature.
16893 (ALL_SIMD): Add fp16fml.
16894 (armv8.2-a): Add fp16fml as an option.
16895 (armv8.3-a): Likewise.
16896 (armv8.4-a): Add fp16fml as part of fp16.
16897 * config/arm/arm.h (TARGET_FP16FML): Define.
16898 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16900 * config/arm/arm-modes.def (V2HF): Define.
16901 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16902 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16903 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16904 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16905 vfmsl_low, vfmsl_high): New set of builtins.
16906 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16907 (vfml_op): New code attribute.
16908 (VFMLHALVES): New int iterator.
16909 (VFML, VFMLSEL): New mode attributes.
16910 (V_reg): Define mapping for V2HF.
16911 (V_hi, V_lo): New mode attributes.
16912 (VF_constraint): Likewise.
16913 (vfml_half, vfml_half_selector): New int attributes.
16914 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16916 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16917 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16919 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16920 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16921 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16922 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16924 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16925 Document new effective target and option set.
16927 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16929 * config/arm/arm-cpus.in (armv8_4): New feature.
16930 (ARMv8_4a): New fgroup.
16931 (armv8.4-a): New arch.
16932 * config/arm/arm-tables.opt: Regenerate.
16933 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16934 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16935 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16936 Add matching rules for -march=armv8.4-a and extensions.
16937 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16939 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16942 * config/rx/rx.md (BW): New mode attribute.
16943 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16945 2018-01-11 Richard Biener <rguenther@suse.de>
16947 PR tree-optimization/83435
16948 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16949 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16950 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16952 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16953 Alan Hayward <alan.hayward@arm.com>
16954 David Sherwood <david.sherwood@arm.com>
16956 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16958 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16959 (aarch64_print_address_internal): Use it to check for a zero offset.
16961 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16962 Alan Hayward <alan.hayward@arm.com>
16963 David Sherwood <david.sherwood@arm.com>
16965 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16966 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16967 Return a poly_int64 rather than a HOST_WIDE_INT.
16968 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16969 rather than a HOST_WIDE_INT.
16970 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16971 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16972 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16973 final_offset from HOST_WIDE_INT to poly_int64.
16974 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16975 to_constant when getting the number of units in an Advanced SIMD
16977 (aarch64_builtin_vectorized_function): Check for a constant number
16979 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16981 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16982 attribute instead of GET_MODE_NUNITS.
16983 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16984 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16985 GET_MODE_SIZE for fixed-size registers.
16986 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16987 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16988 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16989 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16990 (aarch64_print_operand, aarch64_print_address_internal)
16991 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16992 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16993 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16994 Handle polynomial GET_MODE_SIZE.
16995 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16996 wider than SImode without modification.
16997 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16998 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16999 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17000 passing and returning SVE modes.
17001 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17002 rather than GEN_INT.
17003 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17004 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17005 (aarch64_allocate_and_probe_stack_space): Likewise.
17006 (aarch64_layout_frame): Cope with polynomial offsets.
17007 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17008 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17009 polynomial offsets.
17010 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17011 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17012 poly_int64 rather than a HOST_WIDE_INT.
17013 (aarch64_get_separate_components, aarch64_process_components)
17014 (aarch64_expand_prologue, aarch64_expand_epilogue)
17015 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17016 (aarch64_anchor_offset): New function, split out from...
17017 (aarch64_legitimize_address): ...here.
17018 (aarch64_builtin_vectorization_cost): Handle polynomial
17019 TYPE_VECTOR_SUBPARTS.
17020 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17022 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17023 number of elements from the PARALLEL rather than the mode.
17024 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17025 rather than GET_MODE_BITSIZE.
17026 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17027 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17028 (aarch64_expand_vec_perm_const_1): Handle polynomial
17029 d->perm.length () and d->perm elements.
17030 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17031 Apply to_constant to d->perm elements.
17032 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17033 polynomial CONST_VECTOR_NUNITS.
17034 (aarch64_move_pointer): Take amount as a poly_int64 rather
17036 (aarch64_progress_pointer): Avoid temporary variable.
17037 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17038 the mode attribute instead of GET_MODE.
17040 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17041 Alan Hayward <alan.hayward@arm.com>
17042 David Sherwood <david.sherwood@arm.com>
17044 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17045 x exists before using it.
17046 (aarch64_add_constant_internal): Rename to...
17047 (aarch64_add_offset_1): ...this. Replace regnum with separate
17048 src and dest rtxes. Handle the case in which they're different,
17049 including when the offset is zero. Replace scratchreg with an rtx.
17050 Use 2 additions if there is no spare register into which we can
17051 move a 16-bit constant.
17052 (aarch64_add_constant): Delete.
17053 (aarch64_add_offset): Replace reg with separate src and dest
17054 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17055 Use aarch64_add_offset_1.
17056 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17057 an rtx rather than an int. Take the delta as a poly_int64
17058 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17059 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17060 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17061 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17062 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17063 and aarch64_add_sp.
17064 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17065 aarch64_add_constant.
17067 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17069 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17070 Use scalar_float_mode.
17072 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17074 * config/aarch64/aarch64-simd.md
17075 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17076 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17077 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17078 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17079 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17080 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17081 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17082 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17083 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17084 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17086 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17089 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17090 targ_options->x_arm_arch_string is non NULL.
17092 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17094 * config/aarch64/aarch64.h
17095 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17097 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17100 * expmed.c (emit_store_flag_force): Swap if const op0
17101 and change VOIDmode to mode of op0.
17103 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17105 PR rtl-optimization/83761
17106 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17107 than bytes to mode_for_size.
17109 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17111 PR middle-end/83189
17112 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17113 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17116 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17118 PR middle-end/83575
17119 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17120 when in layout mode.
17121 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17122 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17125 2018-01-10 Michael Collison <michael.collison@arm.com>
17127 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17128 * config/aarch64/aarch64-option-extension.def: Add
17129 AARCH64_OPT_EXTENSION of 'fp16fml'.
17130 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17131 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17132 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17133 * config/aarch64/constraints.md (Ui7): New constraint.
17134 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17135 (VFMLA_SEL_W): Ditto.
17138 (VFMLA16_LOW): New int iterator.
17139 (VFMLA16_HIGH): Ditto.
17140 (UNSPEC_FMLAL): New unspec.
17141 (UNSPEC_FMLSL): Ditto.
17142 (UNSPEC_FMLAL2): Ditto.
17143 (UNSPEC_FMLSL2): Ditto.
17144 (f16mac): New code attribute.
17145 * config/aarch64/aarch64-simd-builtins.def
17146 (aarch64_fmlal_lowv2sf): Ditto.
17147 (aarch64_fmlsl_lowv2sf): Ditto.
17148 (aarch64_fmlalq_lowv4sf): Ditto.
17149 (aarch64_fmlslq_lowv4sf): Ditto.
17150 (aarch64_fmlal_highv2sf): Ditto.
17151 (aarch64_fmlsl_highv2sf): Ditto.
17152 (aarch64_fmlalq_highv4sf): Ditto.
17153 (aarch64_fmlslq_highv4sf): Ditto.
17154 (aarch64_fmlal_lane_lowv2sf): Ditto.
17155 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17156 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17157 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17158 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17159 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17160 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17161 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17162 (aarch64_fmlal_lane_highv2sf): Ditto.
17163 (aarch64_fmlsl_lane_highv2sf): Ditto.
17164 (aarch64_fmlal_laneq_highv2sf): Ditto.
17165 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17166 (aarch64_fmlalq_lane_highv4sf): Ditto.
17167 (aarch64_fmlsl_lane_highv4sf): Ditto.
17168 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17169 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17170 * config/aarch64/aarch64-simd.md:
17171 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17172 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17173 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17174 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17175 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17176 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17177 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17178 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17179 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17180 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17181 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17182 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17183 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17184 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17185 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17186 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17187 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17188 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17189 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17190 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17191 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17192 (vfmlsl_low_u32): Ditto.
17193 (vfmlalq_low_u32): Ditto.
17194 (vfmlslq_low_u32): Ditto.
17195 (vfmlal_high_u32): Ditto.
17196 (vfmlsl_high_u32): Ditto.
17197 (vfmlalq_high_u32): Ditto.
17198 (vfmlslq_high_u32): Ditto.
17199 (vfmlal_lane_low_u32): Ditto.
17200 (vfmlsl_lane_low_u32): Ditto.
17201 (vfmlal_laneq_low_u32): Ditto.
17202 (vfmlsl_laneq_low_u32): Ditto.
17203 (vfmlalq_lane_low_u32): Ditto.
17204 (vfmlslq_lane_low_u32): Ditto.
17205 (vfmlalq_laneq_low_u32): Ditto.
17206 (vfmlslq_laneq_low_u32): Ditto.
17207 (vfmlal_lane_high_u32): Ditto.
17208 (vfmlsl_lane_high_u32): Ditto.
17209 (vfmlal_laneq_high_u32): Ditto.
17210 (vfmlsl_laneq_high_u32): Ditto.
17211 (vfmlalq_lane_high_u32): Ditto.
17212 (vfmlslq_lane_high_u32): Ditto.
17213 (vfmlalq_laneq_high_u32): Ditto.
17214 (vfmlslq_laneq_high_u32): Ditto.
17215 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17216 (AARCH64_FL_FOR_ARCH8_4): New.
17217 (AARCH64_ISA_F16FML): New ISA flag.
17218 (TARGET_F16FML): New feature flag for fp16fml.
17219 (doc/invoke.texi): Document new fp16fml option.
17221 2018-01-10 Michael Collison <michael.collison@arm.com>
17223 * config/aarch64/aarch64-builtins.c:
17224 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17225 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17226 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17227 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17228 (AARCH64_ISA_SHA3): New ISA flag.
17229 (TARGET_SHA3): New feature flag for sha3.
17230 * config/aarch64/iterators.md (sha512_op): New int attribute.
17231 (CRYPTO_SHA512): New int iterator.
17232 (UNSPEC_SHA512H): New unspec.
17233 (UNSPEC_SHA512H2): Ditto.
17234 (UNSPEC_SHA512SU0): Ditto.
17235 (UNSPEC_SHA512SU1): Ditto.
17236 * config/aarch64/aarch64-simd-builtins.def
17237 (aarch64_crypto_sha512hqv2di): New builtin.
17238 (aarch64_crypto_sha512h2qv2di): Ditto.
17239 (aarch64_crypto_sha512su0qv2di): Ditto.
17240 (aarch64_crypto_sha512su1qv2di): Ditto.
17241 (aarch64_eor3qv8hi): Ditto.
17242 (aarch64_rax1qv2di): Ditto.
17243 (aarch64_xarqv2di): Ditto.
17244 (aarch64_bcaxqv8hi): Ditto.
17245 * config/aarch64/aarch64-simd.md:
17246 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17247 (aarch64_crypto_sha512su0qv2di): Ditto.
17248 (aarch64_crypto_sha512su1qv2di): Ditto.
17249 (aarch64_eor3qv8hi): Ditto.
17250 (aarch64_rax1qv2di): Ditto.
17251 (aarch64_xarqv2di): Ditto.
17252 (aarch64_bcaxqv8hi): Ditto.
17253 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17254 (vsha512h2q_u64): Ditto.
17255 (vsha512su0q_u64): Ditto.
17256 (vsha512su1q_u64): Ditto.
17257 (veor3q_u16): Ditto.
17258 (vrax1q_u64): Ditto.
17259 (vxarq_u64): Ditto.
17260 (vbcaxq_u16): Ditto.
17261 * config/arm/types.md (crypto_sha512): New type attribute.
17262 (crypto_sha3): Ditto.
17263 (doc/invoke.texi): Document new sha3 option.
17265 2018-01-10 Michael Collison <michael.collison@arm.com>
17267 * config/aarch64/aarch64-builtins.c:
17268 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17269 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17270 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17271 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17272 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17273 (AARCH64_ISA_SM4): New ISA flag.
17274 (TARGET_SM4): New feature flag for sm4.
17275 * config/aarch64/aarch64-simd-builtins.def
17276 (aarch64_sm3ss1qv4si): Ditto.
17277 (aarch64_sm3tt1aq4si): Ditto.
17278 (aarch64_sm3tt1bq4si): Ditto.
17279 (aarch64_sm3tt2aq4si): Ditto.
17280 (aarch64_sm3tt2bq4si): Ditto.
17281 (aarch64_sm3partw1qv4si): Ditto.
17282 (aarch64_sm3partw2qv4si): Ditto.
17283 (aarch64_sm4eqv4si): Ditto.
17284 (aarch64_sm4ekeyqv4si): Ditto.
17285 * config/aarch64/aarch64-simd.md:
17286 (aarch64_sm3ss1qv4si): Ditto.
17287 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17288 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17289 (aarch64_sm4eqv4si): Ditto.
17290 (aarch64_sm4ekeyqv4si): Ditto.
17291 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17292 (sm3part_op): Ditto.
17293 (CRYPTO_SM3TT): Ditto.
17294 (CRYPTO_SM3PART): Ditto.
17295 (UNSPEC_SM3SS1): New unspec.
17296 (UNSPEC_SM3TT1A): Ditto.
17297 (UNSPEC_SM3TT1B): Ditto.
17298 (UNSPEC_SM3TT2A): Ditto.
17299 (UNSPEC_SM3TT2B): Ditto.
17300 (UNSPEC_SM3PARTW1): Ditto.
17301 (UNSPEC_SM3PARTW2): Ditto.
17302 (UNSPEC_SM4E): Ditto.
17303 (UNSPEC_SM4EKEY): Ditto.
17304 * config/aarch64/constraints.md (Ui2): New constraint.
17305 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17306 * config/arm/types.md (crypto_sm3): New type attribute.
17307 (crypto_sm4): Ditto.
17308 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17309 (vsm3tt1aq_u32): Ditto.
17310 (vsm3tt1bq_u32): Ditto.
17311 (vsm3tt2aq_u32): Ditto.
17312 (vsm3tt2bq_u32): Ditto.
17313 (vsm3partw1q_u32): Ditto.
17314 (vsm3partw2q_u32): Ditto.
17315 (vsm4eq_u32): Ditto.
17316 (vsm4ekeyq_u32): Ditto.
17317 (doc/invoke.texi): Document new sm4 option.
17319 2018-01-10 Michael Collison <michael.collison@arm.com>
17321 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17322 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17323 (AARCH64_FL_FOR_ARCH8_4): New.
17324 (AARCH64_FL_V8_4): New flag.
17325 (doc/invoke.texi): Document new armv8.4-a option.
17327 2018-01-10 Michael Collison <michael.collison@arm.com>
17329 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17330 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17331 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17332 * config/aarch64/aarch64-option-extension.def: Add
17333 AARCH64_OPT_EXTENSION of 'sha2'.
17334 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17335 (crypto): Disable sha2 and aes if crypto disabled.
17336 (crypto): Enable aes and sha2 if enabled.
17337 (simd): Disable sha2 and aes if simd disabled.
17338 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17340 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17341 (TARGET_SHA2): New feature flag for sha2.
17342 (TARGET_AES): New feature flag for aes.
17343 * config/aarch64/aarch64-simd.md:
17344 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17345 conditional on TARGET_AES.
17346 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17347 (aarch64_crypto_sha1hsi): Make pattern conditional
17349 (aarch64_crypto_sha1hv4si): Ditto.
17350 (aarch64_be_crypto_sha1hv4si): Ditto.
17351 (aarch64_crypto_sha1su1v4si): Ditto.
17352 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17353 (aarch64_crypto_sha1su0v4si): Ditto.
17354 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17355 (aarch64_crypto_sha256su0v4si): Ditto.
17356 (aarch64_crypto_sha256su1v4si): Ditto.
17357 (doc/invoke.texi): Document new aes and sha2 options.
17359 2018-01-10 Martin Sebor <msebor@redhat.com>
17361 PR tree-optimization/83781
17362 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17365 2018-01-11 Martin Sebor <msebor@gmail.com>
17366 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17368 PR tree-optimization/83501
17369 PR tree-optimization/81703
17371 * tree-ssa-strlen.c (get_string_cst): Rename...
17372 (get_string_len): ...to this. Handle global constants.
17373 (handle_char_store): Adjust.
17375 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17376 Jim Wilson <jimw@sifive.com>
17378 * config/riscv/riscv-protos.h (riscv_output_return): New.
17379 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17380 (riscv_attribute_table, riscv_output_return),
17381 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17382 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17383 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17385 (riscv_expand_prologue): Add early return for naked function.
17386 (riscv_expand_epilogue): Likewise.
17387 (riscv_function_ok_for_sibcall): Return false for naked function.
17388 (riscv_set_current_function): New.
17389 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17390 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17391 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17392 * doc/extend.texi (RISC-V Function Attributes): New.
17394 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17396 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17397 check for 128-bit long double before checking TCmode.
17398 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17399 128-bit long doubles before checking TFmode or TCmode.
17400 (FLOAT128_IBM_P): Likewise.
17402 2018-01-10 Martin Sebor <msebor@redhat.com>
17404 PR tree-optimization/83671
17405 * builtins.c (c_strlen): Unconditionally return zero for the empty
17407 Use -Warray-bounds for warnings.
17408 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17409 for non-constant array indices with COMPONENT_REF, arrays of
17410 arrays, and pointers to arrays.
17411 (gimple_fold_builtin_strlen): Determine and set length range for
17412 non-constant character arrays.
17414 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17416 PR middle-end/81897
17417 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17420 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17422 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17424 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17427 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17428 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17429 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17430 indexed_or_indirect_operand predicate.
17431 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17432 (*vsx_le_perm_load_v8hi): Likewise.
17433 (*vsx_le_perm_load_v16qi): Likewise.
17434 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17435 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17436 (*vsx_le_perm_store_v8hi): Likewise.
17437 (*vsx_le_perm_store_v16qi): Likewise.
17438 (eight unnamed splitters): Likewise.
17440 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17442 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17443 * config/rs6000/emmintrin.h: Likewise.
17444 * config/rs6000/mmintrin.h: Likewise.
17445 * config/rs6000/xmmintrin.h: Likewise.
17447 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17450 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17452 * tree.c (tree_nop_conversion): Return true for location wrapper
17454 (maybe_wrap_with_location): New function.
17455 (selftest::check_strip_nops): New function.
17456 (selftest::test_location_wrappers): New function.
17457 (selftest::tree_c_tests): Call it.
17458 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17459 (maybe_wrap_with_location): New decl.
17460 (EXPR_LOCATION_WRAPPER_P): New macro.
17461 (location_wrapper_p): New inline function.
17462 (tree_strip_any_location_wrapper): New inline function.
17464 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17467 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17468 stack_realign_offset for the largest alignment of stack slot
17470 (ix86_find_max_used_stack_alignment): New function.
17471 (ix86_finalize_stack_frame_flags): Use it. Set
17472 max_used_stack_alignment if we don't realign stack.
17473 * config/i386/i386.h (machine_function): Add
17474 max_used_stack_alignment.
17476 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17478 * config/arm/arm.opt (-mbranch-cost): New option.
17479 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17482 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17485 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17486 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17488 2018-01-10 Richard Biener <rguenther@suse.de>
17491 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17492 early out so it also covers the case where we have a non-NULL
17495 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17497 PR tree-optimization/83753
17498 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17499 for non-strided grouped accesses if the number of elements is 1.
17501 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17504 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17505 * i386.h (TARGET_USE_GATHER): Define.
17506 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17508 2018-01-10 Martin Liska <mliska@suse.cz>
17511 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17512 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17514 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17515 CLEANUP_NO_PARTITIONING is not set.
17517 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17519 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17520 for vectors, as a partial revert of r254296.
17521 * rtl.h (const_vec_p): Delete.
17522 (const_vec_duplicate_p): Don't test for vector CONSTs.
17523 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17524 * expmed.c (make_tree): Likewise.
17527 * common.md (E, F): Use CONSTANT_P instead of checking for
17529 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17530 checking for CONST_VECTOR.
17532 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17534 PR middle-end/83575
17535 * predict.c (force_edge_cold): Handle in more sane way edges
17536 with no prediction.
17538 2018-01-09 Carl Love <cel@us.ibm.com>
17540 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17542 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17543 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17544 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17545 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17546 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17547 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17548 * config/rs6000/rs6000-protos.h: Add extern defition for
17549 rs6000_generate_float2_double_code.
17550 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17552 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17553 (float2_v2df): Add define_expand.
17555 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17558 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17559 op_mode in the force_to_mode call.
17561 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17563 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17564 instead of checking each element individually.
17565 (aarch64_evpc_uzp): Likewise.
17566 (aarch64_evpc_zip): Likewise.
17567 (aarch64_evpc_ext): Likewise.
17568 (aarch64_evpc_rev): Likewise.
17569 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17570 instead of checking each element individually. Return true without
17572 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17573 whether all selected elements come from the same input, instead of
17574 checking each element individually. Remove calls to gen_rtx_REG,
17575 start_sequence and end_sequence and instead assert that no rtl is
17578 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17580 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17581 order of HIGH and CONST checks.
17583 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17585 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17586 if the destination isn't an SSA_NAME.
17588 2018-01-09 Richard Biener <rguenther@suse.de>
17590 PR tree-optimization/83668
17591 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17593 (canonicalize_loop_form): ... here, renamed from ...
17594 (canonicalize_loop_closed_ssa_form): ... this and amended to
17595 swap successor edges for loop exit blocks to make us use
17596 the RPO order we need for initial schedule generation.
17598 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17600 PR tree-optimization/64811
17601 * match.pd: When optimizing comparisons with Inf, avoid
17602 introducing or losing exceptions from comparisons with NaN.
17604 2018-01-09 Martin Liska <mliska@suse.cz>
17607 * asan.c (shadow_mem_size): Add gcc_assert.
17609 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17611 Don't save registers in main().
17614 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17615 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17616 * config/avr/avr.c (avr_set_current_function): Don't error if
17617 naked, OS_task or OS_main are specified at the same time.
17618 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17620 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17622 * common/config/avr/avr-common.c (avr_option_optimization_table):
17623 Switch on -mmain-is-OS_task for optimizing compilations.
17625 2018-01-09 Richard Biener <rguenther@suse.de>
17627 PR tree-optimization/83572
17628 * graphite.c: Include cfganal.h.
17629 (graphite_transform_loops): Connect infinite loops to exit
17630 and remove fake edges at the end.
17632 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17634 * ipa-inline.c (edge_badness): Revert accidental checkin.
17636 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17639 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17640 symbols; not inline clones.
17642 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17645 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17646 hard registers. Formatting fixes.
17648 PR preprocessor/83722
17649 * gcc.c (try_generate_repro): Pass
17650 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17651 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17654 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17655 Kito Cheng <kito.cheng@gmail.com>
17657 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17658 (riscv_leaf_function_p): Delete.
17659 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17661 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17663 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17665 (do_ifelse): New function.
17666 (do_isel): New function.
17667 (do_sub3): New function.
17668 (do_add3): New function.
17669 (do_load_mask_compare): New function.
17670 (do_overlap_load_compare): New function.
17671 (expand_compare_loop): New function.
17672 (expand_block_compare): Call expand_compare_loop() when appropriate.
17673 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17674 option description.
17675 (-mblock-compare-inline-loop-limit): New option.
17677 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17680 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17681 Reverse order of second and third operands in first alternative.
17682 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17683 of first and second elements in UNSPEC_VPERMR vector.
17684 (altivec_expand_vec_perm_le): Likewise.
17686 2018-01-08 Jeff Law <law@redhat.com>
17688 PR rtl-optimizatin/81308
17689 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17690 (process_switch): If group_case_labels makes a change, then set
17692 (pass_convert_switch::execute): If a switch is converted, then
17693 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17695 PR rtl-optimization/81308
17696 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17699 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17701 PR target/83663 - Revert r255946
17702 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17703 generation for cases where splatting a value is not useful.
17704 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17705 across a vec_duplicate and a paradoxical subreg forming a vector
17706 mode to a vec_concat.
17708 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17710 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17711 -march=armv8.3-a variants.
17712 * config/arm/t-multilib: Likewise.
17713 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17715 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17717 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17719 (cceq_ior_compare_complement): Give it a name so I can use it, and
17720 change boolean_or_operator predicate to boolean_operator so it can
17721 be used to generate a crand.
17722 (eqne): New code iterator.
17723 (bd/bd_neg): New code_attrs.
17724 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17725 a single define_insn.
17726 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17727 decrement (bdnzt/bdnzf/bdzt/bdzf).
17728 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17729 with the new names of the branch decrement patterns, and added the
17730 names of the branch decrement conditional patterns.
17732 2018-01-08 Richard Biener <rguenther@suse.de>
17734 PR tree-optimization/83563
17735 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17738 2018-01-08 Richard Biener <rguenther@suse.de>
17740 PR middle-end/83713
17741 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17743 2018-01-08 Richard Biener <rguenther@suse.de>
17745 PR tree-optimization/83685
17746 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17747 references to abnormals.
17749 2018-01-08 Richard Biener <rguenther@suse.de>
17752 * dwarf2out.c (output_indirect_strings): Handle empty
17753 skeleton_debug_str_hash.
17754 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17756 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17758 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17759 (emit_store_direct): Likewise.
17760 (arc_trampoline_adjust_address): Likewise.
17761 (arc_asm_trampoline_template): New function.
17762 (arc_initialize_trampoline): Use asm_trampoline_template.
17763 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17764 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17765 * config/arc/arc.md (flush_icache): Delete pattern.
17767 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17769 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17770 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17773 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17776 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17777 by not USED_FOR_TARGET.
17778 (make_pass_resolve_sw_modes): Likewise.
17780 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17782 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17785 2018-01-08 Richard Biener <rguenther@suse.de>
17787 PR middle-end/83580
17788 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17790 2018-01-08 Richard Biener <rguenther@suse.de>
17792 PR middle-end/83517
17793 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17795 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17797 PR middle-end/81897
17798 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17799 basic blocks with a small number of successors.
17800 (convert_control_dep_chain_into_preds): Improve handling of
17802 (dump_predicates): Split apart into...
17803 (dump_pred_chain): ...here...
17804 (dump_pred_info): ...and here.
17805 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17806 (can_chain_union_be_invalidated_p): Improve check for invalidation
17808 (uninit_uses_cannot_happen): Avoid unnecessary if
17809 convert_control_dep_chain_into_preds yielded nothing.
17811 2018-01-06 Martin Sebor <msebor@redhat.com>
17813 PR tree-optimization/83640
17814 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17815 subtracting negative offset from size.
17816 (builtin_access::overlap): Adjust offset bounds of the access to fall
17817 within the size of the object if possible.
17819 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17821 PR rtl-optimization/83699
17822 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17823 extract_bit_field_as_subreg to cases in which the extracted
17824 value is also a vector.
17826 * lra-constraints.c (process_alt_operands): Test for the equivalence
17827 substitutions when detecting a possible reload cycle.
17829 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17832 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17833 by default if flag_selective_schedling{,2}. Formatting fixes.
17835 PR rtl-optimization/83682
17836 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17837 if it has non-VECTOR_MODE element mode.
17838 (vec_duplicate_p): Likewise.
17840 PR middle-end/83694
17841 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17842 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17844 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17847 * config/i386/i386-builtin.def
17848 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17849 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17850 Require also OPTION_MASK_ISA_AVX512F in addition to
17851 OPTION_MASK_ISA_GFNI.
17852 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17853 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17854 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17855 to OPTION_MASK_ISA_GFNI.
17856 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17857 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17858 OPTION_MASK_ISA_AVX512BW.
17859 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17860 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17861 addition to OPTION_MASK_ISA_GFNI.
17862 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17863 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17864 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17865 to OPTION_MASK_ISA_GFNI.
17866 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17867 a requirement for all ISAs rather than any of them with a few
17869 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17871 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17872 bitmasks to be enabled with 3 exceptions, instead of requiring any
17873 enabled ISA with lots of exceptions.
17874 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17875 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17876 Change avx512bw in isa attribute to avx512f.
17877 * config/i386/sgxintrin.h: Add license boilerplate.
17878 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17879 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17880 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17881 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17883 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17884 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17885 temporarily sse2 rather than sse if not enabled already.
17888 * config/i386/sse.md (VI248_VLBW): Rename to ...
17889 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17890 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17891 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17892 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17893 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17894 mode iterator instead of VI248_VLBW.
17896 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17898 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17899 (record_modified): Skip clobbers; add debug output.
17900 (param_change_prob): Use sreal frequencies.
17902 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17904 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17905 punt for user-aligned variables.
17907 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17909 * tree-chrec.c (chrec_contains_symbols): Return true for
17912 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17915 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17916 of (x|y) == x for BICS pattern.
17918 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17920 PR tree-optimization/83605
17921 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17922 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17925 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17927 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17928 * config/epiphany/rtems.h: New file.
17930 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17931 Uros Bizjak <ubizjak@gmail.com>
17934 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17935 QIreg_operand instead of register_operand predicate.
17936 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17937 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17938 comments instead of -fmitigate[-_]rop.
17940 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17943 * cgraphunit.c (symbol_table::compile): Switch to text_section
17944 before calling assembly_start debug hook.
17945 * run-rtl-passes.c (run_rtl_passes): Likewise.
17948 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17950 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17951 range_int_cst_p rather than !symbolic_range_p before calling
17952 extract_range_from_multiplicative_op_1.
17954 2018-01-04 Jeff Law <law@redhat.com>
17956 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17957 redundant test in assertion.
17959 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17961 * doc/rtl.texi: Document machine_mode wrapper classes.
17963 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17965 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17966 using tree_to_uhwi.
17968 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17970 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17971 the VEC_PERM_EXPR fold to fail.
17973 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17976 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17977 to switched_sections.
17979 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17982 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17983 test for d.testing.
17985 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17988 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17989 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17991 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17994 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17995 is BLKmode and bitpos not zero or mode change is needed.
17997 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18000 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18003 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18006 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18007 instead of MULT rtx. Update all corresponding splitters.
18008 (*saddl_se): Ditto.
18009 (*ssub<modesuffix>): Ditto.
18010 (*ssubl_se): Ditto.
18011 (*cmp_sadd_di): Update split patterns.
18012 (*cmp_sadd_si): Ditto.
18013 (*cmp_sadd_sidi): Ditto.
18014 (*cmp_ssub_di): Ditto.
18015 (*cmp_ssub_si): Ditto.
18016 (*cmp_ssub_sidi): Ditto.
18017 * config/alpha/predicates.md (const23_operand): New predicate.
18018 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18019 Look for ASHIFT, not MULT inner operand.
18020 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18022 2018-01-04 Martin Liska <mliska@suse.cz>
18024 PR gcov-profile/83669
18025 * gcov.c (output_intermediate_file): Add version to intermediate
18027 * doc/gcov.texi: Document new field 'version' in intermediate
18028 file format. Fix location of '-k' option of gcov command.
18030 2018-01-04 Martin Liska <mliska@suse.cz>
18033 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18035 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18037 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18039 2018-01-03 Martin Sebor <msebor@redhat.com>
18041 PR tree-optimization/83655
18042 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18043 checking calls with invalid arguments.
18045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18047 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18048 (vectorizable_mask_load_store): Delete.
18049 (vectorizable_call): Return false for masked loads and stores.
18050 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18051 instead of gimple_assign_rhs1.
18052 (vectorizable_load): Handle IFN_MASK_LOAD.
18053 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18057 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18059 (vectorizable_mask_load_store): ...here.
18060 (vectorizable_load): ...and here.
18062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18064 * tree-vect-stmts.c (vect_build_all_ones_mask)
18065 (vect_build_zero_merge_argument): New functions, split out from...
18066 (vectorizable_load): ...here.
18068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18070 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18072 (vectorizable_mask_load_store): ...here.
18073 (vectorizable_store): ...and here.
18075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18077 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18079 (vectorizable_mask_load_store): ...here.
18081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18083 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18084 (vect_model_store_cost): Take a vec_load_store_type instead of a
18086 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18087 (vect_model_store_cost): Take a vec_load_store_type instead of a
18089 (vectorizable_mask_load_store): Update accordingly.
18090 (vectorizable_store): Likewise.
18091 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18095 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18096 IFN_MASK_LOAD calls here rather than...
18097 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18100 Alan Hayward <alan.hayward@arm.com>
18101 David Sherwood <david.sherwood@arm.com>
18103 * expmed.c (extract_bit_field_1): For vector extracts,
18104 fall back to extract_bit_field_as_subreg if vec_extract
18107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18108 Alan Hayward <alan.hayward@arm.com>
18109 David Sherwood <david.sherwood@arm.com>
18111 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18112 they are variable or constant sized.
18113 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18114 slots for constant-sized data.
18116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18117 Alan Hayward <alan.hayward@arm.com>
18118 David Sherwood <david.sherwood@arm.com>
18120 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18121 handling COND_EXPRs with boolean comparisons, try to find a better
18122 basis for the mask type than the boolean itself.
18124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18126 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18127 is calculated and how it can be overridden.
18128 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18129 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18131 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18135 Alan Hayward <alan.hayward@arm.com>
18136 David Sherwood <david.sherwood@arm.com>
18138 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18139 Remove the mode argument.
18140 (aarch64_simd_valid_immediate): Remove the mode and inverse
18142 * config/aarch64/iterators.md (bitsize): New iterator.
18143 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18144 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18145 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18146 aarch64_simd_valid_immediate.
18147 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18148 (aarch64_reg_or_bic_imm): Likewise.
18149 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18150 with an insn_type enum and msl with a modifier_type enum.
18151 Replace element_width with a scalar_mode. Change the shift
18152 to unsigned int. Add constructors for scalar_float_mode and
18153 scalar_int_mode elements.
18154 (aarch64_vect_float_const_representable_p): Delete.
18155 (aarch64_can_const_movi_rtx_p)
18156 (aarch64_simd_scalar_immediate_valid_for_move)
18157 (aarch64_simd_make_constant): Update call to
18158 aarch64_simd_valid_immediate.
18159 (aarch64_advsimd_valid_immediate_hs): New function.
18160 (aarch64_advsimd_valid_immediate): Likewise.
18161 (aarch64_simd_valid_immediate): Remove mode and inverse
18162 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18163 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18164 and aarch64_float_const_representable_p on the result.
18165 (aarch64_output_simd_mov_immediate): Remove mode argument.
18166 Update call to aarch64_simd_valid_immediate and use of
18167 simd_immediate_info.
18168 (aarch64_output_scalar_simd_mov_immediate): Update call
18171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18172 Alan Hayward <alan.hayward@arm.com>
18173 David Sherwood <david.sherwood@arm.com>
18175 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18176 (mode_nunits): Likewise CONST_MODE_NUNITS.
18177 * machmode.def (ADJUST_NUNITS): Document.
18178 * genmodes.c (mode_data::need_nunits_adj): New field.
18179 (blank_mode): Update accordingly.
18180 (adj_nunits): New variable.
18181 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18183 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18184 listed in adj_nunits.
18185 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18186 listed in adj_nunits. Don't emit case statements for such modes.
18187 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18188 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18189 nothing if adj_nunits is nonnull.
18190 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18191 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18192 (emit_mode_fbit): Update use of print_maybe_const_decl.
18193 (emit_move_size): Likewise. Treat the array as non-const
18195 (emit_mode_adjustments): Handle adj_nunits.
18197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18199 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18200 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18201 (VECTOR_MODES): Use it.
18202 (make_vector_modes): Take the prefix as an argument.
18204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18205 Alan Hayward <alan.hayward@arm.com>
18206 David Sherwood <david.sherwood@arm.com>
18208 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18209 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18210 for MODE_VECTOR_BOOL.
18211 * machmode.def (VECTOR_BOOL_MODE): Document.
18212 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18213 (make_vector_bool_mode): New function.
18214 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18216 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18217 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18219 * stor-layout.c (int_mode_for_mode): Likewise.
18220 * tree.c (build_vector_type_for_mode): Likewise.
18221 * varasm.c (output_constant_pool_2): Likewise.
18222 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18223 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18224 for MODE_VECTOR_BOOL.
18225 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18226 of mode class checks.
18227 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18228 instead of a list of mode class checks.
18229 (expand_vector_scalar_condition): Likewise.
18230 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18233 Alan Hayward <alan.hayward@arm.com>
18234 David Sherwood <david.sherwood@arm.com>
18236 * machmode.h (mode_size): Change from unsigned short to
18238 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18239 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18240 or if measurement_type is not polynomial.
18241 (fixed_size_mode::includes_p): Check for constant-sized modes.
18242 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18243 return a poly_uint16 rather than an unsigned short.
18244 (emit_mode_size): Change the type of mode_size from unsigned short
18245 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18246 (emit_mode_adjustments): Cope with polynomial vector sizes.
18247 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18249 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18251 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18252 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18253 * caller-save.c (setup_save_areas): Likewise.
18254 (replace_reg_with_saved_mem): Likewise.
18255 * calls.c (emit_library_call_value_1): Likewise.
18256 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18257 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18258 (gen_lowpart_for_combine): Likewise.
18259 * convert.c (convert_to_integer_1): Likewise.
18260 * cse.c (equiv_constant, cse_insn): Likewise.
18261 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18262 (cselib_subst_to_values): Likewise.
18263 * dce.c (word_dce_process_block): Likewise.
18264 * df-problems.c (df_word_lr_mark_ref): Likewise.
18265 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18266 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18267 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18268 (rtl_for_decl_location): Likewise.
18269 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18270 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18271 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18272 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18273 (expand_expr_real_1): Likewise.
18274 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18275 (pad_below): Likewise.
18276 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18277 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18278 * ira.c (get_subreg_tracking_sizes): Likewise.
18279 * ira-build.c (ira_create_allocno_objects): Likewise.
18280 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18281 (ira_sort_regnos_for_alter_reg): Likewise.
18282 * ira-costs.c (record_operand_costs): Likewise.
18283 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18284 (resolve_simple_move): Likewise.
18285 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18286 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18287 (lra_constraints): Likewise.
18288 (CONST_POOL_OK_P): Reject variable-sized modes.
18289 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18290 (add_pseudo_to_slot, lra_spill): Likewise.
18291 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18292 * optabs-query.c (get_best_extraction_insn): Likewise.
18293 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18294 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18295 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18296 * recog.c (offsettable_address_addr_space_p): Likewise.
18297 * regcprop.c (maybe_mode_change): Likewise.
18298 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18299 * regrename.c (build_def_use): Likewise.
18300 * regstat.c (dump_reg_info): Likewise.
18301 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18302 (find_reloads, find_reloads_subreg_address): Likewise.
18303 * reload1.c (eliminate_regs_1): Likewise.
18304 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18305 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18306 (simplify_binary_operation_1, simplify_subreg): Likewise.
18307 * targhooks.c (default_function_arg_padding): Likewise.
18308 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18309 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18310 (verify_gimple_assign_ternary): Likewise.
18311 * tree-inline.c (estimate_move_cost): Likewise.
18312 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18313 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18314 (get_address_cost_ainc): Likewise.
18315 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18316 (vect_supportable_dr_alignment): Likewise.
18317 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18318 (vectorizable_reduction): Likewise.
18319 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18320 (vectorizable_operation, vectorizable_load): Likewise.
18321 * tree.c (build_same_sized_truth_vector_type): Likewise.
18322 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18323 * var-tracking.c (emit_note_insn_var_location): Likewise.
18324 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18325 (ADDR_VEC_ALIGN): Likewise.
18327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18328 Alan Hayward <alan.hayward@arm.com>
18329 David Sherwood <david.sherwood@arm.com>
18331 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18333 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18334 or if measurement_type is polynomial.
18335 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18336 * combine.c (make_extraction): Likewise.
18337 * dse.c (find_shift_sequence): Likewise.
18338 * dwarf2out.c (mem_loc_descriptor): Likewise.
18339 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18340 (extract_bit_field, extract_low_bits): Likewise.
18341 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18342 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18343 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18344 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18345 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18346 * reload.c (find_reloads): Likewise.
18347 * reload1.c (alter_reg): Likewise.
18348 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18349 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18350 * tree-if-conv.c (predicate_mem_writes): Likewise.
18351 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18352 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18353 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18354 * valtrack.c (dead_debug_insert_temp): Likewise.
18355 * varasm.c (mergeable_constant_section): Likewise.
18356 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18359 Alan Hayward <alan.hayward@arm.com>
18360 David Sherwood <david.sherwood@arm.com>
18362 * expr.c (expand_assignment): Cope with polynomial mode sizes
18363 when assigning to a CONCAT.
18365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18366 Alan Hayward <alan.hayward@arm.com>
18367 David Sherwood <david.sherwood@arm.com>
18369 * machmode.h (mode_precision): Change from unsigned short to
18371 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18373 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18374 or if measurement_type is not polynomial.
18375 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18376 in which the mode is already known to be a scalar_int_mode.
18377 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18378 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18380 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18381 for GET_MODE_PRECISION.
18382 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18383 for GET_MODE_PRECISION.
18384 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18386 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18387 (expand_field_assignment, make_extraction): Likewise.
18388 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18389 (get_last_value): Likewise.
18390 * convert.c (convert_to_integer_1): Likewise.
18391 * cse.c (cse_insn): Likewise.
18392 * expr.c (expand_expr_real_1): Likewise.
18393 * lra-constraints.c (simplify_operand_subreg): Likewise.
18394 * optabs-query.c (can_atomic_load_p): Likewise.
18395 * optabs.c (expand_atomic_load): Likewise.
18396 (expand_atomic_store): Likewise.
18397 * ree.c (combine_reaching_defs): Likewise.
18398 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18399 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18400 * tree.h (type_has_mode_precision_p): Likewise.
18401 * ubsan.c (instrument_si_overflow): Likewise.
18403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18404 Alan Hayward <alan.hayward@arm.com>
18405 David Sherwood <david.sherwood@arm.com>
18407 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18408 polynomial numbers of units.
18409 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18410 (valid_vector_subparts_p): New function.
18411 (build_vector_type): Remove temporary shim and take the number
18412 of units as a poly_uint64 rather than an int.
18413 (build_opaque_vector_type): Take the number of units as a
18414 poly_uint64 rather than an int.
18415 * tree.c (build_vector_from_ctor): Handle polynomial
18416 TYPE_VECTOR_SUBPARTS.
18417 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18418 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18419 (build_vector_from_val): If the number of units is variable,
18420 use build_vec_duplicate_cst for constant operands and
18421 VEC_DUPLICATE_EXPR otherwise.
18422 (make_vector_type): Remove temporary is_constant ().
18423 (build_vector_type, build_opaque_vector_type): Take the number of
18424 units as a poly_uint64 rather than an int.
18425 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18427 * cfgexpand.c (expand_debug_expr): Likewise.
18428 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18429 (store_constructor, expand_expr_real_1): Likewise.
18430 (const_scalar_mask_from_tree): Likewise.
18431 * fold-const-call.c (fold_const_reduction): Likewise.
18432 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18433 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18434 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18435 (fold_relational_const): Likewise.
18436 (native_interpret_vector): Likewise. Change the size from an
18437 int to an unsigned int.
18438 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18439 TYPE_VECTOR_SUBPARTS.
18440 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18441 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18442 duplicating a non-constant operand into a variable-length vector.
18443 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18444 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18445 * ipa-icf.c (sem_variable::equals): Likewise.
18446 * match.pd: Likewise.
18447 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18448 * print-tree.c (print_node): Likewise.
18449 * stor-layout.c (layout_type): Likewise.
18450 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18451 * tree-cfg.c (verify_gimple_comparison): Likewise.
18452 (verify_gimple_assign_binary): Likewise.
18453 (verify_gimple_assign_ternary): Likewise.
18454 (verify_gimple_assign_single): Likewise.
18455 * tree-pretty-print.c (dump_generic_node): Likewise.
18456 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18457 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18458 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18459 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18460 (vect_shift_permute_load_chain): Likewise.
18461 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18462 (expand_vector_condition, optimize_vector_constructor): Likewise.
18463 (lower_vec_perm, get_compute_type): Likewise.
18464 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18465 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18466 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18467 (vect_recog_mask_conversion_pattern): Likewise.
18468 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18469 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18470 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18471 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18472 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18473 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18474 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18475 (supportable_widening_operation): Likewise.
18476 (supportable_narrowing_operation): Likewise.
18477 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18479 * varasm.c (output_constant): Likewise.
18481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18482 Alan Hayward <alan.hayward@arm.com>
18483 David Sherwood <david.sherwood@arm.com>
18485 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18486 so that both the length == 3 and length != 3 cases set up their
18487 own permute vectors. Add comments explaining why we know the
18488 number of elements is constant.
18489 (vect_permute_load_chain): Likewise.
18491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18492 Alan Hayward <alan.hayward@arm.com>
18493 David Sherwood <david.sherwood@arm.com>
18495 * machmode.h (mode_nunits): Change from unsigned char to
18497 (ONLY_FIXED_SIZE_MODES): New macro.
18498 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18499 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18500 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18502 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18503 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18504 or if measurement_type is not polynomial.
18505 * genmodes.c (ZERO_COEFFS): New macro.
18506 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18508 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18509 Use ZERO_COEFFS when emitting initializers.
18510 * data-streamer.h (bp_pack_poly_value): New function.
18511 (bp_unpack_poly_value): Likewise.
18512 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18513 for GET_MODE_NUNITS.
18514 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18515 for GET_MODE_NUNITS.
18516 * tree.c (make_vector_type): Remove temporary shim and make
18517 the real function take the number of units as a poly_uint64
18518 rather than an int.
18519 (build_vector_type_for_mode): Handle polynomial nunits.
18520 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18521 * emit-rtl.c (const_vec_series_p_1): Likewise.
18522 (gen_rtx_CONST_VECTOR): Likewise.
18523 * fold-const.c (test_vec_duplicate_folding): Likewise.
18524 * genrecog.c (validate_pattern): Likewise.
18525 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18526 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18527 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18528 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18529 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18530 * rtlanal.c (subreg_get_info): Likewise.
18531 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18532 (vect_grouped_load_supported): Likewise.
18533 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18534 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18535 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18536 (simplify_const_unary_operation, simplify_binary_operation_1)
18537 (simplify_const_binary_operation, simplify_ternary_operation)
18538 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18539 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18540 instead of CONST_VECTOR_NUNITS.
18541 * varasm.c (output_constant_pool_2): Likewise.
18542 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18543 explicit-encoded elements in the XVEC for variable-length vectors.
18545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18547 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18550 Alan Hayward <alan.hayward@arm.com>
18551 David Sherwood <david.sherwood@arm.com>
18553 * coretypes.h (fixed_size_mode): Declare.
18554 (fixed_size_mode_pod): New typedef.
18555 * builtins.h (target_builtins::x_apply_args_mode)
18556 (target_builtins::x_apply_result_mode): Change type to
18557 fixed_size_mode_pod.
18558 * builtins.c (apply_args_size, apply_result_size, result_vector)
18559 (expand_builtin_apply_args_1, expand_builtin_apply)
18560 (expand_builtin_return): Update accordingly.
18562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18564 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18565 * cselib.c (cselib_hash_rtx): Likewise.
18566 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18567 CONST_VECTOR encoding.
18569 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18570 Jeff Law <law@redhat.com>
18573 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18574 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18575 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18576 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18579 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18580 explicitly probe *sp in a noreturn function if there were any callee
18581 register saves or frame pointer is needed.
18583 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18586 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18587 BLKmode for ternary, binary or unary expressions.
18590 * var-tracking.c (delete_vta_debug_insn): New inline function.
18591 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18592 insns from get_insns () to NULL instead of each bb separately.
18593 Use delete_vta_debug_insn. No longer static.
18594 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18595 delete_vta_debug_insns callers.
18596 * rtl.h (delete_vta_debug_insns): Declare.
18597 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18598 instead of variable_tracking_main.
18600 2018-01-03 Martin Sebor <msebor@redhat.com>
18602 PR tree-optimization/83603
18603 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18604 arguments past the endof the argument list in functions declared
18605 without a prototype.
18606 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18607 Avoid checking when arguments are null.
18609 2018-01-03 Martin Sebor <msebor@redhat.com>
18612 * doc/extend.texi (attribute const): Fix a typo.
18613 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18614 issuing -Wsuggest-attribute for void functions.
18616 2018-01-03 Martin Sebor <msebor@redhat.com>
18618 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18619 offset_int::from instead of wide_int::to_shwi.
18620 (maybe_diag_overlap): Remove assertion.
18621 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18622 * gimple-ssa-sprintf.c (format_directive): Same.
18623 (parse_directive): Same.
18624 (sprintf_dom_walker::compute_format_length): Same.
18625 (try_substitute_return_value): Same.
18627 2018-01-03 Jeff Law <law@redhat.com>
18629 PR middle-end/83654
18630 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18631 non-constant residual for zero at runtime and avoid probing in
18632 that case. Reorganize code for trailing problem to mirror handling
18635 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18637 PR tree-optimization/83501
18638 * tree-ssa-strlen.c (get_string_cst): New.
18639 (handle_char_store): Call get_string_cst.
18641 2018-01-03 Martin Liska <mliska@suse.cz>
18643 PR tree-optimization/83593
18644 * tree-ssa-strlen.c: Include tree-cfg.h.
18645 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18646 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18647 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18649 (strlen_dom_walker::before_dom_children): Call
18650 gimple_purge_dead_eh_edges. Dump tranformation with details
18652 (strlen_dom_walker::before_dom_children): Update call by adding
18653 new argument cleanup_eh.
18654 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18656 2018-01-03 Martin Liska <mliska@suse.cz>
18659 * cif-code.def (VARIADIC_THUNK): New enum value.
18660 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18663 2018-01-03 Jan Beulich <jbeulich@suse.com>
18665 * sse.md (mov<mode>_internal): Tighten condition for when to use
18666 vmovdqu<ssescalarsize> for TI and OI modes.
18668 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18670 Update copyright years.
18672 2018-01-03 Martin Liska <mliska@suse.cz>
18675 * ipa-visibility.c (function_and_variable_visibility): Skip
18676 functions with noipa attribure.
18678 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18680 * gcc.c (process_command): Update copyright notice dates.
18681 * gcov-dump.c (print_version): Ditto.
18682 * gcov.c (print_version): Ditto.
18683 * gcov-tool.c (print_version): Ditto.
18684 * gengtype.c (create_file): Ditto.
18685 * doc/cpp.texi: Bump @copying's copyright year.
18686 * doc/cppinternals.texi: Ditto.
18687 * doc/gcc.texi: Ditto.
18688 * doc/gccint.texi: Ditto.
18689 * doc/gcov.texi: Ditto.
18690 * doc/install.texi: Ditto.
18691 * doc/invoke.texi: Ditto.
18693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18695 * vector-builder.h (vector_builder::m_full_nelts): Change from
18696 unsigned int to poly_uint64.
18697 (vector_builder::full_nelts): Update prototype accordingly.
18698 (vector_builder::new_vector): Likewise.
18699 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18700 (vector_builder::operator ==): Likewise.
18701 (vector_builder::finalize): Likewise.
18702 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18703 Take the number of elements as a poly_uint64 rather than an
18705 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18706 from unsigned int to poly_uint64.
18707 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18708 (vec_perm_indices::new_vector): Likewise.
18709 (vec_perm_indices::length): Likewise.
18710 (vec_perm_indices::nelts_per_input): Likewise.
18711 (vec_perm_indices::input_nelts): Likewise.
18712 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18713 number of elements per input as a poly_uint64 rather than an
18714 unsigned int. Use the original encoding for variable-length
18715 vectors, rather than clamping each individual element.
18716 For the second and subsequent elements in each pattern,
18717 clamp the step and base before clamping their sum.
18718 (vec_perm_indices::series_p): Handle polynomial element counts.
18719 (vec_perm_indices::all_in_range_p): Likewise.
18720 (vec_perm_indices_to_tree): Likewise.
18721 (vec_perm_indices_to_rtx): Likewise.
18722 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18723 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18724 (tree_vector_builder::new_binary_operation): Handle polynomial
18725 element counts. Return false if we need to know the number
18726 of elements at compile time.
18727 * fold-const.c (fold_vec_perm): Punt if the number of elements
18728 isn't known at compile time.
18730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18732 * vec-perm-indices.h (vec_perm_builder): Change element type
18733 from HOST_WIDE_INT to poly_int64.
18734 (vec_perm_indices::element_type): Update accordingly.
18735 (vec_perm_indices::clamp): Handle polynomial element_types.
18736 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18737 (vec_perm_indices::all_in_range_p): Likewise.
18738 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18740 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18741 polynomial vec_perm_indices element types.
18742 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18743 * fold-const.c (fold_vec_perm): Likewise.
18744 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18745 * tree-vect-generic.c (lower_vec_perm): Likewise.
18746 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18747 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18748 element type to HOST_WIDE_INT.
18750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18751 Alan Hayward <alan.hayward@arm.com>
18752 David Sherwood <david.sherwood@arm.com>
18754 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18755 rather than an int. Use plus_constant.
18756 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18757 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18760 Alan Hayward <alan.hayward@arm.com>
18761 David Sherwood <david.sherwood@arm.com>
18763 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18764 a HOST_WIDE_INT to a poly_int64.
18766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18767 Alan Hayward <alan.hayward@arm.com>
18768 David Sherwood <david.sherwood@arm.com>
18770 * calls.c (load_register_parameters): Cope with polynomial
18771 mode sizes. Require a constant size for BLKmode parameters
18772 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18773 forces a parameter to be padded at the lsb end in order to
18774 fill a complete number of words, require the parameter size
18775 to be ordered wrt UNITS_PER_WORD.
18777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18778 Alan Hayward <alan.hayward@arm.com>
18779 David Sherwood <david.sherwood@arm.com>
18781 * reload1.c (spill_stack_slot_width): Change element type
18782 from unsigned int to poly_uint64_pod.
18783 (alter_reg): Treat mode sizes as polynomial.
18785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18786 Alan Hayward <alan.hayward@arm.com>
18787 David Sherwood <david.sherwood@arm.com>
18789 * reload.c (complex_word_subreg_p): New function.
18790 (reload_inner_reg_of_subreg, push_reload): Use it.
18792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18793 Alan Hayward <alan.hayward@arm.com>
18794 David Sherwood <david.sherwood@arm.com>
18796 * lra-constraints.c (process_alt_operands): Reject matched
18797 operands whose sizes aren't ordered.
18798 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18805 that the mode size is in the set {1, 2, 4, 8, 16}.
18807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18808 Alan Hayward <alan.hayward@arm.com>
18809 David Sherwood <david.sherwood@arm.com>
18811 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18812 Use plus_constant instead of gen_rtx_PLUS.
18814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18815 Alan Hayward <alan.hayward@arm.com>
18816 David Sherwood <david.sherwood@arm.com>
18818 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18819 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18820 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18821 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18822 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18823 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18824 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18825 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18826 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18827 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18829 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18830 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18831 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18832 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18833 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18834 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18835 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18836 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18837 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18838 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18840 * expr.c (emit_move_resolve_push): Treat the input and result
18841 of PUSH_ROUNDING as a poly_int64.
18842 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18843 (emit_push_insn): Likewise.
18844 * lra-eliminations.c (mark_not_eliminable): Likewise.
18845 * recog.c (push_operand): Likewise.
18846 * reload1.c (elimination_effects): Likewise.
18847 * rtlanal.c (nonzero_bits1): Likewise.
18848 * calls.c (store_one_arg): Likewise. Require the padding to be
18849 known at compile time.
18851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18852 Alan Hayward <alan.hayward@arm.com>
18853 David Sherwood <david.sherwood@arm.com>
18855 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18856 Use plus_constant instead of gen_rtx_PLUS.
18858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18859 Alan Hayward <alan.hayward@arm.com>
18860 David Sherwood <david.sherwood@arm.com>
18862 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18863 rather than an int.
18865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18866 Alan Hayward <alan.hayward@arm.com>
18867 David Sherwood <david.sherwood@arm.com>
18869 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18870 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18871 via stack temporaries. Treat the mode size as polynomial too.
18873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18874 Alan Hayward <alan.hayward@arm.com>
18875 David Sherwood <david.sherwood@arm.com>
18877 * expr.c (expand_expr_real_2): When handling conversions involving
18878 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18879 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18880 as a poly_uint64 too.
18882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18883 Alan Hayward <alan.hayward@arm.com>
18884 David Sherwood <david.sherwood@arm.com>
18886 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18889 Alan Hayward <alan.hayward@arm.com>
18890 David Sherwood <david.sherwood@arm.com>
18892 * combine.c (can_change_dest_mode): Handle polynomial
18893 REGMODE_NATURAL_SIZE.
18894 * expmed.c (store_bit_field_1): Likewise.
18895 * expr.c (store_constructor): Likewise.
18896 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18897 and polynomial REGMODE_NATURAL_SIZE.
18898 (gen_lowpart_common): Likewise.
18899 * reginfo.c (record_subregs_of_mode): Likewise.
18900 * rtlanal.c (read_modify_subreg_p): Likewise.
18902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18903 Alan Hayward <alan.hayward@arm.com>
18904 David Sherwood <david.sherwood@arm.com>
18906 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18907 numbers of elements.
18909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18910 Alan Hayward <alan.hayward@arm.com>
18911 David Sherwood <david.sherwood@arm.com>
18913 * match.pd: Cope with polynomial numbers of vector elements.
18915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18916 Alan Hayward <alan.hayward@arm.com>
18917 David Sherwood <david.sherwood@arm.com>
18919 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18920 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
18927 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18928 (ipa_simd_modify_function_body): Likewise.
18930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18931 Alan Hayward <alan.hayward@arm.com>
18932 David Sherwood <david.sherwood@arm.com>
18934 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18935 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18936 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18937 (expand_vector_condition, vector_element): Likewise.
18938 (subparts_gt): New function.
18939 (get_compute_type): Use subparts_gt.
18940 (count_type_subparts): Delete.
18941 (expand_vector_operations_1): Use subparts_gt instead of
18942 count_type_subparts.
18944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18945 Alan Hayward <alan.hayward@arm.com>
18946 David Sherwood <david.sherwood@arm.com>
18948 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18949 (vect_compile_time_alias): ...this new function. Do the calculation
18950 on poly_ints rather than trees.
18951 (vect_prune_runtime_alias_test_list): Update call accordingly.
18953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18954 Alan Hayward <alan.hayward@arm.com>
18955 David Sherwood <david.sherwood@arm.com>
18957 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18959 (vect_schedule_slp_instance): Likewise.
18961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18962 Alan Hayward <alan.hayward@arm.com>
18963 David Sherwood <david.sherwood@arm.com>
18965 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18966 constant and extern definitions for variable-length vectors.
18967 (vect_get_constant_vectors): Note that the number of units
18968 is known to be constant.
18970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18971 Alan Hayward <alan.hayward@arm.com>
18972 David Sherwood <david.sherwood@arm.com>
18974 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18975 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
18983 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18986 Alan Hayward <alan.hayward@arm.com>
18987 David Sherwood <david.sherwood@arm.com>
18989 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18990 vectors as polynomial. Use build_index_vector for
18991 IFN_GOMP_SIMD_LANE.
18993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18994 Alan Hayward <alan.hayward@arm.com>
18995 David Sherwood <david.sherwood@arm.com>
18997 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18998 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18999 for variable-length vectors.
19000 (vectorizable_mask_load_store): Treat the number of units as
19001 polynomial, asserting that it is constant if the condition has
19002 already been enforced.
19003 (vectorizable_store, vectorizable_load): Likewise.
19005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19006 Alan Hayward <alan.hayward@arm.com>
19007 David Sherwood <david.sherwood@arm.com>
19009 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19010 of units as polynomial. Punt if we can't tell at compile time
19011 which vector contains the final result.
19013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19014 Alan Hayward <alan.hayward@arm.com>
19015 David Sherwood <david.sherwood@arm.com>
19017 * tree-vect-loop.c (vectorizable_induction): Treat the number
19018 of units as polynomial. Punt on SLP inductions. Use an integer
19019 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19020 cast of such a series for variable-length floating-point
19023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19024 Alan Hayward <alan.hayward@arm.com>
19025 David Sherwood <david.sherwood@arm.com>
19027 * tree.h (build_index_vector): Declare.
19028 * tree.c (build_index_vector): New function.
19029 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19030 of units as polynomial, forcibly converting it to a constant if
19031 vectorizable_reduction has already enforced the condition.
19032 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19033 to create a {1,2,3,...} vector.
19034 (vectorizable_reduction): Treat the number of units as polynomial.
19035 Choose vectype_in based on the largest scalar element size rather
19036 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
19044 number of units as polynomial.
19046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19047 Alan Hayward <alan.hayward@arm.com>
19048 David Sherwood <david.sherwood@arm.com>
19050 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19051 * target.def (autovectorize_vector_sizes): Return the vector sizes
19052 by pointer, using vector_sizes rather than a bitmask.
19053 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19054 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19055 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19057 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19058 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19059 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19060 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19061 * omp-general.c (omp_max_vf): Likewise.
19062 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19063 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19064 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19065 * tree-vect-slp.c (vect_slp_bb): Likewise.
19066 * doc/tm.texi: Regenerate.
19067 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19069 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19070 the vector size as a poly_uint64 rather than an unsigned int.
19071 (current_vector_size): Change from an unsigned int to a poly_uint64.
19072 (get_vectype_for_scalar_type): Update accordingly.
19073 * tree.h (build_truth_vector_type): Take the size and number of
19074 units as a poly_uint64 rather than an unsigned int.
19075 (build_vector_type): Add a temporary overload that takes
19076 the number of units as a poly_uint64 rather than an unsigned int.
19077 * tree.c (make_vector_type): Likewise.
19078 (build_truth_vector_type): Take the number of units as a poly_uint64
19079 rather than an unsigned int.
19081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19082 Alan Hayward <alan.hayward@arm.com>
19083 David Sherwood <david.sherwood@arm.com>
19085 * target.def (get_mask_mode): Take the number of units and length
19086 as poly_uint64s rather than unsigned ints.
19087 * targhooks.h (default_get_mask_mode): Update accordingly.
19088 * targhooks.c (default_get_mask_mode): Likewise.
19089 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19090 * doc/tm.texi: Regenerate.
19092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19093 Alan Hayward <alan.hayward@arm.com>
19094 David Sherwood <david.sherwood@arm.com>
19096 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19097 * omp-general.c (omp_max_vf): Likewise.
19098 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19099 (expand_omp_simd): Handle polynomial safelen.
19100 * omp-low.c (omplow_simd_context): Add a default constructor.
19101 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19102 (lower_rec_simd_input_clauses): Update accordingly.
19103 (lower_rec_input_clauses): Likewise.
19105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19106 Alan Hayward <alan.hayward@arm.com>
19107 David Sherwood <david.sherwood@arm.com>
19109 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19110 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19111 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19112 (vect_analyze_slp_cost): Likewise.
19113 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19114 (vect_model_load_cost): Likewise.
19116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19117 Alan Hayward <alan.hayward@arm.com>
19118 David Sherwood <david.sherwood@arm.com>
19120 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19121 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19122 from an unsigned int * to a poly_uint64_pod *.
19123 (calculate_unrolling_factor): New function.
19124 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19131 from an unsigned int to a poly_uint64.
19132 (_loop_vec_info::slp_unrolling_factor): Likewise.
19133 (_loop_vec_info::vectorization_factor): Change from an int
19135 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19136 (vect_get_num_vectors): New function.
19137 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19138 (vect_get_num_copies): Use vect_get_num_vectors.
19139 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19140 to an unsigned int *.
19141 (vect_analyze_data_refs): Change min_vf from an int * to a
19143 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19144 than an unsigned HOST_WIDE_INT.
19145 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19146 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19147 to an unsigned int *.
19148 (vect_analyze_data_ref_dependences): Likewise.
19149 (vect_compute_data_ref_alignment): Handle polynomial vf.
19150 (vect_enhance_data_refs_alignment): Likewise.
19151 (vect_prune_runtime_alias_test_list): Likewise.
19152 (vect_shift_permute_load_chain): Likewise.
19153 (vect_supportable_dr_alignment): Likewise.
19154 (dependence_distance_ge_vf): Take the vectorization factor as a
19155 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19156 (vect_analyze_data_refs): Change min_vf from an int * to a
19158 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19159 vfm1 as a poly_uint64 rather than an int. Make the same change
19160 for the returned bound_scalar.
19161 (vect_gen_vector_loop_niters): Handle polynomial vf.
19162 (vect_do_peeling): Likewise. Update call to
19163 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19164 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19166 * tree-vect-loop.c (vect_determine_vectorization_factor)
19167 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19168 (vect_get_known_peeling_cost): Likewise.
19169 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19170 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19171 (vect_transform_loop): Likewise. Use the lowest possible VF when
19172 updating the upper bounds of the loop.
19173 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19174 rather than an int.
19175 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19176 polynomial unroll factors.
19177 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19178 (vect_make_slp_decision): Likewise.
19179 (vect_supported_load_permutation_p): Likewise, and polynomial
19181 (vect_analyze_slp_cost): Handle polynomial vf.
19182 (vect_slp_analyze_node_operations): Likewise.
19183 (vect_slp_analyze_bb_1): Likewise.
19184 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19185 than an unsigned HOST_WIDE_INT.
19186 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19187 (vectorizable_load): Handle polynomial vf.
19188 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19190 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19193 Alan Hayward <alan.hayward@arm.com>
19194 David Sherwood <david.sherwood@arm.com>
19196 * match.pd: Handle bit operations involving three constants
19197 and try to fold one pair.
19199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19201 * tree-vect-loop-manip.c: Include gimple-fold.h.
19202 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19203 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19204 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19205 Add a path that uses a step of VF instead of 1, but disable it
19207 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19208 and niters_no_overflow parameters. Update calls to
19209 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19210 Create a new SSA name if the latter choses to use a ste other
19211 than zero, and return it via niters_vector_mult_vf_var.
19212 * tree-vect-loop.c (vect_transform_loop): Update calls to
19213 vect_do_peeling, vect_gen_vector_loop_niters and
19214 slpeel_make_loop_iterate_ntimes.
19215 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19216 (vect_gen_vector_loop_niters): Update declarations after above changes.
19218 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19220 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19221 128-bit round to integer instructions.
19222 (ceil<mode>2): Likewise.
19223 (btrunc<mode>2): Likewise.
19224 (round<mode>2): Likewise.
19226 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19228 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19229 unaligned VSX load/store on P8/P9.
19230 (expand_block_clear): Allow the use of unaligned VSX
19231 load/store on P8/P9.
19233 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19235 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19237 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19238 swap associated with both a load and a store.
19240 2018-01-02 Andrew Waterman <andrew@sifive.com>
19242 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19243 * config/riscv/riscv.md (clear_cache): Use it.
19245 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19247 * web.c: Remove out-of-date comment.
19249 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19251 * expr.c (fixup_args_size_notes): Check that any existing
19252 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19253 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19254 (emit_single_push_insn): ...here.
19256 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19258 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19259 (const_vector_encoded_nelts): New function.
19260 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19261 (const_vector_int_elt, const_vector_elt): Declare.
19262 * emit-rtl.c (const_vector_int_elt_1): New function.
19263 (const_vector_elt): Likewise.
19264 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19265 of CONST_VECTOR_ELT.
19267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19269 * expr.c: Include rtx-vector-builder.h.
19270 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19271 directly on the tree encoding.
19272 (const_vector_from_tree): Likewise.
19273 * optabs.c: Include rtx-vector-builder.h.
19274 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19275 sequence of "u" values.
19276 * vec-perm-indices.c: Include rtx-vector-builder.h.
19277 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19278 directly on the vec_perm_indices encoding.
19280 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19282 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19283 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19284 * rtx-vector-builder.h: New file.
19285 * rtx-vector-builder.c: Likewise.
19286 * rtl.h (rtx_def::u2): Add a const_vector field.
19287 (CONST_VECTOR_NPATTERNS): New macro.
19288 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19289 (CONST_VECTOR_DUPLICATE_P): Likewise.
19290 (CONST_VECTOR_STEPPED_P): Likewise.
19291 (CONST_VECTOR_ENCODED_ELT): Likewise.
19292 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19293 (unwrap_const_vec_duplicate): Likewise.
19294 (const_vec_series_p): Check for a non-duplicated vector encoding.
19295 Say that the function only returns true for integer vectors.
19296 * emit-rtl.c: Include rtx-vector-builder.h.
19297 (gen_const_vec_duplicate_1): Delete.
19298 (gen_const_vector): Call gen_const_vec_duplicate instead of
19299 gen_const_vec_duplicate_1.
19300 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19301 (gen_const_vec_duplicate): Use rtx_vector_builder.
19302 (gen_const_vec_series): Likewise.
19303 (gen_rtx_CONST_VECTOR): Likewise.
19304 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19305 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19306 Build a new vector rather than modifying a CONST_VECTOR in-place.
19307 (handle_special_swappables): Update call accordingly.
19308 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19309 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19310 Build a new vector rather than modifying a CONST_VECTOR in-place.
19311 (handle_special_swappables): Update call accordingly.
19313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19315 * simplify-rtx.c (simplify_const_binary_operation): Use
19316 CONST_VECTOR_ELT instead of XVECEXP.
19318 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19320 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19321 the selector elements to be different from the data elements
19322 if the selector is a VECTOR_CST.
19323 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19324 ssizetype for the selector.
19326 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19328 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19329 before testing each element individually.
19330 * tree-vect-generic.c (lower_vec_perm): Likewise.
19332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19334 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19335 * selftest-run-tests.c (selftest::run_tests): Call it.
19336 * vector-builder.h (vector_builder::operator ==): New function.
19337 (vector_builder::operator !=): Likewise.
19338 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19339 (vec_perm_indices::all_from_input_p): New function.
19340 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19341 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19342 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19343 instead of reading the VECTOR_CST directly. Detect whether both
19344 vector inputs are the same before constructing the vec_perm_indices,
19345 and update the number of inputs argument accordingly. Use the
19346 utility functions added above. Only construct sel2 if we need to.
19348 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19350 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19351 the broadcast of the low byte.
19352 (expand_mult_highpart): Use an explicit encoding for the permutes.
19353 * optabs-query.c (can_mult_highpart_p): Likewise.
19354 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19355 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19356 (vectorizable_bswap): Likewise.
19357 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19358 explicit encoding for the power-of-2 permutes.
19359 (vect_permute_store_chain): Likewise.
19360 (vect_grouped_load_supported): Likewise.
19361 (vect_permute_load_chain): Likewise.
19363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19365 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19366 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19367 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19368 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19369 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19370 (vect_gen_perm_mask_any): Likewise.
19372 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19374 * int-vector-builder.h: New file.
19375 * vec-perm-indices.h: Include int-vector-builder.h.
19376 (vec_perm_indices): Redefine as an int_vector_builder.
19377 (auto_vec_perm_indices): Delete.
19378 (vec_perm_builder): Redefine as a stand-alone class.
19379 (vec_perm_indices::vec_perm_indices): New function.
19380 (vec_perm_indices::clamp): Likewise.
19381 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19382 (vec_perm_indices::new_vector): New function.
19383 (vec_perm_indices::new_expanded_vector): Update for new
19384 vec_perm_indices class.
19385 (vec_perm_indices::rotate_inputs): New function.
19386 (vec_perm_indices::all_in_range_p): Operate directly on the
19387 encoded form, without computing elided elements.
19388 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19389 encoding. Update for new vec_perm_indices class.
19390 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19391 the given vec_perm_builder.
19392 (expand_vec_perm_var): Update vec_perm_builder constructor.
19393 (expand_mult_highpart): Use vec_perm_builder instead of
19394 auto_vec_perm_indices.
19395 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19396 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19397 or double series encoding as appropriate.
19398 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19399 vec_perm_indices instead of auto_vec_perm_indices.
19400 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19401 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19402 (vect_permute_store_chain): Likewise.
19403 (vect_grouped_load_supported): Likewise.
19404 (vect_permute_load_chain): Likewise.
19405 (vect_shift_permute_load_chain): Likewise.
19406 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19407 (vect_transform_slp_perm_load): Likewise.
19408 (vect_schedule_slp_instance): Likewise.
19409 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19410 (vectorizable_mask_load_store): Likewise.
19411 (vectorizable_bswap): Likewise.
19412 (vectorizable_store): Likewise.
19413 (vectorizable_load): Likewise.
19414 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19415 vec_perm_indices instead of auto_vec_perm_indices. Use
19416 tree_to_vec_perm_builder to read the vector from a tree.
19417 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19418 vec_perm_builder instead of a vec_perm_indices.
19419 (have_whole_vector_shift): Use vec_perm_builder and
19420 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19421 truncation to calc_vec_perm_mask_for_shift.
19422 (vect_create_epilog_for_reduction): Likewise.
19423 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19424 from auto_vec_perm_indices to vec_perm_indices.
19425 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19426 instead of changing individual elements.
19427 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19428 the vector in d.perm.
19429 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19430 from auto_vec_perm_indices to vec_perm_indices.
19431 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19432 instead of changing individual elements.
19433 (arm_vectorize_vec_perm_const): Use new_vector to install
19434 the vector in d.perm.
19435 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19436 Update vec_perm_builder constructor.
19437 (rs6000_expand_interleave): Likewise.
19438 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19439 (rs6000_expand_interleave): Likewise.
19441 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19443 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19444 to qimode could truncate the indices.
19445 * optabs.c (expand_vec_perm_var): Likewise.
19447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19449 * Makefile.in (OBJS): Add vec-perm-indices.o.
19450 * vec-perm-indices.h: New file.
19451 * vec-perm-indices.c: Likewise.
19452 * target.h (vec_perm_indices): Replace with a forward class
19454 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19455 * optabs.h: Include vec-perm-indices.h.
19456 (expand_vec_perm): Delete.
19457 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19458 (expand_vec_perm_const): Declare.
19459 * target.def (vec_perm_const_ok): Replace with...
19460 (vec_perm_const): ...this new hook.
19461 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19462 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19463 * doc/tm.texi: Regenerate.
19464 * optabs.def (vec_perm_const): Delete.
19465 * doc/md.texi (vec_perm_const): Likewise.
19466 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19467 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19468 expand_vec_perm for constant permutation vectors. Assert that
19469 the mode of variable permutation vectors is the integer equivalent
19470 of the mode that is being permuted.
19471 * optabs-query.h (selector_fits_mode_p): Declare.
19472 * optabs-query.c: Include vec-perm-indices.h.
19473 (selector_fits_mode_p): New function.
19474 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19475 is defined, instead of checking whether the vec_perm_const_optab
19476 exists. Use targetm.vectorize.vec_perm_const instead of
19477 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19478 fit in the vector mode before using a variable permute.
19479 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19480 vec_perm_indices instead of an rtx.
19481 (expand_vec_perm): Replace with...
19482 (expand_vec_perm_const): ...this new function. Take the selector
19483 as a vec_perm_indices rather than an rtx. Also take the mode of
19484 the selector. Update call to shift_amt_for_vec_perm_mask.
19485 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19486 Use vec_perm_indices::new_expanded_vector to expand the original
19487 selector into bytes. Check whether the indices fit in the vector
19488 mode before using a variable permute.
19489 (expand_vec_perm_var): Make global.
19490 (expand_mult_highpart): Use expand_vec_perm_const.
19491 * fold-const.c: Includes vec-perm-indices.h.
19492 * tree-ssa-forwprop.c: Likewise.
19493 * tree-vect-data-refs.c: Likewise.
19494 * tree-vect-generic.c: Likewise.
19495 * tree-vect-loop.c: Likewise.
19496 * tree-vect-slp.c: Likewise.
19497 * tree-vect-stmts.c: Likewise.
19498 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19500 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19501 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19502 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19503 (aarch64_vectorize_vec_perm_const): ...this new function.
19504 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19505 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19506 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19507 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19508 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19509 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19510 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19512 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19513 check for NEON modes.
19514 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19515 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19516 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19517 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19519 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19520 the old VEC_PERM_CONST conditions.
19521 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19522 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19523 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19524 (ia64_vectorize_vec_perm_const_ok): Merge into...
19525 (ia64_vectorize_vec_perm_const): ...this new function.
19526 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19527 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19528 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19529 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19530 * config/mips/mips.c (mips_expand_vec_perm_const)
19531 (mips_vectorize_vec_perm_const_ok): Merge into...
19532 (mips_vectorize_vec_perm_const): ...this new function.
19533 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19534 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19535 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19536 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19537 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19538 (rs6000_expand_vec_perm_const): Delete.
19539 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19541 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19542 (altivec_expand_vec_perm_const_le): Take each operand individually.
19543 Operate on constant selectors rather than rtxes.
19544 (altivec_expand_vec_perm_const): Likewise. Update call to
19545 altivec_expand_vec_perm_const_le.
19546 (rs6000_expand_vec_perm_const): Delete.
19547 (rs6000_vectorize_vec_perm_const_ok): Delete.
19548 (rs6000_vectorize_vec_perm_const): New function.
19549 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19550 an element count and rtx array.
19551 (rs6000_expand_extract_even): Update call accordingly.
19552 (rs6000_expand_interleave): Likewise.
19553 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19554 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19555 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19556 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19557 (rs6000_expand_vec_perm_const): Delete.
19558 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19559 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19560 (altivec_expand_vec_perm_const_le): Take each operand individually.
19561 Operate on constant selectors rather than rtxes.
19562 (altivec_expand_vec_perm_const): Likewise. Update call to
19563 altivec_expand_vec_perm_const_le.
19564 (rs6000_expand_vec_perm_const): Delete.
19565 (rs6000_vectorize_vec_perm_const_ok): Delete.
19566 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19567 reference to the SPE evmerge intructions.
19568 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19569 an element count and rtx array.
19570 (rs6000_expand_extract_even): Update call accordingly.
19571 (rs6000_expand_interleave): Likewise.
19572 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19573 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19575 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19579 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19580 vector mode and that that mode matches the mode of the data
19582 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19583 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19584 directly using expand_vec_perm_1 when forcing selectors into
19586 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19590 * optabs-query.h (can_vec_perm_p): Delete.
19591 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19592 * optabs-query.c (can_vec_perm_p): Split into...
19593 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19594 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19595 particular selector is valid.
19596 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19597 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19598 (vect_grouped_load_supported): Likewise.
19599 (vect_shift_permute_load_chain): Likewise.
19600 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19601 (vect_transform_slp_perm_load): Likewise.
19602 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19603 (vectorizable_bswap): Likewise.
19604 (vect_gen_perm_mask_checked): Likewise.
19605 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19606 implementations of variable permutation vectors into account
19607 when deciding which selector to use.
19608 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19609 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19610 with a false third argument.
19611 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19612 to test whether the constant selector is valid and can_vec_perm_var_p
19613 to test whether a variable selector is valid.
19615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19617 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19618 * optabs-query.c (can_vec_perm_p): Likewise.
19619 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19620 instead of vec_perm_indices.
19621 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19622 (vect_gen_perm_mask_checked): Likewise,
19623 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19624 (vect_gen_perm_mask_checked): Likewise,
19626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19628 * optabs-query.h (qimode_for_vec_perm): Declare.
19629 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19630 (qimode_for_vec_perm): ...this new function.
19631 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19633 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19635 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19636 does not have a conditional at the top.
19638 2018-01-02 Richard Biener <rguenther@suse.de>
19640 * ipa-inline.c (big_speedup_p): Fix expression.
19642 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19645 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19648 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19652 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19653 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19654 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19655 cond_taken_branch_cost 3->4.
19657 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19659 PR tree-optimization/83581
19660 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19661 TODO_cleanup_cfg if any changes have been made.
19663 PR middle-end/83608
19664 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19665 convert_modes if target mode has the right side, but different mode
19668 PR middle-end/83609
19669 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19670 last argument when extracting from CONCAT. If either from_real or
19671 from_imag is NULL, use expansion through memory. If result is not
19672 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19673 the parts directly to inner mode, if even that fails, use expansion
19676 PR middle-end/83623
19677 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19678 check for bswap in mode rather than HImode and use that in expand_unop
19681 Copyright (C) 2018 Free Software Foundation, Inc.
19683 Copying and distribution of this file, with or without modification,
19684 are permitted in any medium without royalty provided the copyright
19685 notice and this notice are preserved.