1 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2 Alan Hayward <alan.hayward@arm.com>
3 David Sherwood <david.sherwood@arm.com>
5 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
7 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8 plus and minus and convert them into IFN_COND_FMA-based sequences.
9 (convert_mult_to_fma): Handle conditional plus and minus.
11 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
13 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
14 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
15 (cond_fnms_optab): New optabs.
16 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
18 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
19 * internal-fn.h (get_conditional_internal_fn): Declare.
20 (get_unconditional_internal_fn): Likewise.
21 * internal-fn.c (cond_ternary_direct): New macro.
22 (expand_cond_ternary_optab_fn): Likewise.
23 (direct_cond_ternary_optab_supported_p): Likewise.
24 (FOR_EACH_COND_FN_PAIR): Likewise.
25 (get_conditional_internal_fn): New function.
26 (get_unconditional_internal_fn): Likewise.
27 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
28 (gimple_match_op::gimple_match_op): Add a new overload for 5
30 (gimple_match_op::set_op): Likewise.
31 (gimple_resimplify5): Declare.
32 * genmatch.c (decision_tree::gen): Generate simplifications for
34 * gimple-match-head.c (gimple_simplify): Define an overload for
35 5 operands. Handle calls with 5 arguments in the top-level overload.
36 (convert_conditional_op): Handle conversions from unconditional
37 internal functions to conditional ones.
38 (gimple_resimplify5): New function.
39 (build_call_internal): Pass a fifth operand.
40 (maybe_push_res_to_seq): Likewise.
41 (try_conditional_simplification): Try converting conditional
42 internal functions to unconditional internal functions.
43 Handle 3-operand unconditional forms.
44 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
45 Define ternary equivalents of the current rules for binary conditional
47 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
49 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
50 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
52 (SVE_COND_FP_TERNARY): New int iterator.
53 (sve_fmla_op, sve_fmad_op): New int attributes.
54 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
55 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
56 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
58 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
60 * target.def (preferred_else_value): New target hook.
61 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
62 * doc/tm.texi: Regenerate.
63 * targhooks.h (default_preferred_else_value): Declare.
64 * targhooks.c (default_preferred_else_value): New function.
65 * internal-fn.h (conditional_internal_fn_code): Declare.
66 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
67 (get_conditional_internal_fn): Use it.
68 (conditional_internal_fn_code): New function.
69 * gimple-match.h (gimple_match_cond): New struct.
70 (gimple_match_op): Add a cond member function.
71 (gimple_match_op::gimple_match_op): Update all forms to take a
73 * genmatch.c (expr::gen_transform): Use the same condition as res_op
74 for the suboperation, but don't specify a particular else_value.
75 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
76 (visit_nary_op, visit_reference_op_load): Pass
77 gimple_match_cond::UNCOND to the gimple_match_op constructor.
78 * gimple-match-head.c: Include tree-eh.h
79 (convert_conditional_op): New function.
80 (maybe_resimplify_conditional_op): Likewise.
81 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
82 (gimple_resimplify2): Likewise.
83 (gimple_resimplify3): Likewise.
84 (gimple_resimplify4): Likewise.
85 (maybe_push_res_to_seq): Return null for conditional operations.
86 (try_conditional_simplification): New function.
87 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
89 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
91 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
93 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
95 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
97 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
99 (hash_tree): Do not hash DECL_FCONTEXT
100 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
101 Do not stream DECL_FCONTEXT.
102 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
103 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
105 2018-07-12 Richard Biener <rguenther@suse.de>
108 * dwarf2out.c (gen_block_die): Only output blocks when they have
109 at least one !DECL_IGNORED_P variable.
111 2018-07-12 Richard Biener <rguenther@suse.de>
114 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
115 Remove -mieee-fp handling.
117 2018-07-12 Richard Biener <rguenther@suse.de>
119 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
120 left-over from last patch.
122 2018-07-12 Jakub Jelinek <jakub@redhat.com>
124 PR tree-optimization/86492
125 * gimple-ssa-store-merging.c
126 (imm_store_chain_info::coalesce_immediate_stores): Call
127 check_no_overlap even for the merge_overlapping case. Formatting fix.
129 2018-07-12 Richard Biener <rguenther@suse.de>
132 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
133 move possibly trapping operations into the conditional.
135 2018-07-12 Richard Biener <rguenther@suse.de>
137 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
138 (vn_lookup_simplify_result): Remove recursion limit applied
140 (vn_nary_build_or_lookup_1): Adjust.
141 (try_to_simplify): Likewise.
142 * gimple-match-head.c (gimple_resimplify1): Instead apply one
144 (gimple_resimplify2): Likewise.
145 (gimple_resimplify3): Likewise.
146 (gimple_resimplify4): Likewise.
148 2018-07-11 Jakub Jelinek <jakub@redhat.com>
150 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
151 Use __mmask64 type instead of __mmask8 for __M argument.
152 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
153 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
155 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
156 __mmask16 for __M argument.
157 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
158 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
159 to __mmask16 instead of __mmask8.
160 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
161 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
162 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
163 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
164 instead of __mmask16 for __U argument.
165 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
166 __mmask16 instead of __mmask8 for __U argument.
167 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
169 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
171 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
173 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
175 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
177 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
179 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
181 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
183 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
185 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
187 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
188 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
189 return type as well as __M argument type and all casts from __mmask8
191 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
192 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
193 return type as well as __M argument type and all casts from __mmask8
195 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
196 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
197 return type as well as __M argument type and all casts from __mmask8
199 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
200 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
201 return type as well as __M argument type and all casts from __mmask8
203 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
204 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
207 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
209 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
210 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
213 2018-07-11 Paul Koning <ni1d@arrl.net>
215 * doc/md.texi (define_subst): Document how multiple occurrences of
216 the same argument in the replacement pattern are handled.
218 2018-07-11 Paul Koning <ni1d@arrl.net>
220 * doc/extend.texi (Common Variable Attributes): Move "mode" into
222 (Common Type Attributes): Add "mode" attribute.
224 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
226 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
227 stream DECL_ORIGINAL_TYPE.
228 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
229 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
230 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
231 Do not walk original type.
232 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
234 (write_ts_decl_non_common_tree_pointers): Do not stream
236 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
237 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
239 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
241 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
242 thread twice from the same starting edge.
244 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
246 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
247 * gimple.c (gimple_call_nonnull_result_p): ...here...
248 (gimple_call_nonnull_arg): ...and here.
249 * gimple.h (gimple_call_nonnull_result_p): New.
250 (gimple_call_nonnull_arg): New.
252 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
254 * config/arm/arm-cpus.in: Move information from fpu field of each
255 cpu definition to the isa field.
256 * config/arm/parsecpu.awk (fpu): Delete match rule.
257 (gen_comm_data): Don't add bits from the CPU's FPU entry.
259 2018-07-11 Richard Biener <rguenther@suse.de>
262 * dwarf2out.c (init_sections_and_labels): Use
263 output_asm_line_debug_info consistently.
264 (dwarf2out_early_finish): Likewise.
265 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
268 2018-07-11 Richard Biener <rguenther@suse.de>
270 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
271 Rework father_bb setting in a way to avoid propagating constants
272 multiple times on a loop body.
274 2018-07-10 Mark Wielaard <mark@klomp.org>
277 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
279 2018-07-10 Richard Biener <rguenther@suse.de>
281 * hash-map.h (hash_map::iterator::operator*): Return
282 references to key and value.
284 2018-07-10 Jakub Jelinek <jakub@redhat.com>
287 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
288 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
289 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
290 (gimplify_omp_for): For composite loops, move outer
291 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
292 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
293 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
294 TREE_LIST for both the original class iterator and the "last" helper
295 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
296 loop, remember has_decl_expr from outer composite loops for the
297 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
299 2018-07-09 Martin Sebor <msebor@redhat.com>
303 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
304 accessing implicitly initialized array elements.
305 * expr.c (string_constant): Handle string initializers of
306 character arrays within aggregates.
307 * gimple-fold.c (fold_array_ctor_reference): Add argument.
308 Store element offset. As a special case, handle zero size.
309 (fold_nonarray_ctor_reference): Same.
310 (fold_ctor_reference): Add argument. Store subobject offset.
311 * gimple-fold.h (fold_ctor_reference): Add argument.
313 2018-07-09 Paul Koning <ni1d@arrl.net>
315 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
316 (pdp11_insn_cost): New function.
317 (pdp11_md_asm_adjust): New function.
318 (TARGET_INVALID_WITHIN_DOLOOP): Define.
319 (pdp11_rtx_costs): Update to match machine better.
320 (output_addr_const_pdp11): Correct format mismatch warnings.
321 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
322 * config/pdp11/pdp11.md: General change to add base_cost and/or
323 length attributes for use by new pdp11_insn_cost function.
324 (MIN_BRANCH): Correct definition.
326 (doloop_end): Use standard pattern name for looping pattern.
327 (doloop_end_nocc): New.
328 (movsf): Add another constraint alternative.
329 (zero_extendqihi2): Add constraint alternatives for not in place
331 (zero_extendhisi2): Remove.
332 (shift patterns): Add CC handling variants.
336 (define_peephole2): New peephole to recognize mov that sets CC for
339 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
341 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
342 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
345 2018-07-09 Jakub Jelinek <jakub@redhat.com>
348 * real.c (real_nextafter): Return true if result is denormal.
350 2018-07-09 Martin Liska <mliska@suse.cz>
352 * common.opt: Add back wrongly removed attribute.
354 2018-07-09 Richard Biener <rguenther@suse.de>
357 * dwarf2out.c (gen_block_die): For an early generated DIE
358 always output high/low PC attributes.
360 2018-07-09 Tom de Vries <tdevries@suse.de>
362 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
363 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
364 onto VAR_DECL with abstract origin.
366 2018-07-07 Jim Wilson <jimw@sifive.com>
368 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
370 2018-07-07 Tom de Vries <tdevries@suse.de>
372 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
374 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
376 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
378 (combine_bound): Use wide-int overflow calculation instead of
380 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
382 * fold-const.c (int_const_binop_2): Same.
383 (extract_muldiv_1): Same.
384 (fold_div_compare): Same.
385 (fold_abs_const): Same.
387 * poly-int.h (add): Same.
391 * predict.c (predict_iv_comparison): Same.
392 * profile-count.c (slow_safe_scale_64bit): Same.
393 * simplify-rtx.c (simplify_const_binary_operation): Same.
394 * tree-chrec.c (tree_fold_binomial): Same.
395 * tree-data-ref.c (split_constant_offset_1): Same.
396 * tree-if-conv.c (idx_within_array_bound): Same.
397 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
398 * tree-ssa-phiopt.c (minmax_replacement): Same.
399 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
400 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
401 * vr-values.c (vr_values::adjust_range_with_scev): Same.
402 * wide-int.cc (wi::add_large): Same.
403 (wi::mul_internal): Same.
404 (wi::sub_large): Same.
405 (wi::divmod_internal): Same.
406 * wide-int.h: Change overflow type to overflow_type for neg, add,
407 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
408 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
409 mul_internal, divmod_internal.
410 (overflow_type): New enum.
411 (accumulate_overflow): New.
413 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
415 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
416 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
418 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
420 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
421 argument is checked for zero before entering loop, avoid checking again.
423 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
425 * gimplify.h (generic_expr_could_trap_p): Set as global function.
426 * gimplify.h (generic_expr_could_trap_p): Likwise.
427 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
429 2018-07-06 Jakub Jelinek <jakub@redhat.com>
431 PR tree-optimization/86401
432 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
433 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
434 (fold_bit_and_mask): ... here. New helper function for match.pd.
435 * fold-const.h (fold_bit_and_mask): Declare.
436 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
438 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
441 * target.def (translate_mode_attribute): New hook.
442 * targhooks.h (default_translate_mode_attribute): Declare.
443 * targhooks.c (default_translate_mode_attribute): New function.
444 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
445 * doc/tm.texi: Regenerate.
446 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
447 (rs6000_translate_mode_attribute): New function.
449 2018-07-06 Paul Koning <ni1d@arrl.net>
451 * doc/md.texi (define_split): Document DONE and FAIL.
452 (define_peephole2): Ditto.
454 2018-07-05 Jeff Law <law@redhat.com>
456 PR tree-optimization/86010
457 * tree-ssa-dse.c (compute_trims): More aggressively trim at
458 both the head and tail of mem* and str* calls.
460 2018-07-05 Jim Wilson <jimw@sifive.com>
462 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
464 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
466 * config/aarch64/aarch64-simd.md: correct flags text for
467 MIN_EXPR replacement.
469 2018-07-05 James Clarke <jrtc27@jrtc27.com>
471 * configure: Regenerated.
473 2018-07-05 Carl Love <cel@us.ibm.com>
475 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
476 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
477 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
478 VSX_BUILTIN_DOUBLEL_V4SF.
480 2018-07-05 Martin Sebor <msebor@redhat.com>
483 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
484 than its domain to compute its the upper bound of a char array.
486 2018-07-05 Nathan Sidwell <nathan@acm.org>
488 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
489 * doc/cpp.texi: Update comment.
490 * doc/tm.texi: Rebuilt.
491 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
492 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
493 * doc/extend.texi (Backwards Compatibility): Clarify it is system
494 headers affected by extern "C".
495 * system.h: Poison NO_IMPLICIT_EXTERN_C.
496 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
497 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
498 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
499 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
500 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
501 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
502 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
503 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
504 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
505 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
506 config/sparc/sp64-elf.h, config/spu/spu.h,
507 config/stormy16/stormy16.h, config/v850/v850.h,
508 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
509 define NO_IMPLICIT_EXTERN_C.
510 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
512 2018-07-05 Tamar Christina <tamar.christina@arm.com>
515 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
516 instead of GET_MODE_SIZE when comparing Units.
518 2018-07-05 Tamar Christina <tamar.christina@arm.com>
521 * rtlanal.c (set_noop_p): Constrain on mode change,
522 include hard-reg-set.h
524 2018-07-05 Tamar Christina <tamar.christina@arm.com>
526 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
528 2018-07-05 Jakub Jelinek <jakub@redhat.com>
531 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
534 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
536 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
538 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
541 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
543 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
545 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
548 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
550 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
552 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
555 2018-07-04 Martin Liska <mliska@suse.cz>
557 * tree-switch-conversion.c: Define
558 max_ratio_for_speed and max_ratio_for_size constants.
560 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
561 Martin Liska <mliska@suse.cz>
566 * common.opt: Rename align options with 'str_' prefix.
567 * common/config/i386/i386-common.c (set_malign_value): New
569 (ix86_handle_option): Use it to set -falign-* options/
570 * config/aarch64/aarch64-protos.h (struct tune_params): Change
571 type from int to string.
572 * config/aarch64/aarch64.c: Update default values from int
574 * config/alpha/alpha.c (alpha_override_options_after_change):
576 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
577 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
578 max skip conditionally.
579 * config/i386/freebsd.h (SUBALIGN_LOG): New.
580 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
581 max skip conditionally.
582 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
583 max skip conditionally.
584 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
585 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
586 max skip conditionally.
587 * config/i386/i386.c (struct ptt): Change type from int to
589 (ix86_default_align): Set default values.
590 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
591 max skip conditionally.
592 * config/i386/iamcu.h (SUBALIGN_LOG): New.
593 (ASM_OUTPUT_MAX_SKIP_ALIGN):
594 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
595 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
596 max skip conditionally.
597 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
598 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
599 * config/i386/x86-64.h (SUBALIGN_LOG): New.
600 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
601 max skip conditionally.
602 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
603 * config/ia64/ia64.c (ia64_option_override): Set default values
604 for alignment options.
605 * config/m68k/m68k.c: Handle new str_align_* options.
606 * config/mips/mips.c (mips_set_compression_mode): Change
608 (mips_option_override): Set default values for options.
609 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
611 * config/rs6000/rs6000.c (rs6000_option_override_internal):
613 * config/rx/rx.c (rx_option_override): Likewise.
614 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
615 (LABEL_ALIGN): Use align_labels_log.
616 (LOOP_ALIGN): Use align_loops_align.
617 * config/s390/s390.c (s390_asm_output_function_label): Use new
619 * config/sh/sh.c (sh_override_options_after_change):
620 Change type of constants.
621 * config/spu/spu.c (spu_sched_init): Likewise.
622 * config/sparc/sparc.c (sparc_option_override): Set default
624 * config/visium/visium.c (visium_option_override): Likewise.
625 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
626 emit p2align format with last argument if it's not needed.
627 * doc/invoke.texi: Document extended format of -falign-*.
628 * final.c: Use align_labels alignment.
629 * flags.h (struct target_flag_state): Change type to use
631 (struct align_flags_tuple): New.
632 (struct align_flags): Likewise.
633 (align_loops_log): Redefine macro to use new types.
634 (align_loops_max_skip): Redefine macro to use new types.
635 (align_jumps_log): Redefine macro to use new types.
636 (align_jumps_max_skip): Redefine macro to use new types.
637 (align_labels_log): Redefine macro to use new types.
638 (align_labels_max_skip): Redefine macro to use new types.
639 (align_functions_log): Redefine macro to use new types.
640 (align_loops): Redefine macro to use new types.
641 (align_jumps): Redefine macro to use new types.
642 (align_labels): Redefine macro to use new types.
643 (align_functions): Redefine macro to use new types.
644 (align_functions_max_skip): Redefine macro to use new types.
645 (align_loops_value): New macro.
646 (align_jumps_value): New macro.
647 (align_labels_value): New macro.
648 (align_functions_value): New macro.
649 * function.c (invoke_set_current_function_hook): Propagate
650 alignment values from flags to global variables default in
652 * ipa-icf.c (sem_function::equals_wpa): Use
653 cl_optimization_option_eq instead of memcmp.
654 * lto-streamer.h (cl_optimization_stream_out): Support streaming
656 (cl_optimization_stream_in): Likewise.
657 * optc-save-gen.awk: Support strings in cl_optimization.
658 * opth-gen.awk: Likewise.
659 * opts.c (finish_options): Remove error checking of invalid
661 (MAX_CODE_ALIGN): Remove.
662 (MAX_CODE_ALIGN_VALUE): Likewise.
663 (parse_and_check_align_values): New function.
664 (check_alignment_argument): Likewise.
665 (common_handle_option): Use check_alignment_argument.
666 * opts.h (parse_and_check_align_values): Declare.
667 * toplev.c (init_alignments): Remove.
668 (read_log_maxskip): New.
669 (parse_N_M): Likewise.
670 (parse_alignment_opts): Likewise.
671 (backend_init_target): Remove usage of init_alignments.
672 * toplev.h (parse_alignment_opts): Declare.
673 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
675 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
676 * tree.c (cl_option_hasher::equal): New.
677 * varasm.c: Use new global macros.
679 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
681 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
682 Use a simpler align directive also if MAXSKIP = ALIGN-1.
683 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
684 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
685 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
686 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
687 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
688 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
689 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
690 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
691 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
692 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
693 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
695 2018-07-04 Martin Liska <mliska@suse.cz>
696 Jonathan Wakely <jwakely@redhat.com>
698 * coverage.c: Use correct type.
699 * doc/invoke.texi: Language correction.
701 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
704 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
705 ENDBRANCH for non-tail call which may return via indirect branch.
706 * doc/extend.texi: Document indirect_return attribute.
708 2018-07-03 Martin Sebor <msebor@redhat.com>
710 PR tree-optimization/86274
711 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
713 (format_floating): Correct handling of infinities and NaNs.
715 2018-07-03 Martin Sebor <msebor@redhat.com>
717 * print-tree.c (print_real_cst): New function.
718 (print_node_brief): Call it.
721 2018-07-03 Jeff Law <law@redhat.com>
723 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
724 into a single pattern.
726 * config/h8300/h8300.md (ors code_iterator): New.
727 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
728 a single pattern and single splitter.
729 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
730 (iorqi3_1, xorqi3_1): Likewise.
731 (iorqi3, xorqi3 expanders): Similarly.
733 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
734 (movmd_internal) into a single pattern using the P mode iterator.
735 (movmd splitters): Similarly.
736 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
737 (movsd splitters): Similarly.
739 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
740 ADDB, ADDW and ADDL into a single ADD attribute which selects the
741 right table based on the size of the operand.
742 * config/h8300/h8300.md (length_table): Corresponding changes. All
743 references to "addb", "addw" and "addl" changed to "add".
744 (btst patterns): Merge two variants into a single pattern.
745 (tstqi, tsthi): Likewise.
746 (addhi3_incdec, addsi3_incdec): Likewise.
747 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
748 (mulhi3, mulsi3): Likewise.
749 (udivhi3, udivsi3): Likewise.
750 (divhi3, divsi3): Likewise.
751 (andorqi3, andorhi3, andorsi3): Likewise.
753 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
756 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
757 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
759 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
761 PR tree-optimization/85694
762 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
763 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
765 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
766 (<u>avg<mode>3_ceil): New patterns.
768 2018-07-03 David Malcolm <dmalcolm@redhat.com>
770 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
771 scan-tree-dump directive.
772 * gcc.dg/vect/slp-perm-2.c: Likewise.
773 * gcc.dg/vect/slp-perm-3.c: Likewise.
774 * gcc.dg/vect/slp-perm-5.c: Likewise.
775 * gcc.dg/vect/slp-perm-6.c: Likewise.
776 * gcc.dg/vect/slp-perm-7.c: Likewise.
777 * gcc.dg/vect/slp-perm-8.c: Likewise.
779 2018-07-03 Marek Polacek <polacek@redhat.com>
782 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
784 2018-07-03 Richard Biener <rguenther@suse.de>
787 * tree-ssa-structalias.c (find_func_clobbers): Properly
788 handle indirect calls.
790 2018-07-03 Jeff Law <law@redhat.com>
792 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
793 (shifts): New code iterator.
794 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
795 expander. Fix HImode handling on H8/SX.
796 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
797 (subqi3, subhi3, subsi3 expanders): Likewise.
798 (andqi3, andhi3, andsi3 expanders): Likewise.
799 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
800 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
801 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
802 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
803 (zero_extendqihi2, zero_extendqisi2): Likewise.
804 (extendqihi2, extendqisi2): Likewise.
805 (rotlqi3, rotlhi3, rotlsi3): Likewise.
806 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
807 (rotlqi3_1, rotlhi3_1): Likewise.
808 (logicalhi3_sn, logicalsi3_sn): Likewise.
809 (logicalhi3, logicalsi3): Likewise.
811 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
813 * tree-vect-patterns.c (vect_recog_rotate_pattern)
814 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
815 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
816 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
817 type to append_pattern_def_seq instead of creating a stmt_vec_info
819 (build_mask_conversion): Likewise. Remove vinfo argument.
820 (vect_add_conversion_to_patterm): Likewise, renaming to...
821 (vect_add_conversion_to_pattern): ...this.
822 (vect_recog_mask_conversion_pattern): Update call to
823 build_mask_conversion. Pass the vector type to
824 append_pattern_def_seq here too.
825 (vect_recog_gather_scatter_pattern): Update call to
826 vect_add_conversion_to_pattern.
828 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
830 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
831 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
832 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
833 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
834 STMT_VINFO_PATTERN_DEF_SEQ to null here.
835 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
836 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
837 append_pattern_def_seq instead of new_pattern_def_seq.
838 (vect_recog_divmod_pattern): Do both of the above.
839 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
842 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
844 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
845 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
846 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
847 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
848 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
849 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
850 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
851 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
852 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
853 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
854 parameter with a single stmt_vec_info.
855 (vect_recog_func_ptr): Likewise.
856 (vect_recog_gather_scatter_pattern): Likewise, folding in...
857 (vect_try_gather_scatter_pattern): ...this.
858 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
859 the stmt_vec_info of the statement to be matched. Don't clear
860 STMT_VINFO_RELATED_STMT.
861 (vect_pattern_recog): Update call accordingly.
863 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
865 PR tree-optimization/85694
866 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
867 (uavgM3_ceil): Document new optabs.
868 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
869 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
871 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
872 (savg_ceil_optab): New optabs.
873 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
874 (vect_vect_recog_func_ptrs): Add it.
875 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
876 constant directly from the associated lhs.
878 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
880 * tree-vect-patterns.c (vect_split_statement): New function.
881 (vect_convert_input): Use it to try to split an existing cast.
883 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
885 * poly-int.h (print_hex): New function.
886 * dumpfile.h (dump_dec, dump_hex): Declare.
887 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
888 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
889 min_input_precision, operation_precision and operation_sign.
890 * tree-vect-patterns.c (vect_get_range_info): New function.
891 (vect_same_loop_or_bb_p, vect_single_imm_use)
892 (vect_operation_fits_smaller_type): Delete.
893 (vect_look_through_possible_promotion): Add an optional
894 single_use_p parameter.
895 (vect_recog_over_widening_pattern): Rewrite to use new
896 stmt_vec_info infomration. Handle one operation at a time.
897 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
898 (vect_truncatable_operation_p, vect_set_operation_type)
899 (vect_set_min_input_precision): New functions.
900 (vect_determine_min_output_precision_1): Likewise.
901 (vect_determine_min_output_precision): Likewise.
902 (vect_determine_precisions_from_range): Likewise.
903 (vect_determine_precisions_from_users): Likewise.
904 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
905 (vect_vect_recog_func_ptrs): Put over_widening first.
907 (vect_pattern_recog): Call vect_determine_precisions.
909 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
911 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
912 statements that have been replaced by further pattern statements.
913 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
915 2018-07-03 Richard Biener <rguenther@suse.de>
917 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
918 always set *dt. Dump vectype in vectype overload.
919 * dumpfile.h (dump_gimple_expr): New function.
920 (dump_gimple_expr_loc): Likewise.
921 * dumpfile.c (dump_gimple_expr): New function.
922 (dump_gimple_expr_loc): Likewise.
924 2018-07-02 Jeff Law <law@redhat.com>
926 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
927 the H8/300, H8/300H and H8/S variants into a single pattern.
928 (movhi_h8300, movqi_h8300hs): Similarly.
929 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
930 (QHI mode iterator): New.
932 * config/h8300/h8300.md: Remove trailing whitespace.
934 2018-07-02 Jim Wilson <jimw@sifive.com>
936 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
937 instead of emit_insn for interrupt returns.
938 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
939 (riscv_sret, riscv_uret): Likewise.
941 2018-07-02 David Malcolm <dmalcolm@redhat.com>
943 * pretty-print.c (selftest::test_pp_format): Move save and restore
944 of quotes to class auto_fix_quotes, and add an instance.
945 * selftest.c: Include "intl.h".
946 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
947 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
948 * selftest.h (selftest::auto_fix_quotes): New class.
950 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
952 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
953 (aarch64_sve_prepare_conditional_op): Remove.
954 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
955 Allow aarch64_simd_reg_or_zero as select operand; remove
956 the aarch64_sve_prepare_conditional_op call.
957 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
958 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
959 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
960 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
961 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
962 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
963 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
964 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
965 and a splitters to match all of the *_any patterns.
966 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
968 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
969 (SVE_COND_FP_BINARY_REV): Remove.
970 (sve_int_op_rev, sve_fp_op_rev): New.
971 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
972 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
973 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
974 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
975 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
976 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
977 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
978 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
979 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
981 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
982 Remove match_dup 1 from the inner unspec.
983 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
985 * config/aarch64/aarch64.md (movprfx): New attr.
986 (length): Default movprfx to 8.
987 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
988 (*madd<SVE_I>, *msub<SVE_I): Likewise.
989 (*<su>mul<SVE_I>3_highpart): Likewise.
990 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
991 (*v<ASHIFT><SVE_I>3): Likewise.
992 (*<su><MAXMIN><SVE_I>3): Likewise.
993 (*<su><MAXMIN><SVE_F>3): Likewise.
994 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
995 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
996 (*div<SVE_F>4): Likewise.
998 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1000 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1003 2018-07-02 Richard Biener <rguenther@suse.de>
1005 PR tree-optimization/86363
1006 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1007 memset argument refers to a non-variable address.
1009 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1011 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1012 {PLUS,MINUS}_EXPR code to...
1013 (adjust_symbolic_bound): ...here,
1014 (combine_bound): ...here,
1015 (set_value_range_with_overflow): ...and here.
1017 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1019 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1021 (extract_range_from_abs_expr): ...here.
1023 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1025 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1026 -fno-omit-frame-pointer when not optimizing.
1028 2018-07-02 Martin Liska <mliska@suse.cz>
1031 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1032 (propagate_nothrow): Likewise.
1034 2018-07-02 Martin Liska <mliska@suse.cz>
1037 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1039 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1041 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1042 function in r262149, changing "loc" param from source_location to
1043 const dump_location_t &.
1044 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1045 declaration, as above.
1047 2018-07-01 Paul Koning <ni1d@arrl.net>
1049 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1050 -munit-asm, -mgnu-asm, -mdec-asm.
1051 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1052 (pdp11_output_labelref): New.
1053 (pdp11_output_def): New.
1054 (pdp11_output_addr_vec_elt): New.
1055 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1056 %# and %@ format codes.
1057 (pdp11_option_override): New.
1058 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1059 (pdp11_output_ident): New.
1060 (pdp11_asm_named_section): New.
1061 (pdp11_asm_init_sections): New.
1062 (pdp11_file_start): New.
1063 (pdp11_file_end): New.
1064 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1065 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1066 %o, like %c but octal.
1067 (pdp11_option_override): New.
1068 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1070 (DATA_SECTION_ASM_OP): Ditto.
1071 (READONLY_DATA_SECTION_ASM_OP): New.
1072 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1073 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1074 (ASM_OUTPUT_LABELREF): Ditto.
1075 (ASM_OUTPUT_DEF): Ditto.
1076 (ASM_OUTPUT_EXTERNAL): New.
1077 (ASM_OUTPUT_SOURCE_FILENAME): New.
1078 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1079 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1080 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1081 %# and %@ format codes.
1082 * config/pdp11/pdp11.opt (mgnu-asm): New.
1083 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1084 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1085 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1087 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1089 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1090 dereferencing path[] beyond its length.
1092 (debug_all_paths): New.
1093 (rewire_first_differing_edge): New.
1094 (adjust_paths_after_duplication): New.
1095 (duplicate_thread_path): Call adjust_paths_after_duplication.
1097 (thread_through_all_blocks): Add new argument to
1098 duplicate_thread_path.
1100 2018-06-30 Jim Wilson <jimw@sifive.com>
1102 * config/riscv/predicates.md (p2m1_shift_operand): New.
1103 (high_mask_shift_operand): New.
1104 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1105 pattern using p2m1_shift_operand.
1106 (lshsi3_zero_extend_3+2): New combiner pattern using
1107 high_mask_shift_operand.
1109 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1111 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1113 (vect_recog_rotate_pattern): ...here.
1114 (vect_convert_input): Try to insert casts of invariants in the
1116 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1117 preheader to be empty.
1119 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1121 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1122 vector type. If given, install it in the new statement's
1124 (vect_element_precision): New function.
1125 (vect_unpromoted_value): New struct.
1126 (vect_unpromoted_value::vect_unpromoted_value): New function.
1127 (vect_unpromoted_value::set_op): Likewise.
1128 (vect_look_through_possible_promotion): Likewise.
1129 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1130 (vect_widened_op_tree, vect_convert_input): Likewise.
1131 (vect_convert_inputs, vect_convert_output): Likewise.
1132 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1133 to handle the optional cast of the multiplication result and
1134 vect_widened_op_tree to detect the widened multiplication itself.
1135 Do not require the input and output of promotion casts to have
1136 the same sign, but base the signedness of the operation on the
1137 input rather than the result. If the pattern includes two
1138 promotions, check that those promotions have the same sign.
1139 Do not restrict the MULT_EXPR handling to a double-width result;
1140 handle quadruple-width results and wider. Use vect_convert_inputs
1141 to convert the inputs to the common type.
1142 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1143 to handle the optional cast of the ABS result. Also allow a sign
1144 change or a sign extension between the ABS and MINUS.
1145 Use vect_widened_op_tree to detect the widened subtraction and use
1146 vect_convert_inputs to convert the inputs to the common type.
1147 (vect_handle_widen_op_by_const): Delete.
1148 (vect_recog_widen_op_pattern): New function.
1149 (vect_recog_widen_mult_pattern): Use it.
1150 (vect_recog_widen_shift_pattern): Likewise.
1151 (vect_recog_widen_sum_pattern): Use
1152 vect_look_through_possible_promotion to handle the promoted
1155 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1157 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1158 the containing gimple_seq *.
1159 * gimple-iterator.h (gsi_for_stmt): Declare it.
1160 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1161 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1162 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1163 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1164 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1166 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1168 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1169 statement being replaced is part of an existing pattern
1170 definition sequence, inserting the new pattern statements before
1172 (vect_pattern_recog_1): Don't return a bool. If the statement
1173 is already part of a pattern, instead apply pattern matching
1174 to the pattern definition statements. Don't clear the
1175 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1176 (vect_pattern_recog): Don't break after the first match;
1177 continue processing the pattern definition statements instead.
1178 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1180 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1182 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1183 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1184 (vect_recog_widen_sum_pattern): Use it.
1186 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1188 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1189 phi is not a pattern statement and has not been replaced by
1190 a pattern statement.
1191 * tree-vect-patterns.c (type_conversion_p): Don't check
1192 STMT_VINFO_IN_PATTERN_P.
1193 (vect_recog_vector_vector_shift_pattern): Likewise.
1194 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1195 the pattern statement rather than the original statement; check
1196 directly for a WIDEN_MULT_EXPR here.
1197 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1198 vect_is_simple_use to return the pattern statement rather
1199 than the original statement; use is_pattern_stmt_p to check
1200 for such a pattern statement.
1201 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1202 to return the pattern statement rather than the original statement;
1203 don't do the same transformation here.
1204 (vect_is_simple_use): If the defining statement has been replaced
1205 by a pattern statement, return the pattern statement instead.
1206 Remove the corresponding (local) transformation from the vectype
1209 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1211 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1212 end and default to null.
1213 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1214 (vectorizable_reduction): Update calls accordingly, dropping the
1215 gimple ** argument if the passed-back statement isn't needed.
1216 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1217 (vect_recog_rotate_pattern): Likewise.
1218 (vect_recog_mask_conversion_pattern): Likewise.
1219 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1220 (vect_mask_constant_operand_p): Likewise.
1221 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1222 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1223 (get_group_load_store_type, get_load_store_type): Likewise.
1224 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1225 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1226 (vectorizable_conversion, vectorizable_assignment): Likewise.
1227 (vectorizable_shift, vectorizable_operation): Likewise.
1228 (vectorizable_store, vect_is_simple_cond): Likewise.
1229 (vectorizable_condition, vectorizable_comparison): Likewise.
1230 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1231 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1232 and move it to the end. Cope with null def_stmt_outs.
1234 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1236 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1238 2018-06-29 Jeff Law <law@redhat.com>
1240 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1241 displacements for TARGET_V850E2V3 and newer.
1242 (TARGET_LRA_P): Remove. Defaults to LRA now.
1243 * config/v850/v850.md (sign23byte_load): Remove.
1244 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1245 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1247 2018-06-29 Martin Liska <mliska@suse.cz>
1250 * coverage.c (coverage_init): Mangle full path name.
1251 * doc/invoke.texi: Document the change.
1252 * gcov-io.c (mangle_path): New.
1253 * gcov-io.h (mangle_path): Likewise.
1254 * gcov.c (mangle_name): Use mangle_path for path mangling.
1256 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1258 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1259 if starting source register is not even.
1261 2018-06-29 Martin Liska <mliska@suse.cz>
1263 PR tree-optimization/86263
1264 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1265 Make edge redirection.
1267 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1269 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1270 (dump_scope_depth): New variable.
1271 (get_dump_scope_depth): New function.
1272 (dump_begin_scope): New function.
1273 (dump_end_scope): New function.
1274 * dumpfile.h (get_dump_scope_depth): New declaration.
1275 (dump_begin_scope): New declaration.
1276 (dump_end_scope): New declaration.
1277 (class auto_dump_scope): New class.
1278 (AUTO_DUMP_SCOPE): New macro.
1279 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1282 2018-06-29 Richard Biener <rguenther@suse.de>
1284 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1285 compute_all_dependences succeeds.
1286 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1287 exceed --param loop-max-datarefs-for-datadeps.
1289 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1291 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1293 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1296 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1297 alternative 0 in preferred_for_speed attribute.
1299 2018-06-28 Paul Koning <ni1d@arrl.net>
1301 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1302 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1303 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1304 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1305 for shift insn patterns.
1307 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1309 * cgraph.c (cgraph_node::get_body): Replace assignments to
1310 "dump_file" with calls to set_dump_file.
1311 * dumpfile.c (alt_dump_file): Make static, and group with...
1312 (alt_flags): ...this definition.
1313 (dumps_are_enabled): New variable.
1314 (refresh_dumps_are_enabled): New function.
1315 (set_dump_file): New function.
1316 (set_alt_dump_file): New function.
1317 (gcc::dump_manager::dump_start): Replace assignments to
1318 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1320 (gcc::dump_manager::dump_finish): Likewise.
1321 * dumpfile.h (alt_dump_file): Delete decl.
1322 (dumps_are_enabled): New variable decl.
1323 (set_dump_file): New function decl.
1324 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1326 * tree-nested.c (lower_nested_functions): Replace assignments to
1327 "dump_file" with calls to set_dump_file.
1329 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1331 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1332 goto_locus of each outgoing edge of each basic block.
1334 2018-06-28 Richard Biener <rguenther@suse.de>
1336 * dwarf2out.c (decl_scope_table): Remove.
1337 (push_decl_scope): Likewise.
1338 (pop_decl_scope): Likewise.
1339 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1340 (gen_struct_or_union_type_die): Likewise.
1341 (gen_tagged_type_die): Likewise.
1342 (dwarf2out_init): Do not initialize decl_scope_table.
1343 (dwarf2out_c_finalize): Do not free it.
1345 2018-06-28 Richard Biener <rguenther@suse.de>
1347 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1348 deciding whether to not re-use a DIE.
1350 2018-06-28 Richard Biener <rguenther@suse.de>
1352 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1353 DW_AT_abstract_origin attribute.
1355 2018-06-28 Martin Liska <mliska@suse.cz>
1357 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1358 Use newly introduced constants.
1359 * tree-switch-conversion.h (struct jump_table_cluster):
1360 Define max_ratio_for_size and max_ratio_for_speed.
1362 2018-06-28 Martin Liska <mliska@suse.cz>
1364 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1365 Add new checking assert to catch invalid state.
1366 (jump_table_cluster::can_be_handled): Handle single case
1368 (jump_table_cluster::is_beneficial): Bail out for such case.
1369 (bit_test_cluster::find_bit_tests):
1370 Add new checking assert to catch invalid state.
1371 (bit_test_cluster::can_be_handled): Handle single case
1373 (bit_test_cluster::is_beneficial): Bail out for such case.
1374 (switch_decision_tree::analyze_switch_statement):
1377 2018-06-28 Martin Liska <mliska@suse.cz>
1379 * common.opt: Introduce -completion option.
1380 * gcc.c (driver_handle_option): Handle it.
1381 (driver::main): Print completions if completion
1383 * opt-suggestions.c (option_proposer::get_completions):
1385 (option_proposer::suggest_completion): Likewise.
1386 (option_proposer::find_param_completions): Likewise.
1387 (verify_autocompletions): Likewise.
1388 (test_completion_valid_options): Likewise.
1389 (test_completion_valid_params): Likewise.
1390 (in_completion_p): Likewise.
1391 (empty_completion_p): Likewise.
1392 (test_completion_partial_match): Likewise.
1393 (test_completion_garbage): Likewise.
1394 (opt_proposer_c_tests): Likewise.
1395 * opt-suggestions.h: Declare new functions.
1396 * opts.c (common_handle_option): Handle OPT__completion_.
1397 * selftest-run-tests.c (selftest::run_tests): Add
1398 opt_proposer_c_tests.
1399 * selftest.c (assert_str_startswith): New.
1400 * selftest.h (assert_str_startswith): Likewise.
1401 (opt_proposer_c_tests): New.
1402 (ASSERT_STR_STARTSWITH): Likewise.
1404 2018-06-28 Martin Liska <mliska@suse.cz>
1406 * Makefile.in: Add opt-suggestions.o.
1407 * gcc-main.c: Include opt-suggestions.h.
1408 * gcc.c (driver::driver): Likewise.
1409 (driver::~driver): Remove m_option_suggestions.
1410 (driver::build_option_suggestions): Moved to option_proposer.
1411 (driver::suggest_option): Likewise.
1412 (driver::handle_unrecognized_options): Use option_proposer.
1413 * gcc.h (class driver): Add new memver m_option_proposer.
1414 * opt-suggestions.c: New file.
1415 * opt-suggestions.h: New file.
1417 2018-06-28 Martin Liska <mliska@suse.cz>
1419 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1420 (auto_string_vec::~auto_string_vec): Likewise.
1422 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1424 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1425 prevent_decl_creation_for_types fields up and add reset_location field.
1426 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1427 statement if id->reset_location is true.
1428 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1429 id->reset_location is true.
1430 (copy_phis_for_bb): Force input_location on the arguments if
1431 id->reset_location is true.
1432 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1433 is set on the function to be inlined.
1435 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1437 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1439 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1441 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1442 registers for Pmode.
1443 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1444 hard registers for the clobbered pseudo.
1446 2018-06-27 Paul Koning <ni1d@arrl.net>
1448 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1449 mutually exclusive options.
1450 * config/pdp11/constraints.md (h): New constraint.
1451 (O): Update definition to match shift code generation.
1452 (D): New constraint.
1453 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1455 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1457 (output_jump): Change arguments.
1458 (pdp11_fixed_cc_regs): New function.
1459 (pdp11_cc_mode): Ditto.
1460 (pdp11_expand_shift): Ditto.
1461 (pdp11_assemble_shift): Ditto.
1462 (pdp11_small_shift): Ditto.
1463 (pdp11_branch_cost): Remove.
1464 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1466 (pdp11_register_move_cost): Update for CC registers.
1467 (pdp11_rtx_costs): Add case for LSHIFTRT.
1468 (pdp11_output_jump): Add CCNZ mode conditional branches.
1469 (notice_update_cc_on_set): Remove.
1470 (pdp11_cc_mode): New function.
1471 (simple_memory_operand): Correct pre/post decrement case.
1472 (no_side_effect_operand): New function.
1473 (pdp11_regno_reg_class): Add CC_REGS class.
1474 (pdp11_fixed_cc_regs): New function.
1475 (pdp11_small_shift): New function.
1476 (pdp11_expand_shift): New function to expand shift insns.
1477 (pdp11_assemble_shift): New function to output shifts.
1478 (pdp11_branch_cost): Remove.
1479 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1480 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1481 (WCHAR_TYPE): Ditto.
1482 (PTRDIFF_TYPE): Ditto.
1483 (ADJUST_INSN_LENGTH): New macro.
1484 (FIXED_REGISTERS): Add CC registers.
1485 (CALL_USED_REGISTERS): Ditto.
1487 (REG_CLASS_NAMES): Ditto.
1488 (REG_CLASS_CONTENTS): Ditto.
1489 (SELECT_CC_MODE): Use new function.
1490 (TARGET_FLAGS_REGNUM): New macro.
1491 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1492 (cc0_reg_rtx): Remove.
1493 (CC_STATUS_MDEP): Remove.
1494 (CC_STATUS_MDEFP_INIT): Remove.
1495 (CC_IN_FPU): Remove.
1496 (NOTICE_UPDATE_CC): Remove.
1497 (REGISTER_NAMES): Add CC registers.
1498 (BRANCH_COST): Change to constant 1.
1499 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1501 * config/pdp11/pdp11.opt (mbcopy): Remove.
1502 (mbcopy-builtin): Remove.
1503 (mbranch-cheap): Remove.
1504 (mbranch-expensive): Remove.
1505 * config/pdp11/predicates.md (expand_shift_operand): Update to
1506 match shift code generation.
1507 (ccnz_operator): New predicate.
1508 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1509 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1510 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1512 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1513 description of O constraint.
1515 2018-06-27 Jeff Law <law@redhat.com>
1516 Austin Law <austinklaw@gmail.com>
1518 * config/v850/v850.md (addsi3_set_flags): New pattern.
1519 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1520 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1521 (zero_extendhisi2_v850_set_flags): Likewise.
1522 (zero_extendqisi2_v850_set_flags): Likewise.
1523 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1524 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1525 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1527 * config/v850/v850-protos.h (notice_update_cc): Remove.
1528 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1529 (v850_print_operand): Handle 'D' and "d".
1530 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1531 Add handling of arithmetic/logical operations compared against zero.
1532 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1533 Do not look at v850_compare_op, instead get mode from last argument.
1534 (v850_gen_compare): Remove
1535 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1536 after reload for prologue insns.
1537 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1539 (construct_save_jarl): Likewise.
1540 (TARGET_FLAGS_REGNUM): Define.
1541 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1542 (NOTICE_UPDATE_CC): Remove.
1543 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1544 than cc0. Conditionalize on reload_completed.
1545 (cmpsi_insn, setfcc_insn): Likewise.
1546 (tst1 splitter): Turn into define_and_split which sets the flags
1548 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1549 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1550 (cstoresf4, cstoredf4): Clobber the flags.
1551 (cmpsi, cmpsf, cmpdf): Remove expanders.
1552 (setf_insn): Remove pattern.
1553 (addsi3): Turn into define_and_split which clobbers the flags after
1554 reload and a suitable pattern (addsi3_clobber_flags) for use after
1556 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1557 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1558 (ashrsi3, ashrsi3_v850e2): Likewise.
1559 (bins): Clobber the flags.
1560 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1561 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1562 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1563 (call_value_internal_short, call_value_internal_long): Likewise.
1564 (callt_save_interrupt, callt_return_interrupt): Likewise.
1565 (save_interrupt, return_interrupt): Likewise.
1566 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1567 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1568 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1569 (All FP comparisons): Only allow after reload has completed.
1571 (divh, divhu): Tweak output template.
1572 (branch_z_normal, branch_z_invert): Remove
1573 (branch_nz_normal, branch_nz_invert): Likewise.
1574 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1576 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1577 * config/v850/v850.c (notice_update_cc): Remove.
1578 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1579 (CC_NO_CARRY): Likewise.
1580 (NOTICE_UPDATE_CC): Define to nothing.
1581 * config/v850/v850.md: Remove block comment on cc0 handling
1582 Remove "cc" attribute from all patterns. Remove cc_status handling
1583 from all patterns. Minor formatting fixes.
1585 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1587 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1588 (cortex-a76.cortex-a55): Likewise.
1589 * config/aarch64/aarch64-tune.md: Regenerate.
1590 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1591 cortex-a76.cortex-a55.
1593 2018-06-27 Jeff Law <law@redhat.com>
1595 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1596 (MULTILIB_DIRNAMES): Similarly.
1598 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1600 * gimple.h (gimple_return_retbnd): Delete.
1601 (gimple_return_set_retbnd): Likewise.
1602 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1603 gimple_return_set_retbnd.
1604 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1605 gimple_return_retbnd and adjust.
1606 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1607 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1608 Explicitly return NULL in a couple more cases. Move assertion
1609 on debug statements and remove unreachable code.
1610 (reset_debug_binding): Do not test id->retbnd.
1611 (expand_call_inline): Do not set it.
1613 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1615 * configure.ac: Add --disable-gcov option.
1616 * configure: Regenerate.
1617 * Makefile.in: Honour @enable_gcov@.
1618 * doc/install.texi: Document --disable-gcov.
1620 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1622 * config/arm/arm-cpus.in (cortex-a76): New entry.
1623 (cortex-a76.cortex-a55): Likewise.
1624 * config/arm/arm-tables.opt: Regenerate.
1625 * config/arm/arm-tune.md: Likewise.
1626 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1627 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1628 cortex-a76.cortex-a55.
1630 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1633 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1635 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1637 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1639 (EPILOGUE_USES): Likewise.
1641 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1643 * tree-inline.c (remap_location): New function extracted from...
1644 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1645 (copy_phis_for_bb): ...here. Call remap_location.
1646 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1648 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1650 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1651 unaligned vsx for 16B memset.
1653 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1656 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1657 ieee128_float_type_node to long_double_type_node unless
1658 TARGET_LONG_DOUBLE_128 is set.
1660 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1662 * cfgloop.c (get_loop_location): Convert return type from
1663 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1664 by implicit construction from rtx_insn *, and using
1665 dump_user_location_t::from_function_decl for the fallback case.
1666 * cfgloop.h (get_loop_location): Convert return type from
1667 location_t to dump_user_location_t.
1668 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1669 dump_printf_loc to pass in a dump_location_t rather than a
1670 location_t, via the gimple stmt.
1671 * coverage.c (get_coverage_counts): Update calls to
1672 dump_printf_loc to pass in dump_location_t rather than a
1674 * doc/optinfo.texi (Dump types): Convert example of
1675 dump_printf_loc from taking "locus" to taking "insn". Update
1676 description of the "_loc" calls to cover dump_location_t.
1677 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1679 (dump_user_location_t::dump_user_location_t): New constructors,
1680 from gimple *stmt and rtx_insn *.
1681 (dump_user_location_t::from_function_decl): New function.
1682 (dump_loc): Make static.
1683 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1684 const dump_location_t &.
1685 (dump_generic_expr_loc): Delete.
1686 (dump_printf_loc): Convert param "loc" from location_t to
1687 const dump_location_t &.
1688 (selftest::test_impl_location): New function.
1689 (selftest::dumpfile_c_tests): New function.
1690 * dumpfile.h: Include "profile-count.h".
1691 (class dump_user_location_t): New class.
1692 (struct dump_impl_location_t): New struct.
1693 (class dump_location_t): New class.
1694 (dump_printf_loc): Convert 2nd param from source_location to
1695 const dump_location_t &.
1696 (dump_generic_expr_loc): Delete.
1697 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1698 const dump_location_t &.
1699 * gimple-fold.c (fold_gimple_assign): Update call to
1700 dump_printf_loc to pass in a dump_location_t rather than a
1701 location_t, via the gimple stmt.
1702 (gimple_fold_call): Likewise.
1703 * gimple-loop-interchange.cc
1704 (loop_cand::analyze_iloop_reduction_var): Update for change to
1705 check_reduction_path.
1706 (tree_loop_interchange::interchange): Update for change to
1708 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1709 change in return-type of find_loop_location.
1710 (graphite_regenerate_ast_isl): Likewise.
1711 * graphite-optimize-isl.c (optimize_isl): Likewise.
1712 * graphite.c (graphite_transform_loops): Likewise.
1713 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1714 pass in a dump_location_t rather than a location_t, via the
1716 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1717 * ipa.c (walk_polymorphic_call_targets): Likewise.
1718 * loop-unroll.c (report_unroll): Convert "locus" param from
1719 location_t to dump_location_t.
1720 (decide_unrolling): Update for change to get_loop_location's
1722 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1723 location_t to dump_user_location_t.
1724 (grid_find_single_omp_among_assignments_1): Updates calls to
1725 dump_printf_loc to pass in a dump_location_t rather than a
1726 location_t, via the gimple stmt.
1727 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1728 location_t to dump_location_t. Updates calls to dump_printf_loc
1729 to pass in a dump_location_t rather than a location_t, via the
1731 (grid_inner_loop_gridifiable_p): Likewise.
1732 (grid_dist_follows_simple_pattern): Likewise.
1733 (grid_gfor_follows_tiling_pattern): Likewise.
1734 (grid_target_follows_gridifiable_pattern): Likewise.
1735 (grid_attempt_target_gridification): Convert initialization
1736 of local "grid" from memset to zero-initialization; FIXME: does
1737 this require C++11? Update call to dump_printf_loc to pass in a
1738 optinfo_location rather than a location_t, via the gimple stmt.
1739 * profile.c (read_profile_edge_counts): Updates call to
1740 dump_printf_loc to pass in a dump_location_t rather than a
1742 (compute_branch_probabilities): Likewise.
1743 * selftest-run-tests.c (selftest::run_tests): Call
1745 * selftest.h (dumpfile_c_tests): New decl.
1746 * tree-loop-distribution.c (pass_loop_distribution::execute):
1747 Update for change in return type of find_loop_location.
1748 * tree-parloops.c (parallelize_loops): Likewise.
1749 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1750 "locus" from location_t to dump_user_location_t.
1751 (canonicalize_loop_induction_variables): Likewise.
1752 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1753 for change in return type of find_loop_location.
1754 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1755 to dump_printf_loc to pass in a dump_location_t rather than a
1756 location_t, via the stmt.
1757 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1759 * tree-vect-loop-manip.c (find_loop_location): Convert return
1760 type from source_location to dump_user_location_t.
1761 (vect_do_peeling): Update for above change.
1762 (vect_loop_versioning): Update for change in type of
1764 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1765 from location_t to dump_user_location_t.
1766 (vect_estimate_min_profitable_iters): Update for change in type
1768 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1769 location_t to dump_location_t.
1770 (vect_slp_bb): Update for change in type of vect_location.
1771 * tree-vectorizer.c (vect_location): Convert from source_location
1772 to dump_user_location_t.
1773 (try_vectorize_loop_1): Update for change in vect_location's type.
1774 (vectorize_loops): Likewise.
1775 (increase_alignment): Likewise.
1776 * tree-vectorizer.h (vect_location): Convert from source_location
1777 to dump_user_location_t.
1778 (find_loop_location): Convert return type from source_location to
1779 dump_user_location_t.
1780 (check_reduction_path): Convert 1st param from location_t to
1781 dump_user_location_t.
1782 * value-prof.c (check_counter): Update call to dump_printf_loc to
1783 pass in a dump_user_location_t rather than a location_t; update
1784 call to error_at for change in type of "locus".
1785 (check_ic_target): Update call to dump_printf_loc to
1786 pass in a dump_user_location_t rather than a location_t, via the
1789 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1791 * config/s390/s390.h (enum processor_flags): Do not use
1792 default tune parameter when -march was specified.
1794 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1797 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1798 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1801 2018-06-26 Richard Biener <rguenther@suse.de>
1803 PR tree-optimization/86287
1805 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1806 (vect_analyze_loop): Initialize n_stmts.
1808 2018-06-26 Richard Biener <rguenther@suse.de>
1811 * fold-const.c (fold_convertible_p): Pointer extension
1814 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1817 * dwarf2out.c (loc_list_has_views): Adjust comments.
1818 (dw_loc_list): Split single cross-partition range with
1821 2018-06-25 Jeff Law <law@redhat.com>
1823 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1824 on -mbig-switch by default.
1826 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1828 (const_float_0_operand): Remove unused predicate.
1829 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1830 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1831 (recipsf2): New expander. Original pattern now called
1833 (recipdf2, recipdf2_insn): Similarly.
1834 (rsqrtsf2, rsqrtsf2_insn): Similarly
1835 (rsqrtdf2, rsqrtdf2_insn): Similarly
1837 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1839 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1840 Simplify logic for FreeBSD (twice).
1842 2018-06-25 Martin Sebor <msebor@redhat.com>
1844 PR tree-optimization/86204
1845 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1846 a strnlen result if it's less than the length of the string.
1848 2018-06-25 Martin Sebor <msebor@redhat.com>
1850 PR tree-optimization/85700
1851 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1852 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1853 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1855 2018-06-25 Martin Sebor <msebor@redhat.com>
1857 * doc/extend.texi (Zero-length arrays): Update and clarify.
1859 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1861 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1862 added IEEE/IBM long double multilib support on PowerPC little
1863 endian Linux systems.
1864 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1865 (MULTILIB_DEFAULTS): Likewise.
1866 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1868 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1869 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1870 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1872 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1875 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1876 (REORDER_45): Likewise.
1878 2018-06-25 Jeff Law <law@redhat.com>
1880 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1881 dividend to 32 bits. Adjust length.
1882 (udivmodhi4): Cleanup output template. Fix length.
1884 2018-06-25 Carl Love <cel@us.ibm.com>
1886 * config/rs6000/vsx.md: Change word selector to prefered location.
1888 2018-06-25 Richard Biener <rguenther@suse.de>
1890 PR tree-optimization/86304
1891 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1892 epilogue-if-converted loops as well.
1894 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1896 * lto-section-out.c (lto_begin_section): Do not print section
1897 name for noaddr and unnumbered dumps.
1899 2018-06-25 Richard Biener <rguenther@suse.de>
1901 * tree-vectorizer.h (struct vec_info_shared): New structure
1902 with parts split out from struct vec_info and loop_nest from
1903 struct _loop_vec_info.
1904 (struct vec_info): Adjust accordingly.
1905 (struct _loop_vec_info): Likewise.
1906 (LOOP_VINFO_LOOP_NEST): Adjust.
1907 (LOOP_VINFO_DATAREFS): Likewise.
1908 (LOOP_VINFO_DDRS): Likewise.
1909 (struct _bb_vec_info): Likewise.
1910 (BB_VINFO_DATAREFS): Likewise.
1911 (BB_VINFO_DDRS): Likewise.
1912 (struct _stmt_vec_info): Add dr_aux member.
1913 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1914 (DR_MISALIGNMENT_UNINITIALIZED): New.
1915 (set_dr_misalignment): Adjust.
1916 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1917 (vect_analyze_loop): Adjust prototype.
1918 (vect_analyze_loop_form): Likewise.
1919 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1920 Compute dependences lazily.
1921 (vect_record_base_alignments): Use shared datarefs/ddrs.
1922 (vect_verify_datarefs_alignment): Likewise.
1923 (vect_analyze_data_refs_alignment): Likewise.
1924 (vect_analyze_data_ref_accesses): Likewise.
1925 (vect_analyze_data_refs): Likewise.
1926 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1927 constructor parameter for shared part.
1928 (vect_analyze_loop_form): Pass in shared part and adjust.
1929 (vect_analyze_loop_2): Pass in storage for the number of
1930 stmts. Move loop nest finding to the caller. Compute
1932 (vect_analyze_loop): Pass in shared part.
1933 (vect_transform_loop): Verify shared datarefs are unchanged.
1934 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1935 constructor parameter for shared part.
1936 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1937 (vect_slp_bb): Verify shared datarefs are unchanged before
1939 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1941 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1942 DR_MISALIGNMENT_UNINITIALIZED.
1943 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1944 parameter for shared part.
1945 (vec_info::~vec_info): Adjust.
1946 (vec_info_shared::vec_info_shared): New.
1947 (vec_info_shared::~vec_info_shared): Likewise.
1948 (vec_info_shared::save_datarefs): Likewise.
1949 (vec_info_shared::check_datarefs): Likewise.
1950 (try_vectorize_loop_1): Construct shared part live for analyses
1951 of a single loop for multiple vector sizes.
1952 * tree-parloops.c (gather_scalar_reductions): Adjust.
1954 2018-06-25 Richard Biener <rguenther@suse.de>
1956 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1957 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1958 (vect_analyze_data_refs): Remove similar code from here and
1959 simplify accordingly.
1961 2018-06-25 Richard Biener <rguenther@suse.de>
1963 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1964 for reverse storage order accesses rather than asserting
1965 they cannot happen here.
1967 2018-06-25 Tom de Vries <tdevries@suse.de>
1970 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1971 Use data16 instead of .byte for insn prefix.
1973 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1976 * parser.c (make_char_string_pack): Pass this literal chars
1977 through cpp_interpret_string.
1978 (cp_parser_userdef_numeric_literal): Check the result of
1979 make_char_string_pack.
1981 2018-06-24 Maya Rashish <coypu@sdf.org>
1983 * ginclude/stddef.h: Simplify conditions around avoiding
1984 re-definition of __size_t.
1986 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1988 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1989 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1991 2018-06-22 Maya Rashish <coypu@sdf.org>
1993 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1994 made to OpenBSD and NetBSD through the years.
1996 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1998 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1999 behavior of vec_pack (vector double, vector double) to match
2000 behavior of vec_float2 (vector double, vector double).
2002 2018-06-22 Olivier Hainque <hainque@adacore.com>
2004 * gimplify.c (gimplify_function_tree): Prevent creation
2005 of a trampoline for the address of the current function
2006 passed to entry/exit instrumentation hooks.
2008 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2011 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2014 2018-06-22 Martin Liska <mliska@suse.cz>
2016 PR tree-optimization/86263
2017 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2018 Bail out if is_enabled is false.
2019 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2021 (jump_table_cluster::is_enabled): New function.
2023 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2025 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2026 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2027 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2028 (lto_input_ts_binfo_tree_pointers): Likewise.
2029 * tree-streamer-out.c (streamer_write_tree_bitfields,
2030 write_ts_binfo_tree_pointers): Likewise.
2031 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2033 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2035 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2037 2018-06-22 Martin Liska <mliska@suse.cz>
2039 * symbol-summary.h (get): Make it pure and inline move
2040 functionality from ::get function.
2041 (get): Remove and inline into ::get and ::get_create.
2042 (get_create): Move code from ::get function.
2044 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2047 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2048 -x assembler-with-cpp.
2050 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2052 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2053 _FILE_OFFSET_BITS=64 for C++.
2055 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2057 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2058 conversion insn that shows up when pr85657-3.c is compiled using
2059 IEEE 128-bit long double.
2060 (neg<mode>2_internal): Use the correct mode to check whether the
2061 mode is IBM extended.
2062 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2063 multiply and divide external functions from being created more
2066 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2068 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2070 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2071 the edge can be forwarded.
2072 (cfg_layout_merge_blocks): Likewise.
2074 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2076 * except.c (finish_eh_generation): Commit edge insertions only after
2077 the EH edges have been redirected from post-landing to landing pads.
2079 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2081 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2082 create_tmp_var_for to create the FRAME decl.
2083 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2085 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2087 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2088 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2089 debug statement when resetting its value.
2090 (expand_call_inline): Copy the locus of the call onto the assignment
2091 of the return value, if any. Use local variable in more cases.
2093 2018-06-21 Martin Liska <mliska@suse.cz>
2095 * ipa-pure-const.c (propagate_nothrow): Use
2096 funct_state_summaries->get.
2097 (dump_malloc_lattice): Likewise.
2098 (propagate_malloc): Likewise.
2100 2018-06-21 Richard Biener <rguenther@suse.de>
2102 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2103 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2104 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2106 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2107 BLOCK_ABSTRACT_ORIGIN unconditionally.
2109 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2111 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2113 * ipa-reference.c (ipa_reference_c_finalize): Delete
2114 ipa_ref_opt_sum_summaries and set it to NULL.
2116 2018-06-21 Tom de Vries <tdevries@suse.de>
2118 PR tree-optimization/85859
2119 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2120 test with comment from bb_no_side_effects_p.
2122 2018-06-21 Richard Biener <rguenther@suse.de>
2124 PR tree-optimization/86232
2125 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2126 max for constant niter.
2128 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2130 * config/aarch64/aarch64-simd.md
2131 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2133 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2135 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2136 Make opernads of the unspec commutative.
2138 2018-06-21 Richard Biener <rguenther@suse.de>
2140 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2141 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2142 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2143 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2145 (vect_record_base_alignments): Do not record base alignment
2146 for gathers or scatters.
2147 (vect_compute_data_ref_alignment): Drop return value that is always
2148 true. Bail out early for gathers or scatters.
2149 (vect_enhance_data_refs_alignment): Bail out early for gathers
2151 (vect_find_same_alignment_drs): Likewise.
2152 (vect_analyze_data_refs_alignment): Remove dead code.
2153 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2154 (vect_analyze_data_refs): For possible gathers or scatters do
2155 not create an alternate DR, just check their possible validity
2156 and mark them. Adjust DECL_NONALIASED handling to not rely
2158 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2159 update inits of gathers or scatters.
2160 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2161 Also copy gather/scatter flag to pattern vinfo.
2163 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2165 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2166 behavior of vec_packsu (vector unsigned long long, vector unsigned
2167 long long) to match behavior of vec_packs with same signature.
2169 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2170 Thomas Schwinge <thomas@codesourcery.com>
2171 Cesar Philippidis <cesar@codesourcery.com>
2173 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2174 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2175 (gimplify_adjust_omp_clauses): Likewise.
2176 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2177 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2178 (gimplify_omp_target_update): Update handling of acc update and
2180 * omp-low.c (install_var_field): Remove unused parameter
2181 base_pointers_restrict.
2182 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2183 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2185 (omp_target_base_pointers_restrict_p): Delete.
2186 (scan_omp_target): Update call to scan_sharing_clauses.
2187 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2189 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2190 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2191 (convert_local_omp_clauses): Likewise.
2192 * tree-pretty-print.c (dump_omp_clause): Likewise.
2193 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2195 (omp_clause_code_name): Likewise.
2197 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2200 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2203 PR tree-optimization/86231
2204 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2205 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2207 2018-06-20 Tom de Vries <tdevries@suse.de>
2209 PR tree-optimization/86097
2210 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2211 iv type if signedness of iv type is not the same as that of *nit.
2213 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2215 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2216 EDGE_EH edges, verify they are all EDGE_EH.
2218 2018-06-20 Maya Rashish <coypu@sdf.org>
2220 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2222 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2224 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2225 * config/aarch64/aarch64.c (xgene1_tunings): Add
2226 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2227 (aarch64_mode_valid_for_sched_fusion_p):
2228 Allow 16-byte modes.
2229 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2230 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2232 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2234 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2235 * config/aarch64/iterators.md (VQ2): New mode iterator.
2237 2018-06-20 Martin Liska <mliska@suse.cz>
2239 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2240 Change default ratio from 10 to 8.
2242 2018-06-20 Martin Liska <mliska@suse.cz>
2244 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2246 (bit_test_cluster::find_bit_tests): Likewise.
2247 (switch_decision_tree::analyze_switch_statement): Find clusters.
2248 * tree-switch-conversion.h (struct jump_table_cluster): Document
2251 2018-06-20 Martin Liska <mliska@suse.cz>
2253 * tree-switch-conversion.c (switch_conversion::collect):
2254 Record m_uniq property.
2255 (switch_conversion::expand): Bail out for special conditions.
2256 (group_cluster::~group_cluster): New.
2257 (group_cluster::group_cluster): Likewise.
2258 (group_cluster::dump): Likewise.
2259 (jump_table_cluster::emit): New.
2260 (switch_decision_tree::fix_phi_operands_for_edges): New.
2261 (struct case_node): Remove struct.
2262 (jump_table_cluster::can_be_handled): New.
2263 (case_values_threshold): Moved to header.
2264 (reset_out_edges_aux): Likewise.
2265 (jump_table_cluster::is_beneficial): New.
2266 (bit_test_cluster::can_be_handled): Likewise.
2267 (add_case_node): Remove.
2268 (bit_test_cluster::is_beneficial): New.
2269 (case_bit_test::cmp): New.
2270 (bit_test_cluster::emit): New.
2271 (expand_switch_as_decision_tree_p): Remove.
2272 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2273 (fix_phi_operands_for_edge): Likewise.
2274 (switch_decision_tree::analyze_switch_statement): New.
2275 (compute_cases_per_edge): Move ...
2276 (switch_decision_tree::compute_cases_per_edge): ... here.
2277 (try_switch_expansion): Likewise.
2278 (switch_decision_tree::try_switch_expansion): Likewise.
2279 (record_phi_operand_mapping): Likewise.
2280 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2281 (emit_case_decision_tree): Likewise.
2282 (switch_decision_tree::emit): Likewise.
2283 (balance_case_nodes): Likewise.
2284 (switch_decision_tree::balance_case_nodes): Likewise.
2285 (dump_case_nodes): Likewise.
2286 (switch_decision_tree::dump_case_nodes): Likewise.
2287 (emit_jump): Likewise.
2288 (switch_decision_tree::emit_jump): Likewise.
2289 (emit_cmp_and_jump_insns): Likewise.
2290 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2291 (emit_case_nodes): Likewise.
2292 (switch_decision_tree::emit_case_nodes): Likewise.
2293 (conditional_probability): Remove.
2294 * tree-switch-conversion.h (enum cluster_type): New.
2296 (struct cluster): Likewise.
2297 (cluster::cluster): Likewise.
2298 (struct simple_cluster): Likewise.
2299 (simple_cluster::simple_cluster): Likewise.
2300 (struct group_cluster): Likewise.
2301 (struct jump_table_cluster): Likewise.
2302 (struct bit_test_cluster): Likewise.
2303 (struct min_cluster_item): Likewise.
2304 (struct case_tree_node): Likewise.
2305 (case_tree_node::case_tree_node): Likewise.
2306 (jump_table_cluster::case_values_threshold): Likewise.
2307 (struct case_bit_test): Likewise.
2308 (struct switch_decision_tree): Likewise.
2309 (struct switch_conversion): Likewise.
2310 (switch_decision_tree::reset_out_edges_aux): Likewise.
2312 2018-06-20 Martin Liska <mliska@suse.cz>
2314 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2315 (hoist_edge_and_branch_if_true): Likewise.
2316 (expand_switch_using_bit_tests_p): Likewise.
2317 (struct case_bit_test): Likewise.
2318 (case_bit_test_cmp): Likewise.
2319 (emit_case_bit_tests): Likewise.
2320 (switch_conversion::switch_conversion): New class.
2321 (struct switch_conv_info): Remove old struct.
2322 (collect_switch_conv_info): More to ...
2323 (switch_conversion::collect): ... this.
2324 (check_range): Likewise.
2325 (switch_conversion::check_range): Likewise.
2326 (check_all_empty_except_final): Likewise.
2327 (switch_conversion::check_all_empty_except_final): Likewise.
2328 (check_final_bb): Likewise.
2329 (switch_conversion::check_final_bb): Likewise.
2330 (create_temp_arrays): Likewise.
2331 (switch_conversion::create_temp_arrays): Likewise.
2332 (free_temp_arrays): Likewise.
2333 (gather_default_values): Likewise.
2334 (switch_conversion::gather_default_values): Likewise.
2335 (build_constructors): Likewise.
2336 (switch_conversion::build_constructors): Likewise.
2337 (constructor_contains_same_values_p): Likewise.
2338 (switch_conversion::contains_same_values_p): Likewise.
2339 (array_value_type): Likewise.
2340 (switch_conversion::array_value_type): Likewise.
2341 (build_one_array): Likewise.
2342 (switch_conversion::build_one_array): Likewise.
2343 (build_arrays): Likewise.
2344 (switch_conversion::build_arrays): Likewise.
2345 (gen_def_assigns): Likewise.
2346 (switch_conversion::gen_def_assigns): Likewise.
2347 (prune_bbs): Likewise.
2348 (switch_conversion::prune_bbs): Likewise.
2349 (fix_phi_nodes): Likewise.
2350 (switch_conversion::fix_phi_nodes): Likewise.
2351 (gen_inbound_check): Likewise.
2352 (switch_conversion::gen_inbound_check): Likewise.
2353 (process_switch): Use the newly created class.
2354 (switch_conversion::expand): New.
2355 (switch_conversion::~switch_conversion): New.
2356 * tree-switch-conversion.h: New file.
2358 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2360 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2361 tree-vect-patterns.c.
2362 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2363 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2364 (vect_recog_sad_pattern): Likewise.
2365 (vect_recog_widen_sum_pattern): Likewise.
2366 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2367 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2368 (vect_recog_rotate_pattern): Likewise.
2369 (vect_recog_mult_pattern): Likewise.
2370 (vect_recog_vector_vector_shift_pattern): Likewise.
2371 (vect_recog_divmod_pattern): Likewise.
2372 (vect_recog_mixed_size_cond_pattern): Likewise.
2373 (vect_recog_bool_pattern): Likewise.
2374 (vect_recog_mask_conversion_pattern): Likewise.
2375 (vect_try_gather_scatter_pattern): Likewise.
2376 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2377 (vect_recog_over_widening_pattern): Likewise.
2378 (vect_recog_gather_scatter_pattern): Likewise.
2379 (vect_recog_func_ptr): Move from tree-vectorizer.h
2380 (vect_vect_recog_func_ptrs): Move further down the file.
2381 (vect_recog_func): Likewise. Remove the third argument.
2382 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2383 (vect_pattern_recog_1): Expect the pattern function to do any
2384 necessary target tests. Also expect it to provide a vector type.
2385 Remove the type_in handling.
2387 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2389 * tree-vect-patterns.c (vect_pattern_detected): New function.
2390 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2391 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2392 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2393 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2394 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2395 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2396 (vect_recog_mask_conversion_pattern)
2397 (vect_try_gather_scatter_pattern): Likewise.
2399 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2401 * tree-vect-patterns.c (vect_get_internal_def): New function.
2402 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2403 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2404 (search_type_for_mask_1): Use it.
2406 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2408 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2409 redundant WIDEN_SUM_EXPR handling.
2410 (vect_recog_sad_pattern): Likewise.
2412 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2414 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2415 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2416 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2417 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2418 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2420 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2422 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2423 use the stmt_vec_info of the original bb statement for the
2424 new zero assignment, even if the call is part of a pattern.
2426 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2428 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2429 that the sequence is attached to the original statement rather
2430 than the pattern statement.
2431 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2432 PATTERN_DEF_SEQ from the original statement rather than
2433 the main pattern statement.
2434 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2435 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2436 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2438 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2440 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2441 definition statements before the early exit for statements that aren't
2443 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2445 (vect_transform_loop): ...here. Process pattern definition
2446 statements without first checking whether the main pattern
2447 statement is live or relevant.
2449 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2451 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2452 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2454 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2456 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2457 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2458 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2459 (expand_block_compare): Change select_block_compare_mode call.
2460 (expand_strncmp_align_check): Use new functions, fix comment.
2461 (emit_final_str_compare_gpr): New function.
2462 (expand_strn_compare): Refactor and clean up code.
2463 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2465 2018-06-19 Tony Reix <tony.reix@atos.com>
2466 Damien Bergamini <damien.bergamini@atos.com>
2467 David Edelsohn <dje.gcc@gmail.com>
2469 * collect2.c (static_obj): New variable.
2470 (static_libs): New variable.
2471 (is_in_list): Uncomment declaration.
2472 (main): Track AIX libraries linked statically.
2473 (is_in_list): Uncomment definition.
2474 (scan_prog_file): Don't add AIX shared libraries initializer
2475 to constructor list if linking statically.
2477 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2479 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2481 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2483 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2485 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2488 2018-06-19 Martin Liska <mliska@suse.cz>
2490 * config/i386/i386.c (ix86_can_inline_p): Do not use
2491 ipa_fn_summaries::get_create.
2492 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2494 (devirtualization_time_bonus): Likewise.
2495 (ipcp_propagate_stage): Likewise.
2496 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2497 (edge_set_predicate): Likewise.
2498 (evaluate_conditions_for_known_args): Likewise.
2499 (evaluate_properties_for_edge): Likewise.
2500 (ipa_call_summary::reset): Tranform to ...
2501 (ipa_call_summary::~ipa_call_summary): ... this.
2502 (ipa_fn_summary::reset): Transform to ...
2503 (ipa_fn_summary::~ipa_fn_summary): ... this.
2504 (ipa_fn_summary_t::remove): Rename to ...
2505 (ipa_fn_summary_t::remove_callees): ... this.
2506 (ipa_fn_summary_t::duplicate): Use placement new
2507 instead of memory copy.
2508 (ipa_call_summary_t::duplicate): Likewise.
2509 (ipa_call_summary_t::remove): Remove.
2510 (dump_ipa_call_summary): Change get_create to get.
2511 (ipa_dump_fn_summary): Dump only when summary exists.
2512 (analyze_function_body): Use symbol_summary::get instead
2514 (compute_fn_summary): Likewise.
2515 (estimate_edge_devirt_benefit): Likewise.
2516 (estimate_edge_size_and_time): Likewise.
2517 (inline_update_callee_summaries): Likewise.
2518 (remap_edge_change_prob): Likewise.
2519 (remap_edge_summaries): Likewise.
2520 (ipa_merge_fn_summary_after_inlining): Likewise.
2521 (write_ipa_call_summary): Likewise.
2522 (ipa_fn_summary_write): Likewise.
2523 (ipa_free_fn_summary): Likewise.
2524 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2525 (struct ipa_call_summary): Likewise.
2526 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2528 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2529 (estimate_size_after_inlining): Likewise.
2530 (estimate_growth): Likewise.
2531 (growth_likely_positive): Likewise.
2532 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2533 (inline_call): Likewise.
2534 * ipa-inline.c (caller_growth_limits): Likewise.
2535 (can_inline_edge_p): Likewise.
2536 (can_inline_edge_by_limits_p): Likewise.
2537 (compute_uninlined_call_time): Likewise.
2538 (compute_inlined_call_time): Likewise.
2539 (want_inline_small_function_p): Likewise.
2540 (edge_badness): Likewise.
2541 (update_caller_keys): Likewise.
2542 (update_callee_keys): Likewise.
2543 (inline_small_functions): Likewise.
2544 (inline_to_all_callers_1): Likewise.
2545 (dump_overall_stats): Likewise.
2546 (early_inline_small_functions): Likewise.
2547 (early_inliner): Likewise.
2548 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2549 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2550 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2551 * ipa-split.c (execute_split_functions): Likewise.
2552 * symbol-summary.h: Likewise.
2553 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2555 2018-06-19 Richard Biener <rguenther@suse.de>
2557 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2558 (vectorize_loops): ... here. Fix dbgcnt handling.
2559 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2561 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2564 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2565 ieee128 argument takes up only one (vector) register, not two (floating
2568 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2570 * gimplify.c (gimplify_init_constructor): Really never clear for an
2571 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2573 2018-06-19 Richard Biener <rguenther@suse.de>
2575 PR tree-optimization/86179
2576 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2577 after failed recognition.
2579 2018-06-18 Martin Sebor <msebor@redhat.com>
2582 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2583 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2584 Handle integer subtraction.
2585 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2586 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2588 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2590 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2591 param from rtx to rtx_insn *.
2592 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2594 (frv_ifcvt_modify_insn): Likwise.
2595 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2596 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2597 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2598 as_a <rtx_insn *> cast to local "unprotected_region" once
2599 it's been established that it's not NULL or pc_rtx.
2600 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2601 param "sethi" from rtx to rtx_insn *.
2602 (nds32_group_float_insns): Likewise for param "insn".
2603 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2605 (vax_output_int_subtract): Likewise.
2606 * config/vax/vax.c (vax_output_int_add): Likewise for param
2608 (vax_output_int_subtract): Likewise.
2609 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2610 (emit_pattern_after): Likewise for param "after".
2611 (emit_insn_after): Likewise.
2612 (emit_jump_insn_after): Likewise.
2613 (emit_call_insn_after): Likewise.
2614 (emit_debug_insn_after): Likewise.
2615 (emit_pattern_before): Likewise for param "before".
2616 (emit_insn_before): Likewise.
2617 (emit_jump_insn_before): Likewise.
2618 * final.c (get_insn_template): Likewise for param "insn", removing
2620 * output.h (get_insn_template): Likewise for 2nd param.
2621 * rtl.h (emit_insn_before): Likewise.
2622 (emit_jump_insn_before): Likewise.
2623 (emit_debug_insn_before_noloc): Likewise.
2624 (emit_insn_after): Likewise.
2625 (emit_jump_insn_after): Likewise.
2626 (emit_call_insn_after): Likewise.
2627 (emit_debug_insn_after): Likewise.
2628 (set_insn_deleted): Likewise for param.
2630 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2633 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2634 floating point modes, so that IFmode is numerically greater than
2635 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2636 to declare the ordering. This prevents IFmode from being
2637 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2638 machine. Include rs6000-modes.h to share the fractional values
2639 between genmodes* and the rest of the compiler.
2643 * config/rs6000/rs6000-modes.h: New file.
2644 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2645 meaning of rs6000_long_double_size so that 126..128 selects an
2646 appropriate 128-bit floating point type.
2647 (rs6000_option_override_internal): Likewise.
2648 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2649 (TARGET_LONG_DOUBLE_128): Change the meaning of
2650 rs6000_long_double_size so that 126..128 selects an appropriate
2651 128-bit floating point type.
2652 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2653 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2654 source and destination to match the standard usage.
2655 (truncifkf2): Likewise.
2656 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2657 ISA 2.07 to use an explicit clobber, instead of passing in a
2659 (copysign<mode>3_soft): Likewise.
2661 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2663 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2664 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2665 (vect_slp_analyze_instance_dependence): Likewise.
2666 (vect_enhance_data_refs_alignment): Likewise.
2667 (vect_analyze_data_refs_alignment): Likewise.
2668 (vect_slp_analyze_and_verify_instance_alignment
2669 (vect_analyze_data_ref_accesses): Likewise.
2670 (vect_prune_runtime_alias_test_list): Likewise.
2671 (vect_analyze_data_refs): Likewise.
2672 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2673 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2674 (vect_analyze_scalar_cycles_1): Likewise.
2675 (vect_get_loop_niters): Likewise.
2676 (vect_analyze_loop_form_1): Likewise.
2677 (vect_update_vf_for_slp): Likewise.
2678 (vect_analyze_loop_operations): Likewise.
2679 (vect_analyze_loop): Likewise.
2680 (vectorizable_induction): Likewise.
2681 (vect_transform_loop): Likewise.
2682 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2683 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2684 (vect_make_slp_decision): Likewise.
2685 (vect_detect_hybrid_slp): Likewise.
2686 (vect_slp_analyze_operations): Likewise.
2687 (vect_slp_bb): Likewise.
2688 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2689 (vectorizable_bswap): Likewise.
2690 (vectorizable_call): Likewise.
2691 (vectorizable_simd_clone_call): Likewise.
2692 (vectorizable_conversion): Likewise.
2693 (vectorizable_assignment): Likewise.
2694 (vectorizable_shift): Likewise.
2695 (vectorizable_operation): Likewise.
2696 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2698 2018-06-18 Martin Sebor <msebor@redhat.com>
2700 PR tree-optimization/81384
2701 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2702 * builtins.c (expand_builtin_strnlen): New function.
2703 (expand_builtin): Call it.
2704 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2705 * builtins.def (BUILT_IN_STRNLEN): New.
2706 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2707 Warn for bounds in excess of maximum object size.
2708 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2709 single-value ranges. Handle strnlen.
2710 (handle_builtin_strlen): Handle strnlen.
2711 (strlen_check_and_optimize_stmt): Same.
2712 * doc/extend.texi (Other Builtins): Document strnlen.
2714 2018-06-18 Maya Rashish <coypu@sdf.org>
2716 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2717 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2718 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2720 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2722 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2724 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2726 * tree.c (escaped_string::escape): Replace cast to char * by
2727 const_cast<char *> (unescaped).
2729 2018-06-18 Nick Clifton <nickc@redhat.com>
2732 * tree.c (escaped_string): New class. Converts an unescaped
2733 string into its escaped equivalent.
2734 (warn_deprecated_use): Use the new class to convert the
2735 deprecation message, if present.
2736 (test_escaped_strings): New self test.
2737 (test_c_tests): Add test_escaped_strings.
2738 * doc/extend.texi (deprecated): Add a note that the
2739 deprecation message is affected by the -fmessage-length
2740 option, and that control characters will be escaped.
2741 (#pragma GCC error): Document this pragma.
2742 (#pragma GCC warning): Likewise.
2743 * doc/invoke.texi (-fmessage-length): Document this option's
2744 effect on the #warning and #error preprocessor directives and
2745 the deprecated attribute.
2747 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2749 * tree.c (decl_value_expr_lookup): Revert latest change.
2750 (decl_value_expr_insert): Likewise.
2752 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2754 * gimplify.c (nonlocal_vlas): Delete.
2755 (nonlocal_vla_vars): Likewise.
2756 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2758 (gimplify_body): Do not create and destroy nonlocal_vlas.
2759 * tree-nested.c: Include diagnostic.h.
2760 (use_pointer_in_frame): Tweak.
2761 (lookup_field_for_decl): Add assertion and declare the transformation.
2762 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2763 internal error when the reference is in a wrong context. Do not
2764 create a debug decl by default.
2765 (note_nonlocal_block_vlas): Delete.
2766 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2767 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2768 create a debug decl by default.
2769 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2770 call is in a wrong context.
2771 (fixup_vla_decls): New function.
2772 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2773 debug variables were created.
2774 * tree.c (decl_value_expr_lookup): Add checking assertion.
2775 (decl_value_expr_insert): Likewise.
2777 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2780 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2781 * tree-scalar-evolution.c (interpret_expr): Likewise.
2782 (expression_expensive_p): Likewise.
2783 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2784 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2785 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2786 (ssa_defined_by_minus_one_stmt_p): New.
2788 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2791 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2792 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2793 * dojump.c (do_jump): Likewise.
2794 * expr.c (expand_expr_real_2): Check operand type's sign.
2795 * fold-const.c (const_unop): Handle ABSU_EXPR.
2796 (fold_abs_const): Likewise.
2797 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2798 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2799 (strip_sign_op_1): Likesise.
2800 * match.pd: Add new pattern to generate ABSU_EXPR.
2801 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2802 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2803 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2804 * tree-inline.c (estimate_operator_cost): Likewise.
2805 * tree-pretty-print.c (dump_generic_node): Likewise.
2806 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2807 * tree.def (ABSU_EXPR): New.
2809 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2812 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2813 documented as preserved for backward compatibility only.
2814 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2816 PR rtl-optimization/86108
2817 * bb-reorder.c (create_forwarder_block): Renamed to ...
2818 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2819 jump from new landing pad to the second part.
2820 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2823 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2826 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2827 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2828 Only call store_expr for halves if the mode is the same.
2831 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2832 Fix up comment formatting.
2834 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2836 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2837 (typed_splay_tree::closure,
2838 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2839 (typed_splay_tree::typed_splay_tree,
2840 typed_splay_tree::operator =): Declared private.
2841 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2842 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2843 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2844 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2845 typed_splay_tree::splay_tree_splay,
2846 typed_splay_tree::splay_tree_foreach_helper,
2847 typed_splay_tree::splay_tree_insert,
2848 typed_splay_tree::splay_tree_remove,
2849 typed_splay_tree::splay_tree_lookup,
2850 typed_splay_tree::splay_tree_predecessor,
2851 typed_splay_tree::splay_tree_successor,
2852 typed_splay_tree::splay_tree_min,
2853 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2854 (typed_splay_tree::root, typed_splay_tree::comp,
2855 typed_splay_tree::delete_key,
2856 typed_splay_tree::delete_value): New data members.
2857 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2858 typed_splay_tree::remove.
2860 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2862 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2863 -mginv and -mno-ginv to the assembler.
2864 * config/mips/mips.opt (-mcrc): New option.
2866 * doc/invoke.text (-mcrc): Document.
2869 2018-06-15 Nick Clifton <nickc@redhat.com>
2872 * tree.c (escaped_string): New class. Converts an unescaped
2873 string into its escaped equivalent.
2874 (warn_deprecated_use): Use the new class to convert the
2875 deprecation message, if present.
2876 (test_escaped_strings): New self test.
2877 (test_c_tests): Add test_escaped_strings.
2878 * doc/extend.texi (deprecated): Add a note that the
2879 deprecation message is affected by the -fmessage-length
2880 option, and that control characters will be escaped.
2881 (#pragma GCC error): Document this pragma.
2882 (#pragma GCC warning): Likewise.
2883 * doc/invoke.texi (-fmessage-length): Document this option's
2884 effect on the #warning and #error preprocessor directives and
2885 the deprecated attribute.
2887 2018-06-15 Richard Biener <rguenther@suse.de>
2889 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2890 here, also noting vector size used.
2891 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2892 size used in MSG_OPTIMIZED_LOCATIONS dump.
2893 (pass_slp_vectorize::execute): Adjust.
2895 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2898 * config/arc/arc.c (arc_return_address_register): Fix
2901 2018-06-15 Richard Biener <rguenther@suse.de>
2904 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2905 leave useless conversion stripping to force_gimple_operand_gsi.
2906 (gimplify_build2): Likewise.
2907 (gimplify_build1): Likewise.
2909 2018-06-15 Richard Biener <rguenther@suse.de>
2912 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2913 before adjusting their block.
2915 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2917 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2918 multilibs for *-*-rtems*.
2919 * config/riscv/t-rtems: New file.
2921 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2924 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2925 unsigned_type_for returns NULL.
2928 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2929 subregs of multi-word pseudos unless the float mode has word size.
2931 2018-06-14 Richard Biener <rguenther@suse.de>
2934 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2935 duplicate and harmful type_hash_canon.
2936 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2938 2018-06-14 Richard Biener <rguenther@suse.de>
2941 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2944 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2946 * config/rtems.h (STDINT_LONG32): Define.
2948 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2949 Prachi Godbole <prachi.godbole@imgtec.com>
2951 * config/mips/mips-cpus.def: Define P6600.
2952 * config/mips/mips-tables.opt: Regenerate.
2953 * config/mips/mips.c (mips_ucbranch_type): New enum.
2954 (mips_rtx_cost_data): Add support for P6600.
2955 (mips_issue_rate): Likewise.
2956 (mips_multipass_dfa_lookahead): Likewise.
2957 (mips_avoid_hazard): Likewise.
2958 (mips_reorg_process_insns): Likewise.
2959 (mips_classify_branch_p6600): New function.
2960 * config/mips/mips.h (TUNE_P6600): New define.
2961 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2962 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2963 * config/mips/mips.md: Include p6600.md.
2964 (processor): Add p6600.
2965 * config/mips/p6600.md: New file.
2966 * doc/invoke.texi: Add p6600 to supported architectures.
2968 2018-06-13 Martin Sebor <msebor@redhat.com>
2970 PR tree-optimization/86114
2971 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2973 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2975 2018-06-13 Richard Biener <rguenther@suse.de>
2977 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2978 Properly set vector type of the intermediate stmt.
2979 * tree-vect-stmts.c (vectorizable_operation): The destination
2980 var always has vectype_out type.
2982 2018-06-13 Jeff Law <law@redhat.com>
2984 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2985 integer 0 for argument to print_rtl_with_bb.
2986 (rl78_reorg): Likewise.
2988 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2990 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2991 from rtx to rtx_insn *.
2992 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2994 (add_sched_insns_for_speculation): Likewise for local "target",
2995 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2996 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2997 from rtx_insn ** to rtx_code_label **.
2998 (reorg_emit_nops): Likewise.
2999 (c6x_reorg): Likewise for local "call_labels".
3000 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3002 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3003 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3004 the loops over LABEL_REFS.
3005 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3007 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3008 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3009 (split_branches): Strengthen local "olabel" from rtx to
3010 rtx_insn *, adding a safe_as_a cast.
3011 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3013 (add_insn_after): Likewise for first two params.
3014 (add_insn_before): Likewise.
3015 (remove_insn): Likewise for param.
3016 (emit_pattern_before_noloc): Likewise for second and third params.
3017 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3018 (emit_call_insn_before_noloc): Likewise.
3019 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3021 (emit_barrier_before): Likewise.
3022 (emit_label_before): Strengthen "label" param from "rtx" to
3023 "rtx_code_label *". Strengthen "before" param from "rtx" to
3025 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3027 (emit_pattern_after_noloc): Likewise.
3028 (emit_insn_after_noloc): Likewise.
3029 (emit_jump_insn_after_noloc): Likewise.
3030 (emit_call_insn_after_noloc): Likewise.
3031 (emit_debug_insn_after_noloc): Likewise.
3032 (emit_barrier_after): Likewise.
3033 (emit_label_after): Likewise for both params.
3034 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3035 "loc" param from "int" to "location_t".
3036 (emit_insn_after_setloc): Likewise.
3037 (emit_jump_insn_after_setloc): Likewise.
3038 (emit_call_insn_after_setloc): Likewise.
3039 (emit_debug_insn_after_setloc): Likewise.
3040 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3041 "loc" param from "int" to "location_t".
3042 (emit_pattern_before): Convert NULL_RTX to NULL.
3043 (emit_insn_before_setloc): Convert "loc" param from "int" to
3045 (emit_jump_insn_before_setloc): Likewise.
3046 (emit_call_insn_before_setloc): Likewise.
3047 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3048 rtx_insn *. Convert "loc" param from "int" to "location_t".
3049 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3050 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3051 Convert 3rd param from "int" to "location_t".
3052 (emit_barrier_before, emit_barrier_after, next_real_insn):
3053 Strengthen param from rtx to rtx_insn *.
3054 (emit_label_before): Strengthen 1st param from "rtx" to
3055 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3057 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3058 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3059 Strengthen 2nd param from "rtx" to "rtx_insn *".
3060 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3061 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3062 Likewise. Convert 3rd param from "int" to "location_t".
3063 (emit_label_after): Strengthen 1st param from "rtx" to
3065 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3067 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3068 from "rtx" to "rtx_insn *".
3070 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3072 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3073 bodies streamed in with -Q.
3074 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3075 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3076 * gimple-streamer-out.c: Include gimple-pretty-print.h
3077 (output_bb): Dump stmts streamed.
3078 * lto-section-out.c: Include print-tree.h
3079 (lto_begin_section): Dump sections created.
3080 (lto_output_decl_index): Dump decl encoded.
3081 * lto-streamer-out.c: Include print-tree.h
3082 (create_output_block): Dump output block created.
3083 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3084 (output_function): Dump function output.
3085 (output_constructor): Dump constructor streamed.
3086 (write_global_stream): Output indexes encoded.
3087 (produce_asm_for_decls): Dump streams encoded.
3088 * lto-streamer.c (streamer_dump_file): New global var.
3089 * lto-streamer.h (streamer_dump_file): Declare.
3090 * passes.c (ipa_write_summaries): Initialize streamer dump.
3091 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3094 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3097 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3098 offsets for register save directives. Emit a second batch of save
3099 directives, if need be, when the function accesses prior frames.
3101 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3103 * config/arc/fpu.md (fmasf4): Force operand to register.
3104 (fnmasf4): Likewise.
3106 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3108 * config/arc/arc-protos.h (arc_pad_return): Remove.
3109 * config/arc/arc.c (machine_function): Remove force_short_suffix
3111 (arc_print_operand): Adjust printing of '&'.
3112 (arc_verify_short): Remove conditional printing of short suffix.
3113 (arc_final_prescan_insn): Remove reference to size_reason.
3114 (pad_return): New function.
3115 (arc_reorg): Call pad_return.
3116 (arc_pad_return): Remove.
3117 (arc_init_machine_status): Remove reference to force_short_suffix.
3118 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3119 (attr length): When attribute iscompact is true force to 2
3120 regardless; in the case of maybe check if we want to force the
3121 instruction to have 4 bytes length.
3122 (nopv): Change it to generate 4 byte long nop as well.
3123 (blockage): New pattern.
3124 (simple_return): Remove call to arc_pad_return.
3125 (p_return_i): Likewise.
3127 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3129 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3131 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3133 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3136 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3138 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3139 for ARC700 and ARCv2.
3141 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3144 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3145 operands[2] instead of operands[1].
3148 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3150 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3151 case, check whether the outer register overlaps an unallocatable
3152 register, not just whether it fits the required class.
3154 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3156 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3158 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3159 (memrefs_conflict_p): Likewise.
3160 (init_alias_analysis): Likewise.
3161 * cfgexpand.c (expand_debug_expr): Likewise.
3162 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3163 * cse.c (fold_rtx): Likewise.
3164 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3165 * expr.c (emit_block_move_hints): Likewise.
3166 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3167 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3168 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3169 (emit_group_store): Likewise.
3170 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3171 to read the PRE/POST_MODIFY increment.
3172 * calls.c (store_one_arg): Use strip_offset.
3173 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3175 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3177 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3178 (simplify_binary_operation_1): Extend CONST_INT handling to
3180 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3181 than a HOST_WIDE_INT.
3182 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3184 (adjust_mems, add_stores): Update accodingly.
3185 (vt_canonicalize_addr): Track polynomial offsets.
3186 (emit_note_insn_var_location): Likewise.
3187 (vt_add_function_parameter): Likewise.
3188 (vt_initialize): Likewise.
3190 2018-06-12 Jeff Law <law@redhat.com>
3192 * config.gcc (alpha*-*-freebsd*): Remove.
3193 * config/alpha/freebsd.h: Remove.
3195 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3198 * spellcheck-tree.c (levenshtein_distance): Rename to...
3199 (get_edit_distance): ...this, and update for underlying renaming.
3200 * spellcheck-tree.h (levenshtein_distance): Rename to...
3201 (get_edit_distance): ...this.
3202 * spellcheck.c (levenshtein_distance): Rename to...
3203 (get_edit_distance): ...this. Convert from Levenshtein distance
3204 to Damerau-Levenshtein distance by supporting transpositions of
3205 adjacent characters. Rename "v1" to "v_next" and "v0" to
3207 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3208 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3209 update for underlying renaming.
3210 (selftest::levenshtein_distance_unit_test): Rename to...
3211 (selftest::test_get_edit_distance_unit): ...this, and update for
3212 underlying renaming.
3213 (selftest::test_find_closest_string): Add example from PR 69968
3214 where transposition helps
3215 (selftest::test_metric_conditions): Update for renaming.
3216 (selftest::test_metric_conditions): Likewise.
3217 (selftest::spellcheck_c_tests): Likewise.
3218 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3219 (get_edit_distance): ...this.
3220 (best_match::consider): Update for renaming.
3222 2018-06-12 Martin Sebor <msebor@redhat.com>
3224 PR tree-optimization/85259
3225 * builtins.c (compute_objsize): Handle constant offsets.
3226 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3227 true iff a warning has been issued.
3228 * gimple.h (gimple_nonartificial_location): New function.
3229 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3230 gimple_nonartificial_location and handle -Wno-system-headers.
3231 (handle_builtin_stxncpy): Same.
3233 2018-06-12 Martin Sebor <msebor@redhat.com>
3236 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3238 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3240 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3241 BUILTIN_VEC_XST entries for pointer to double and long long.
3243 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3246 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3248 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3251 2018-06-12 Martin Liska <mliska@suse.cz>
3253 * doc/options.texi: Document IntegerRange.
3255 2018-06-12 Martin Liska <mliska@suse.cz>
3257 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3258 * opt-functions.awk: Handle Deprecated flag.
3259 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3261 (read_cmdline_option): Report warning for a deprecated option.
3262 * opts.h (struct cl_option): Add new field cl_deprecated.
3263 (CL_ERR_DEPRECATED): New.
3265 2018-06-12 Martin Liska <mliska@suse.cz>
3267 * doc/options.texi: Document Deprecated option flag.
3269 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3271 * config/arc/arc-arch.h (arc_extras): New enum.
3272 (arc_cpu_t):Add field extra.
3273 (arc_cpu_types): Consider the extras.
3274 * config/arc/arc-cpus.def: Add extras info.
3275 * config/arc/arc-opts.h (processor_type): Consider extra field.
3276 * config/arc/arc.c (arc_override_options): Handle extra field.
3278 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3280 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3281 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3283 * config/arc/arc.c (arc_init): Update pic warning.
3284 (irq_range): Update irq range parsing warnings.
3285 (arc_override_options): Update various warning messages.
3286 (arc_handle_aux_attribute): Likewise.
3288 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3290 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3292 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3294 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3295 compiler messages directives.
3297 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3299 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3300 * config/mips/mips-tables.opt: Regenerate.
3301 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3303 * doc/invoke.texi: Document -march=i6500.
3305 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3307 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3308 (i6400_gpmul): Add cpu_unit.
3309 (i6400_gpdiv): Likewise.
3310 (i6400_msa_add_d): Update reservations.
3311 (i6400_msa_int_add) Likewise.
3312 (i6400_msa_short_logic3) Likewise.
3313 (i6400_msa_short_logic2) Likewise.
3314 (i6400_msa_short_logic) Likewise.
3315 (i6400_msa_move) Likewise.
3316 (i6400_msa_cmp) Likewise.
3317 (i6400_msa_short_float2) Likewise.
3318 (i6400_msa_div_d) Likewise.
3319 (i6400_msa_long_logic1) Likewise.
3320 (i6400_msa_long_logic2) Likewise.
3321 (i6400_msa_mult) Likewise.
3322 (i6400_msa_long_float2) Likewise.
3323 (i6400_msa_long_float4) Likewise.
3324 (i6400_msa_long_float5) Likewise.
3325 (i6400_msa_long_float8) Likewise.
3326 (i6400_fpu_fadd): Include frint type.
3327 (i6400_fpu_store): New define_insn_reservation.
3328 (i6400_fpu_load): Likewise.
3329 (i6400_fpu_move): Likewise.
3330 (i6400_fpu_fcmp): Likewise.
3331 (i6400_fpu_fmadd): Likewise.
3332 (i6400_int_mult): Include imul3nc type and update reservation.
3333 (i6400_int_div): Include idiv3 type and update reservation.
3334 (i6400_int_load): Update to check type not move_type.
3335 (i6400_int_store): Likewise.
3336 (i6400_int_prefetch): Set zero latency.
3338 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3340 * gcc.c: Document new %@{...} sequence.
3341 (LINK_COMMAND_SPEC): Use it for the -L switches.
3342 (cpp_unique_options): Use it for the -I switches.
3343 (at_file_argbuf): New global variable.
3344 (in_at_file): Likewise.
3345 (alloc_args): Create at_file_argbuf.
3346 (clear_args): Truncate at_file_argbuf.
3347 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3348 (open_at_file): New function.
3349 (close_at_file): Likewise.
3350 (create_at_file): Delete.
3351 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3354 (validate_switches_from_spec): Deal with %@{...} sequence.
3355 (validate_switches): Likewise.
3356 (driver::finalize): Call clear_args.
3358 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3360 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3362 2018-06-11 Martin Sebor <msebor@redhat.com>
3364 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3365 (Wno-class-memaccess): Add @opindex.
3366 (Wno-templates, Wno-multiple-inheritance): Same.
3367 (Wno-virtual-inheritance, Wno-namespaces): Same.
3368 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3369 (Wno-format-overflow, Wno-format-truncation): Same.
3370 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3371 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3372 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3373 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3374 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3375 (Wno-misspelled-isr): Same.
3377 2018-06-11 Martin Sebor <msebor@redhat.com>
3379 * PR tree-optimization/86083
3380 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3382 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3384 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3386 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3389 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3390 on the correct operand.
3391 (*movdi_internal64): Ditto.
3393 2018-06-11 Martin Liska <mliska@suse.cz>
3395 PR tree-optimization/86089
3396 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3398 2018-06-11 Julia Koval <julia.koval@intel.com>
3400 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3401 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3402 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3404 2018-06-11 Olivier Hainque <hainque@adacore.com>
3406 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3407 for Ada with strict dwarf2.
3409 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3412 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3415 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3417 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3419 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3421 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3424 2018-06-08 Martin Liska <mliska@suse.cz>
3426 * tree-cfg.h (debug_function): Fix argument type to match
3429 2018-06-08 Martin Liska <mliska@suse.cz>
3431 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3432 Remove usage of MPX-related (and removed) fields.
3433 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3435 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3437 * cfg.c (debug): Use TDF_NONE rather than 0.
3438 * cfghooks.c (debug): Likewise.
3439 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3440 (struct dump_option_value_info): Convert to...
3441 (struct kv_pair): ...this template type.
3442 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3444 (optinfo_verbosity_options): Likewise.
3445 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3447 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3448 than int for "optgroup_flags" param.
3449 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3451 (dump_dec): Likewise.
3452 (dump_finish): Use TDF_NONE rather than 0.
3453 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3454 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3455 than 0. Update for change to option_ptr.
3456 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3457 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3458 0. Update for changes to optinfo_verbosity_options and
3460 (opt_info_switch_p): Convert optgroup_flags from int to
3462 (dump_basic_block): Use dump_flags_t rather than int
3463 for "dump_kind" param.
3464 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3465 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3466 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3467 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3468 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3469 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3470 TDF_NONE): Convert from macros to...
3471 (enum dump_flag): ...this new enum.
3472 (dump_flags_t): Update to use enum.
3473 (operator|, operator&, operator~, operator|=, operator&=):
3474 Implement for dump_flags_t.
3475 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3476 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3477 Convert from macros to...
3478 (enum optgroup_flag): ...this new enum.
3479 (optgroup_flags_t): New typedef.
3480 (operator|, operator|=): Implement for optgroup_flags_t.
3481 (struct dump_file_info): Convert field "alt_flags" to
3482 dump_flags_t. Convert field "optgroup_flags" to
3484 (dump_basic_block): Use dump_flags_t rather than int for param.
3485 (dump_generic_expr_loc): Likewise.
3486 (dump_dec): Likewise.
3487 (dump_register): Convert param "optgroup_flags" to
3489 (opt_info_enable_passes): Likewise.
3490 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3492 * gimple-pretty-print.c (debug): Likewise.
3493 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3494 (merged_store_group::apply_stores): Likewise.
3495 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3496 * gimple.c (verify_gimple_pp): Likewise.
3497 * graphite-poly.c (print_pbb_body): Likewise.
3498 * passes.c (pass_manager::register_one_dump_file): Convert
3499 local "optgroup_flags" to optgroup_flags_t.
3500 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3502 (debug_body): Likewise.
3503 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3504 to optgroup_flags_t.
3505 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3507 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3508 (convert_mult_to_fma): Likewise.
3509 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3510 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3511 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3512 "dump_kind" to dump_flags_t.
3514 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3516 * config/rs6000/rs6000.c (min, max): Delete.
3518 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3520 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3521 -mabi=spe and -mabi=no-spe.
3523 2018-06-08 Martin Liska <mliska@suse.cz>
3525 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3526 where we expect an existing summary.
3528 2018-06-08 Martin Liska <mliska@suse.cz>
3530 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3531 * ipa-inline.h (estimate_edge_growth): Likewise.
3533 2018-06-08 Martin Liska <mliska@suse.cz>
3535 * cgraph.c (function_version_hasher::hash): Use
3536 cgraph_node::get_uid ().
3537 (function_version_hasher::equal):
3538 * cgraph.h (cgraph_node::get_uid): New method.
3539 * ipa-inline.c (update_caller_keys): Use
3540 cgraph_node::get_uid ().
3541 (update_callee_keys): Likewise.
3542 * ipa-utils.c (searchc): Likewise.
3543 (ipa_reduced_postorder): Likewise.
3544 * lto-cgraph.c (input_node): Likewise.
3545 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3546 * symbol-summary.h (symtab_insertion): Likewise.
3547 (symtab_removal): Likewise.
3548 (symtab_duplication): Likewise.
3549 * tree-pretty-print.c (dump_function_header): Likewise.
3550 * tree-sra.c (convert_callers_for_node): Likewise.
3552 2018-06-08 Martin Liska <mliska@suse.cz>
3554 * cgraph.c (symbol_table::create_edge): Always assign a new
3556 (symbol_table::free_edge): Do not recycle numbers.
3557 * cgraph.h (cgraph_edge::get): New method.
3558 * symbol-summary.h (symtab_removal): Use it.
3559 (symtab_duplication): Likewise.
3560 (call_summary::hashable_uid): Remove.
3562 2018-06-08 Martin Liska <mliska@suse.cz>
3564 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3565 (initialize_growth_caches): Remove.
3566 (free_growth_caches): Likewise.
3567 (do_estimate_edge_time): Use edge_growth_cache.
3568 (do_estimate_edge_size): Likewise.
3569 (do_estimate_edge_hints): Likewise.
3570 * ipa-inline.c (reset_edge_caches): Likewise.
3571 (recursive_inlining): Likewise.
3572 (inline_small_functions): Likewise.
3573 * ipa-inline.h (initialize_growth_caches): Remove.
3574 (estimate_edge_size): Likewise.
3575 (estimate_edge_time): Likewise.
3576 (estimate_edge_hints): Likewise.
3577 (reset_edge_growth_cache): Likewise.
3578 * symbol-summary.h (call_summary::remove): New method.
3580 2018-06-08 Martin Liska <mliska@suse.cz>
3582 * ipa-cp.c (class edge_clone_summary): New summary.
3583 (grow_edge_clone_vectors): Remove.
3584 (ipcp_edge_duplication_hook): Remove.
3585 (class edge_clone_summary_t): New call_summary class.
3586 (ipcp_edge_removal_hook): Remove.
3587 (edge_clone_summary_t::duplicate): New function.
3588 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3589 (create_specialized_node): Likewise.
3590 (ipcp_driver): Initialize edge_clone_summaries and do not
3593 2018-06-08 Martin Liska <mliska@suse.cz>
3595 * symbol-summary.h (get): New function.
3596 (call_summary::m_initialize_when_cloning): New class member.
3598 2018-06-08 Martin Liska <mliska@suse.cz>
3600 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3601 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3602 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3603 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3604 (remove_cgraph_node_from_order): Use the removed_nodes set.
3605 (do_per_function_toporder): Likwise.
3606 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3607 instead of summary_uid.
3608 (symtab_removal): Likewise.
3609 (symtab_duplication): Likewise.
3611 2018-06-08 Martin Liska <mliska@suse.cz>
3613 * ipa-cp.c (ipcp_store_bits_results): Use
3614 ipcp_transformation_sum.
3615 (ipcp_store_vr_results): Likewise.
3616 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3618 (ipcp_transformation_initialize): ... this.
3619 (ipa_set_node_agg_value_chain):
3620 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3621 (write_ipcp_transformation_info): Likewise.
3622 (read_ipcp_transformation_info): Likewise.
3623 (ipcp_update_bits): Likewise.
3624 (ipcp_update_vr): Likewise.
3625 (ipcp_transform_function): Likewise.
3626 * ipa-prop.h: Rename ipcp_transformation_summary to
3627 ipcp_transformation.
3628 (class ipcp_transformation_t): New function summary.
3629 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3630 (ipa_get_agg_replacements_for_node): Likewise.
3632 2018-06-08 Martin Liska <mliska@suse.cz>
3634 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3636 (class funct_state_summary_t): New function_summary class.
3637 (has_function_state): Remove.
3638 (get_function_state): Likewise.
3639 (set_function_state): Likewise.
3640 (add_new_function): Likewise.
3641 (funct_state_summary_t::insert): New function.
3642 (duplicate_node_data): Remove.
3643 (remove_node_data): Remove.
3644 (funct_state_summary_t::duplicate): New function.
3645 (register_hooks): Create new funct_state_summaries.
3646 (pure_const_generate_summary): Use it.
3647 (pure_const_write_summary): Likewise.
3648 (pure_const_read_summary): Likewise.
3649 (propagate_pure_const): Likewise.
3650 (propagate_nothrow): Likewise.
3651 (dump_malloc_lattice): Likewise.
3652 (propagate_malloc): Likewise.
3653 (execute): Do not register hooks, just remove summary
3655 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3658 2018-06-08 Martin Liska <mliska@suse.cz>
3660 * ipa-reference.c (remove_node_data): Remove.
3661 (duplicate_node_data): Likewise.
3662 (class ipa_ref_var_info_summary_t): New class.
3663 (class ipa_ref_opt_summary_t): Likewise.
3664 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3665 (get_reference_optimization_summary): Use
3666 ipa_ref_opt_sum_summaries.
3667 (set_reference_vars_info): Remove.
3668 (set_reference_optimization_summary): Likewise.
3669 (ipa_init): Create summaries.
3670 (init_function_info): Use function summary.
3671 (ipa_ref_opt_summary_t::duplicate): New function.
3672 (ipa_ref_opt_summary_t::remove): New function.
3673 (get_read_write_all_from_node): Fix GNU coding style.
3674 (propagate): Use function summary.
3675 (write_node_summary_p): Fix GNU coding style.
3676 (stream_out_bitmap): Likewise.
3677 (ipa_reference_read_optimization_summary): Use function summary.
3678 (ipa_reference_c_finalize): Do not release hooks.
3680 2018-06-08 Martin Liska <mliska@suse.cz>
3682 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3683 (analyze_function_body): Extract multiple calls of get_create.
3684 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3685 * ipa-inline.c (recursive_inlining): Use ::get method.
3686 * ipa-inline.h (estimate_edge_growth): Likewise.
3688 2018-06-08 Martin Liska <mliska@suse.cz>
3690 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3692 (hsa_function_summary::hsa_function_summary): Use the new enum
3694 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3695 * hsa-gen.c (hsa_get_host_function): Likewise.
3696 (get_brig_function_name): Likewise.
3697 * ipa-hsa.c (process_hsa_functions): Likewise.
3698 (ipa_hsa_write_summary): Likewise.
3699 * symbol-summary.h (symtab_duplication): Use ::get function/
3700 (get): New function.
3702 2018-06-08 Martin Liska <mliska@suse.cz>
3704 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3706 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3707 (hsa_register_kernel): Likewise.
3708 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3709 * hsa-gen.c (hsa_get_host_function): Likewise.
3710 (get_brig_function_name): Likewise.
3711 (generate_hsa): Likewise.
3712 (pass_gen_hsail::execute): Likewise.
3713 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3714 (devirtualization_time_bonus): Likewise.
3715 (ipcp_propagate_stage): Likewise.
3716 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3717 (edge_set_predicate): Likewise.
3718 (evaluate_conditions_for_known_args): Likewise.
3719 (evaluate_properties_for_edge): Likewise.
3720 (ipa_fn_summary::reset): Likewise.
3721 (ipa_fn_summary_t::duplicate): Likewise.
3722 (dump_ipa_call_summary): Likewise.
3723 (ipa_dump_fn_summary): Likewise.
3724 (analyze_function_body): Likewise.
3725 (compute_fn_summary): Likewise.
3726 (estimate_edge_devirt_benefit): Likewise.
3727 (estimate_edge_size_and_time): Likewise.
3728 (estimate_calls_size_and_time): Likewise.
3729 (estimate_node_size_and_time): Likewise.
3730 (inline_update_callee_summaries): Likewise.
3731 (remap_edge_change_prob): Likewise.
3732 (remap_edge_summaries): Likewise.
3733 (ipa_merge_fn_summary_after_inlining): Likewise.
3734 (ipa_update_overall_fn_summary): Likewise.
3735 (read_ipa_call_summary): Likewise.
3736 (inline_read_section): Likewise.
3737 (write_ipa_call_summary): Likewise.
3738 (ipa_fn_summary_write): Likewise.
3739 (ipa_free_fn_summary): Likewise.
3740 * ipa-hsa.c (process_hsa_functions): Likewise.
3741 (ipa_hsa_write_summary): Likewise.
3742 (ipa_hsa_read_section): Likewise.
3743 * ipa-icf.c (sem_function::merge): Likewise.
3744 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3745 (do_estimate_edge_time): Likewise.
3746 (estimate_size_after_inlining): Likewise.
3747 (estimate_growth): Likewise.
3748 (growth_likely_positive): Likewise.
3749 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3750 (inline_call): Likewise.
3751 * ipa-inline.c (caller_growth_limits): Likewise.
3752 (can_inline_edge_p): Likewise.
3753 (can_inline_edge_by_limits_p): Likewise.
3754 (compute_uninlined_call_time): Likewise.
3755 (compute_inlined_call_time): Likewise.
3756 (want_inline_small_function_p): Likewise.
3757 (edge_badness): Likewise.
3758 (update_caller_keys): Likewise.
3759 (update_callee_keys): Likewise.
3760 (recursive_inlining): Likewise.
3761 (inline_small_functions): Likewise.
3762 (inline_to_all_callers_1): Likewise.
3763 (dump_overall_stats): Likewise.
3764 (early_inline_small_functions): Likewise.
3765 (early_inliner): Likewise.
3766 * ipa-inline.h (estimate_edge_growth): Likewise.
3767 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3768 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3769 * ipa-prop.h (IPA_NODE_REF): Likewise.
3770 (IPA_EDGE_REF): Likewise.
3771 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3772 (propagate_malloc): Likewise.
3773 * ipa-split.c (execute_split_functions): Likewise.
3774 * symbol-summary.h: Rename get to get_create.
3776 (get_create): Likewise.
3777 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3779 2018-06-08 Martin Liska <mliska@suse.cz>
3781 * symbol-summary.h (release): Move definition out of class
3783 (symtab_removal): Likewise.
3784 (symtab_duplication): Likewise.
3786 2018-06-08 Martin Liska <mliska@suse.cz>
3788 * symbol-summary.h (function_summary): Move constructor
3789 implementation out of class declaration.
3790 (release): Likewise.
3791 (symtab_insertion): Likewise.
3792 (symtab_removal): Likewise.
3793 (symtab_duplication): Likewise.
3796 2018-06-08 Martin Liska <mliska@suse.cz>
3798 * Makefile.in: Remove support for MPX (macros, related functions,
3799 fields in cgraph_node, ...).
3800 * builtin-types.def (BT_BND): Likewise.
3801 (BT_FN_BND_CONST_PTR): Likewise.
3802 (BT_FN_CONST_PTR_BND): Likewise.
3803 (BT_FN_VOID_PTR_BND): Likewise.
3804 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3805 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3806 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3807 (expand_builtin_mempcpy_with_bounds): Likewise.
3808 (expand_builtin_memset_with_bounds): Likewise.
3809 (expand_builtin_memset_args): Likewise.
3810 (std_expand_builtin_va_start): Likewise.
3811 (expand_builtin): Likewise.
3812 (expand_builtin_with_bounds): Likewise.
3813 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3814 (DEF_LIB_BUILTIN_CHKP): Likewise.
3815 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3816 (DEF_CHKP_BUILTIN): Likewise.
3817 (BUILT_IN_MEMCPY): Likewise.
3818 (BUILT_IN_MEMMOVE): Likewise.
3819 (BUILT_IN_MEMPCPY): Likewise.
3820 (BUILT_IN_MEMSET): Likewise.
3821 (BUILT_IN_STPCPY): Likewise.
3822 (BUILT_IN_STRCAT): Likewise.
3823 (BUILT_IN_STRCHR): Likewise.
3824 (BUILT_IN_STRCPY): Likewise.
3825 (BUILT_IN_STRLEN): Likewise.
3826 (BUILT_IN_MEMCPY_CHK): Likewise.
3827 (BUILT_IN_MEMMOVE_CHK): Likewise.
3828 (BUILT_IN_MEMPCPY_CHK): Likewise.
3829 (BUILT_IN_MEMSET_CHK): Likewise.
3830 (BUILT_IN_STPCPY_CHK): Likewise.
3831 (BUILT_IN_STRCAT_CHK): Likewise.
3832 (BUILT_IN_STRCPY_CHK): Likewise.
3833 * calls.c (store_bounds): Likewise.
3834 (emit_call_1): Likewise.
3835 (special_function_p): Likewise.
3836 (maybe_warn_nonstring_arg): Likewise.
3837 (initialize_argument_information): Likewise.
3838 (finalize_must_preallocate): Likewise.
3839 (compute_argument_addresses): Likewise.
3840 (expand_call): Likewise.
3841 * cfgexpand.c (expand_call_stmt): Likewise.
3842 (expand_return): Likewise.
3843 (expand_gimple_stmt_1): Likewise.
3844 (pass_expand::execute): Likewise.
3845 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3846 (cgraph_node::remove): Likewise.
3847 (cgraph_node::dump): Likewise.
3848 (cgraph_node::verify_node): Likewise.
3849 * cgraph.h (chkp_function_instrumented_p): Likewise.
3850 (symtab_node::get_alias_target): Likewise.
3851 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3852 (cgraph_local_p): Likewise.
3853 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3854 (cgraph_edge::rebuild_references): Likewise.
3855 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3856 (walk_polymorphic_call_targets): Likewise.
3857 (cgraph_node::expand_thunk): Likewise.
3858 (symbol_table::output_weakrefs): Likewise.
3859 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3860 (ix86_handle_option): Likewise.
3861 * config/i386/constraints.md: Likewise.
3862 * config/i386/i386-builtin-types.def (BND): Likewise.
3866 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3867 (BDESC_FIRST): Likewise.
3869 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3870 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3871 * config/i386/i386.c (enum reg_class): Likewise.
3872 (ix86_target_string): Likewise.
3873 (ix86_option_override_internal): Likewise.
3874 (ix86_conditional_register_usage): Likewise.
3875 (ix86_valid_target_attribute_inner_p): Likewise.
3876 (ix86_set_indirect_branch_type): Likewise.
3877 (ix86_set_current_function): Likewise.
3878 (ix86_function_arg_regno_p): Likewise.
3879 (init_cumulative_args): Likewise.
3880 (ix86_function_arg_advance): Likewise.
3881 (ix86_function_arg): Likewise.
3882 (ix86_pass_by_reference): Likewise.
3883 (ix86_function_value_regno_p): Likewise.
3884 (ix86_function_value_1): Likewise.
3885 (ix86_function_value_bounds): Likewise.
3886 (ix86_return_in_memory): Likewise.
3887 (ix86_setup_incoming_vararg_bounds): Likewise.
3888 (ix86_va_start): Likewise.
3889 (indirect_thunk_need_prefix): Likewise.
3890 (print_reg): Likewise.
3891 (ix86_print_operand): Likewise.
3892 (ix86_expand_call): Likewise.
3893 (ix86_output_function_return): Likewise.
3894 (reg_encoded_number): Likewise.
3895 (BDESC_VERIFYS): Likewise.
3896 (ix86_init_mpx_builtins): Likewise.
3897 (ix86_init_builtins): Likewise.
3898 (ix86_emit_cmove): Likewise.
3899 (ix86_emit_move_max): Likewise.
3900 (ix86_expand_builtin): Likewise.
3901 (ix86_builtin_mpx_function): Likewise.
3902 (ix86_get_arg_address_for_bt): Likewise.
3903 (ix86_load_bounds): Likewise.
3904 (ix86_store_bounds): Likewise.
3905 (ix86_load_returned_bounds): Likewise.
3906 (ix86_store_returned_bounds): Likewise.
3907 (ix86_class_likely_spilled_p): Likewise.
3908 (ix86_hard_regno_mode_ok): Likewise.
3909 (x86_order_regs_for_local_alloc): Likewise.
3910 (ix86_mitigate_rop): Likewise.
3911 (ix86_bnd_prefixed_insn_p): Likewise.
3912 (ix86_mpx_bound_mode): Likewise.
3913 (ix86_make_bounds_constant): Likewise.
3914 (ix86_initialize_bounds): Likewise.
3915 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3916 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3917 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3918 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3919 (TARGET_CHKP_BOUND_MODE): Likewise.
3920 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3921 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3922 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3923 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3924 * config/i386/i386.h (TARGET_MPX): Likewise.
3925 (TARGET_MPX_P): Likewise.
3926 (VALID_BND_REG_MODE): Likewise.
3927 (FIRST_BND_REG): Likewise.
3928 (LAST_BND_REG): Likewise.
3929 (enum reg_class): Likewise.
3930 (BND_REG_P): Likewise.
3931 (BND_REGNO_P): Likewise.
3932 (BNDmode): Likewise.
3933 (ADJUST_INSN_LENGTH): Likewise.
3934 * config/i386/i386.md: Likewise.
3935 * config/i386/i386.opt: Likewise.
3936 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3937 (defined): Likewise.
3938 (LINK_MPX): Likewise.
3939 (MPX_SPEC): Likewise.
3940 (LIBMPX_SPEC): Likewise.
3941 (LIBMPXWRAPPERS_SPEC): Likewise.
3942 (CHKP_SPEC): Likewise.
3943 * config/i386/predicates.md: Likewise.
3944 * dbxout.c (dbxout_type): Likewise.
3945 * doc/extend.texi: Likewise.
3946 * doc/invoke.texi: Likewise.
3947 * doc/md.texi: Likewise.
3948 * doc/tm.texi: Likewise.
3949 * doc/tm.texi.in: Likewise.
3950 * dwarf2out.c (is_base_type): Likewise.
3951 (gen_formal_types_die): Likewise.
3952 (gen_subprogram_die): Likewise.
3953 (gen_type_die_with_usage): Likewise.
3954 (gen_decl_die): Likewise.
3955 (dwarf2out_late_global_decl): Likewise.
3956 * expr.c (expand_assignment): Likewise.
3957 (emit_storent_insn): Likewise.
3958 (store_expr_with_bounds): Likewise.
3959 (store_expr): Likewise.
3960 (expand_expr_real_1): Likewise.
3961 * expr.h (store_expr_with_bounds): Likewise.
3962 * function.c (use_register_for_decl): Likewise.
3963 (struct bounds_parm_data): Likewise.
3964 (assign_parms_augmented_arg_list): Likewise.
3965 (assign_parm_find_entry_rtl): Likewise.
3966 (assign_parm_is_stack_parm): Likewise.
3967 (assign_parm_load_bounds): Likewise.
3968 (assign_bounds): Likewise.
3969 (assign_parms): Likewise.
3970 (expand_function_start): Likewise.
3971 * gcc.c (CHKP_SPEC): Likewise.
3972 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3973 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3974 (wrestrict_dom_walker::check_call): Likewise.
3975 * gimple.c (gimple_build_call_from_tree): Likewise.
3976 * gimple.h (enum gf_mask): Likewise.
3977 (gimple_call_with_bounds_p): Likewise.
3978 (gimple_call_set_with_bounds): Likewise.
3979 * gimplify.c (gimplify_init_constructor): Likewise.
3980 * ipa-cp.c (initialize_node_lattices): Likewise.
3981 (propagate_constants_across_call): Likewise.
3982 (find_more_scalar_values_for_callers_subset): Likewise.
3983 * ipa-hsa.c (process_hsa_functions): Likewise.
3984 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3985 * ipa-icf.c (sem_function::merge): Likewise.
3986 * ipa-inline.c (early_inliner): Likewise.
3987 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3988 (warn_function_cold): Likewise.
3989 (propagate_pure_const): Likewise.
3990 * ipa-ref.h (enum GTY): Likewise.
3991 * ipa-split.c (find_retbnd): Likewise.
3992 (consider_split): Likewise.
3993 (split_function): Likewise.
3994 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3995 * ipa.c (walk_polymorphic_call_targets): Likewise.
3996 (symbol_table::remove_unreachable_nodes): Likewise.
3997 (process_references): Likewise.
3998 (cgraph_build_static_cdtor_1): Likewise.
3999 * lto-cgraph.c (lto_output_node): Likewise.
4000 (output_refs): Likewise.
4001 (compute_ltrans_boundary): Likewise.
4002 (input_overwrite_node): Likewise.
4003 (input_node): Likewise.
4004 (input_cgraph_1): Likewise.
4005 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4006 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4007 (class pass_chkp_instrumentation_passes): Likewise.
4008 (make_pass_chkp_instrumentation_passes): Likewise.
4009 * passes.def: Likewise.
4010 * rtl.h (struct GTY): Likewise.
4011 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4012 * stor-layout.c (layout_type): Likewise.
4013 * symtab.c: Likewise.
4014 * target.def: Likewise.
4015 * targhooks.c (default_chkp_bound_type): Likewise.
4016 (default_chkp_bound_mode): Likewise.
4017 (default_builtin_chkp_function): Likewise.
4018 (default_chkp_function_value_bounds): Likewise.
4019 (default_chkp_make_bounds_constant): Likewise.
4020 (default_chkp_initialize_bounds): Likewise.
4021 * targhooks.h (default_chkp_bound_type): Likewise.
4022 (default_chkp_bound_mode): Likewise.
4023 (default_builtin_chkp_function): Likewise.
4024 (default_chkp_function_value_bounds): Likewise.
4025 (default_chkp_make_bounds_constant): Likewise.
4026 (default_chkp_initialize_bounds): Likewise.
4027 * toplev.c (compile_file): Likewise.
4028 (process_options): Likewise.
4029 * tree-core.h (DEF_BUILTIN): Likewise.
4030 (DEF_BUILTIN_CHKP): Likewise.
4031 * tree-inline.c (declare_return_variable): Likewise.
4032 (remap_gimple_stmt): Likewise.
4033 (copy_bb): Likewise.
4034 (initialize_inlined_parameters): Likewise.
4035 (expand_call_inline): Likewise.
4036 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4037 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4038 (make_pass_ipa_chkp_produce_thunks): Likewise.
4039 (make_pass_chkp): Likewise.
4040 (make_pass_chkp_opt): Likewise.
4041 (make_pass_chkp_instrumentation_passes): Likewise.
4042 * tree-pretty-print.c (dump_generic_node): Likewise.
4043 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4044 * tree-ssa-dce.c (propagate_necessity): Likewise.
4045 (eliminate_unnecessary_stmts): Likewise.
4046 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4047 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4048 * tree-ssa-sccvn.h: Likewise.
4049 * tree-ssa-strlen.c (get_string_length): Likewise.
4050 (valid_builtin_call): Likewise.
4051 (adjust_last_stmt): Likewise.
4052 (handle_builtin_strchr): Likewise.
4053 (handle_builtin_strcpy): Likewise.
4054 (handle_builtin_stxncpy): Likewise.
4055 (handle_builtin_memcpy): Likewise.
4056 (handle_builtin_strcat): Likewise.
4057 (strlen_check_and_optimize_stmt): Likewise.
4058 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4059 * tree-streamer-in.c: Likewise.
4060 * tree-streamer.c (record_common_node): Likewise.
4061 * tree.c (tree_code_size): Likewise.
4062 (wide_int_to_tree_1): Likewise.
4063 (type_contains_placeholder_1): Likewise.
4064 (build_common_tree_nodes): Likewise.
4065 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4066 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4067 (POINTER_BOUNDS_P): Likewise.
4068 (BOUNDED_TYPE_P): Likewise.
4069 (BOUNDED_P): Likewise.
4070 (CALL_WITH_BOUNDS_P): Likewise.
4071 (pointer_bounds_type_node): Likewise.
4072 * value-prof.c (gimple_ic): Likewise.
4073 * var-tracking.c (vt_add_function_parameters): Likewise.
4074 * varasm.c (make_decl_rtl): Likewise.
4075 (assemble_start_function): Likewise.
4076 (output_constant): Likewise.
4077 (maybe_assemble_visibility): Likewise.
4078 * varpool.c (ctor_for_folding): Likewise.
4079 * chkp-builtins.def: Remove.
4080 * ipa-chkp.c: Remove.
4081 * ipa-chkp.h: Remove.
4082 * rtl-chkp.c: Remove.
4083 * rtl-chkp.h: Remove.
4084 * tree-chkp-opt.c: Remove.
4085 * tree-chkp.c: Remove.
4086 * tree-chkp.h: Remove.
4088 2018-06-07 Carl Love <cel@us.ibm.com>
4090 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4091 vextract_fp_from_shortl): Add BE support.
4093 2018-06-07 Paul Koning <ni1d@arrl.net>
4095 * compare-elim.c (try_merge_compare): Don't merge compare if
4096 address contains a side effect.
4097 (try_eliminate_compare): Likewise.
4099 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4101 * config.gcc: Support "tremont".
4102 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4103 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4105 * config/i386/i386.c (m_TREMONT): Define.
4106 (processor_target_table): Add "tremont".
4107 (PTA_TREMONT): Define.
4108 (ix86_lea_outperforms): Add TARGET_TREMONT.
4109 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4110 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4111 and M_INTEL_GOLDMONT_PLUS.
4112 (fold_builtin_cpu): Add "tremont".
4113 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4114 (ix86_option_override_internal): Add "tremont".
4115 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4116 (processor_type): Add PROCESSOR_TREMONT.
4117 * config/i386/x86-tune.def: Add m_TREMONT.
4118 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4120 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4122 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4123 symbol defined for msp430i* devices to be lower case.
4125 gcc/testsuite/gcc.target/msp430/
4126 * msp430i-device-symbol.c: New test.
4127 * msp430f-device-symbol.c: New test.
4128 * msp430.h: New test header file.
4130 2018-06-07 Richard Biener <rguenther@suse.de>
4132 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4133 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4134 Properly wrap signed arithmetic if overflow wraps.
4136 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4138 PR tree-optimization/69615
4139 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4140 of a cast from a same precision integral SSA_NAME in a bb dominated
4141 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4142 cast to utype if rhs2 has already a compatible type.
4144 2018-06-07 Richard Biener <rguenther@suse.de>
4146 PR tree-optimization/85935
4147 * graphite-scop-detection.c (find_params_in_bb): Analyze
4148 condition operands with respect to the correct loop. Assert
4149 the analysis doesn't fail.
4151 2018-06-04 Carl Love <cel@us.ibm.com>
4153 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4154 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4155 as it is slightly cheaper.
4156 (first_match_or_eos_index_<mode>):
4157 Calculate index using natural element order.
4158 (first_match_index_<mode>):
4159 Calculate index using natural element order.
4160 (first_match_or_eos_index_<mode>):
4161 Calculate index using natural order.
4162 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4163 for BE and LE modes.
4164 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4165 P9V_BUILTIN_VCLZLSBB_V16QI.
4166 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4169 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4171 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4172 indentation and line wrap for many prototypes. Add missing
4173 @smallexample directives around block of prototypes for vec_xl and
4176 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4178 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4179 track if we pass or return IEEE 128-bit floating point.
4180 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4181 C++ mangling that is compatible with GCC 8.1.
4182 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4183 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4184 floating point types.
4185 (rs6000_function_arg_advance_1): Likewise.
4186 (rs6000_mangle_type): Optionally generate mangled names that match
4187 what GCC 8.1 generated for IEEE 128-bit floating point types.
4188 (rs6000_globalize_decl_name): If we have an external function that
4189 passes or returns IEEE 128-bit types, generate a weak reference
4190 from the mangled name used in GCC 8.1 to the current mangled
4192 (rs6000_init_builtins): Make __ibm128 use the long double type if
4193 long double is IBM extended double. Make __float128 use the long
4194 double type if long double is IEEE 128-bit.
4197 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4198 macro for __ibm128 built-in functions.
4199 (PACK_IF): Add __ibm128 pack/unpack functions.
4200 (UNPACK_IF): Likewise.
4201 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4202 enable long double built-in functions if long double is IEEE
4203 128-bit floating point.
4204 (rs6000_invalid_builtin): Update long double built-in function
4206 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4207 functions, adjust the built-in function to use the long double
4208 built-in function if __ibm128 and long double are the same type.
4209 * doc/extend.texi (PowerPC builtins): Update documention for
4210 __builtin_{,un}pack_longdouble. Add documentation for
4211 __builtin_{,un}pack_ibm128.
4213 2018-06-06 Jim Wilson <jimw@sifive.com>
4215 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4216 (struct machine_function): New field interrupt_mode.
4217 (riscv_handle_type_attribute): New function. Add forward declaration.
4218 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4219 (riscv_expand_epilogue): Check interrupt_mode field.
4220 (riscv_set_current_function): Check interrupt attribute args and
4221 set interrupt_mode field.
4222 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4223 (riscv_sret, riscv_uret): New.
4224 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4225 new arguments to interrupt attribute.
4227 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4230 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4231 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4233 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4235 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4236 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4237 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4238 several redundant entries.
4240 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4242 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4243 type from "rtx" to "rtx_insn *".
4244 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4245 for local "call_insn", removing cast.
4246 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4248 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4250 PR tree-optimization/86066
4251 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4252 for BIT_INSERT_EXPR stores.
4254 2018-06-06 Richard Biener <rguenther@suse.de>
4256 PR tree-optimization/86062
4257 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4258 component refs ontop
4259 of to be offsetted base.
4261 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4263 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4264 to be static and remove check on interrupt attribute name.
4266 gcc/testsuite/gcc.target/msp430/
4267 * function-attributes-4.c: New test.
4268 * static-interrupts.c: New test.
4270 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4272 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4273 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4275 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4278 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4280 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4281 Remove second argument, change how error is called.
4282 (aarch64_layout_arg): Remove second argument from
4283 aarch64_err_no_fpadvsimd call.
4284 (aarch64_init_cumulative_args): Ditto.
4285 (aarch64_gimplify_va_arg_expr): Ditto.
4286 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4288 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4290 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4291 (*simple_return_indirect_internal<mode>): Rename from
4292 simple_return_indirect_internal. Use W mode iterator.
4293 (rstorssp): New expander.
4294 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4295 (clrssbsy): New expander.
4296 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4298 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4300 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4302 (cmse_check_pointed_object): Likewise.
4304 2018-06-05 Martin Liska <mliska@suse.cz>
4306 PR gcov-profile/47618
4307 * doc/invoke.texi: Document how -fprofile-dir format
4310 2018-06-05 Richard Biener <rguenther@suse.de>
4312 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4313 removal pretend DOM info isn't available so we do not update
4314 it and only remove edges, not dominated blocks. Actually free
4315 DOM info in case we removed something. Remove unreachable blocks.
4316 (mfb_keep_latches): Work with either DOM info or marked backedges.
4317 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4318 first. Mark backedges if DOM info isn't available.
4319 (Re-)compute DOM info after cleanup_control_flow_pre.
4321 2018-06-05 Richard Biener <rguenther@suse.de>
4323 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4324 (locus_discrim_hasher::hash): Adjust.
4325 (locus_discrim_hasher::equal): Likewise.
4326 (next_discriminator_for_locus): Work on line directly.
4327 (same_line_p): Pass in expanded locus1 as well.
4328 (assign_discriminators): Avoid redundant location expansions.
4330 2018-06-05 Richard Biener <rguenther@suse.de>
4332 PR tree-optimization/86046
4333 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4334 if required after clearing TREE_ADDRESSABLE.
4336 2018-06-05 Richard Biener <rguenther@suse.de>
4338 PR tree-optimization/86047
4339 * tree-ssa-loop.c (for_each_index): Glob handling of all
4340 decls and constants and really handle all of them.
4342 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4345 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4346 qualifier_void_pointer and qualifier_const_void_pointer.
4347 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4348 (arm_init_builtins): Handle the above.
4349 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4350 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4351 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4354 2018-06-05 Martin Liska <mliska@suse.cz>
4356 * auto-profile.c (read_autofdo_file): Do not use
4357 gcov_ctr_summary struct.
4358 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4359 * coverage.c (struct counts_entry): Likewise.
4360 (read_counts_file): Read just single summary entry.
4361 (get_coverage_counts): Use gcov_summary struct.
4362 * coverage.h (get_coverage_counts): Likewise.
4363 * gcov-dump.c (dump_working_sets): Likewise.
4364 (tag_summary): Dump just single summary.
4365 * gcov-io.c (gcov_write_summary): Write just histogram
4367 (gcov_read_summary): Read just single summary.
4368 (compute_working_sets): Use gcov_summary struct.
4369 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4370 of GCOV_COUNTERS_SUMMABLE.
4371 (GCOV_COUNTERS_SUMMABLE): Remove.
4372 (GCOV_FIRST_VALUE_COUNTER): Replace with
4373 GCOV_COUNTER_V_INTERVAL.
4374 (struct gcov_ctr_summary): Remove.
4375 (struct gcov_summary): Directly use fields of former
4377 (compute_working_sets): Use gcov_summary struct.
4378 * gcov.c (read_count_file): Do not use ctrs fields.
4379 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4381 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4383 * profile.c: Likewise.
4384 * profile.h: Likewise.
4386 2018-06-05 Martin Liska <mliska@suse.cz>
4388 PR gcov-profile/84846
4389 * gcov.c (output_lines): Print working directory only
4390 in intermediate format.
4392 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4394 * config/s390/s390-builtin-types.def: Add void function type.
4395 * config/s390/s390-builtins.def: Use the function type for the
4398 2018-06-04 Jim Wilson <jimw@sifive.com>
4400 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4402 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4403 and maybe_eh_return. Change regno to unsigned int. Use new args to
4404 handle EH_RETURN_DATA_REGNO registers properly.
4405 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4406 (riscv_expand_epilogue): Update comment. Change argument name and
4407 type. Update code to use new name and type. Pass new args to
4408 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4410 * config/riscv/riscv.md (NORMAL_RETURN): New.
4411 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4412 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4413 (eh_return): Call gen_eh_return_internal and emit barrier.
4414 (eh_return_internal): Call riscv_expand_epilogue.
4416 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4418 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4419 bit_insertion field and declare can_be_merged_into method.
4420 (merged_store_group::can_be_merged_into): New method.
4421 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4422 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4423 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4425 2018-06-04 Richard Biener <rguenther@suse.de>
4427 PR tree-optimization/85955
4428 * builtins.c (fold_builtin_sincos): Convert pointers to
4429 destination to appropriate type before dereferencing.
4431 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4433 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4435 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4437 * expr.c (expand_expr_real_1): Force the operand into memory if
4438 its TYPE_MODE is BLKmode and if there is no integer mode for
4439 the number of bits being extracted.
4441 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4445 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4446 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4448 2018-06-04 Richard Biener <rguenther@suse.de>
4450 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4451 (cleanup_tree_cfg_noloop): ... single caller. Do
4452 start_recording_case_labels later.
4454 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4456 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4457 to _IMMINTRIN_H_INCLUDED.
4458 * config/i386/pconfigintrin.h: Ditto.
4459 * config/i386/waitpkgintrin.h: Ditto.
4460 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4461 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4462 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4463 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4464 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4465 waitpkgintrin.h and cldemoteintrin.h.
4467 2018-06-04 Richard Biener <rguenther@suse.de>
4469 PR tree-optimization/86038
4470 * tracer.c (find_best_successor): Check probability for
4471 being initialized, bail out if not.
4473 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4476 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4477 of bits to ignore when comparing architectures.
4479 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4481 PR tree-optimization/69615
4482 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4483 maximum or minimum of the type, try to merge it also as if
4484 range1 is + [-, x - 1] or + [x + 1, -].
4487 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4489 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4491 PR tree-optimization/86034
4492 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4493 the unsigned bitfield type in a bit insertion sequence if it does not
4494 have a larger precision than the bitfield size.
4495 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4497 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4499 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4501 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4503 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4504 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4505 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4506 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4508 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4510 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4511 Disable -fdelete-null-pointer-checks for ELF toolchain.
4513 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4514 Kito Cheng <kito.cheng@gmail.com>
4516 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4517 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4518 (nds32*-*-*): Add glibc and uclibc conditions.
4519 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4520 (TARGET_EXCEPT_UNWIND_INFO): Define.
4521 * config/nds32/elf.h: New file.
4522 * config/nds32/linux.h: New file.
4523 * config/nds32/nds32-elf.opt: New file.
4524 * config/nds32/nds32-linux.opt: New file.
4525 * config/nds32/nds32-fp-as-gp.c
4526 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4527 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4529 (nds32_asm_file_end): Ditto.
4530 (nds32_print_operand): Ditto.
4531 (nds32_insert_attributes): Ditto.
4532 (nds32_init_libfuncs): New function.
4533 (TARGET_HAVE_TLS): Define.
4534 (TARGET_INIT_LIBFUNCS): Define.
4535 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4537 (TARGET_ELF): Apply different mcmodel setting.
4538 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4539 been migrated into elf.h and linux.h files.
4540 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4541 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4542 (mcmodel): The content has been migrated into nds32-elf.opt and
4543 nds32-linux.opt files.
4544 * config/nds32/t-elf: New file.
4545 * config/nds32/t-linux: New file.
4547 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4548 Shiva Chen <shiva0217@gmail.com>
4550 * config/nds32/constants.md (unspec_volatile_element): Add
4551 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4552 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4554 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4555 (make_pass_nds32_fp_as_gp): Declare.
4556 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4558 (nds32_asm_function_end_prologue): Remove unused asm output.
4559 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4560 (nds32_asm_file_start): Output necessary fp_as_gp information.
4561 (nds32_option_override): Adjust register usage.
4562 (nds32_expand_prologue): Consider fp_as_gp situation.
4563 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4564 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4567 (simple_return): Ditto.
4568 (omit_fp_begin): Output special directive for fp_as_gp.
4569 (omit_fp_end): Output special directive for fp_as_gp.
4570 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4571 mforbid-fp-as-gp): New options.
4573 2018-06-01 Mark Wielaard <mark@klomp.org>
4575 * dwarf2out.c (dwarf2out_finish): Remove generation of
4576 DW_AT_loclists_base.
4578 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4580 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4581 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4582 (struct merged_store_group): Add bit_insertion field.
4583 (dump_char_array): Use standard hexadecimal format.
4584 (merged_store_group::merged_store_group): Set bit_insertion to false.
4585 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4586 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4587 also print the mask in the dump file.
4588 (pass_store_merging::gate): Minor tweak.
4589 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4590 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4591 stores with INTEGER_CST stores.
4592 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4593 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4594 and use it throughout. Generate bit insertion sequences if need be.
4595 (pass_store_merging::process_store): Remove redundant condition.
4596 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4598 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4600 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4601 the 128-bit floating point types. Fix function comment.
4603 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4605 * config/aarch64/aarch64-simd.md
4606 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4608 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4611 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4613 PR tree-optimization/85989
4614 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4616 (backprop::intersect_uses): Check it when deciding whether this
4617 is a backedge reference.
4618 (backprop::process_block): Add each phi to m_visited_phis
4619 after visiting it, then clear it at the end.
4621 2018-06-01 Richard Biener <rguenther@suse.de>
4623 * tree-vectorizer.h (vect_dr_stmt): New function.
4624 (vect_get_load_cost): Adjust.
4625 (vect_get_store_cost): Likewise.
4626 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4627 Use vect_dr_stmt instead of DR_SMTT.
4628 (vect_record_base_alignments): Likewise.
4629 (vect_calculate_target_alignment): Likewise.
4630 (vect_compute_data_ref_alignment): Likewise and make static.
4631 (vect_update_misalignment_for_peel): Likewise.
4632 (vect_verify_datarefs_alignment): Likewise.
4633 (vector_alignment_reachable_p): Likewise.
4634 (vect_get_data_access_cost): Likewise. Pass down
4635 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4636 (vect_get_peeling_costs_all_drs): Likewise.
4637 (vect_peeling_hash_get_lowest_cost): Likewise.
4638 (vect_enhance_data_refs_alignment): Likewise.
4639 (vect_find_same_alignment_drs): Likewise.
4640 (vect_analyze_data_refs_alignment): Likewise.
4641 (vect_analyze_group_access_1): Likewise.
4642 (vect_analyze_group_access): Likewise.
4643 (vect_analyze_data_ref_access): Likewise.
4644 (vect_analyze_data_ref_accesses): Likewise.
4645 (vect_vfa_segment_size): Likewise.
4646 (vect_small_gap_p): Likewise.
4647 (vectorizable_with_step_bound_p): Likewise.
4648 (vect_prune_runtime_alias_test_list): Likewise.
4649 (vect_analyze_data_refs): Likewise.
4650 (vect_supportable_dr_alignment): Likewise.
4651 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4652 (vect_gen_prolog_loop_niters): Likewise.
4653 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4654 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4656 (vect_recog_mask_conversion_pattern): Likewise.
4657 (vect_try_gather_scatter_pattern): Likewise.
4658 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4659 to vect_get_store_cost.
4660 (vect_get_store_cost): Get stmt_info instead of DR.
4661 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4662 (vect_get_load_cost): Get stmt_info instead of DR.
4664 2018-06-01 Richard Biener <rguenther@suse.de>
4667 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4668 vars wrapped in MEM_REFs.
4670 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4672 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4673 Fix subreg tests so that we only return a choice between
4674 GENERAL_REGS and FP_REGS if the original classes included both.
4676 2018-06-01 Richard Biener <rguenther@suse.de>
4679 * tree-ssa-structalias.c (get_function_part_constraint):
4680 Handle NULL fi->decl.
4681 (find_func_aliases_for_call): Properly handle indirect
4682 fi from direct call.
4683 (find_func_clobbers): Likewise.
4684 (ipa_pta_execute): Likewise.
4685 (create_variable_info_for): For functions that are ifunc_resolver
4686 resolve to a varinfo that contains the result of the resolver call.
4687 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4690 2018-05-31 Michael Collison <michael.collison@arm.com>
4692 * config/aarch64/aarch64.md:
4693 (*fix_to_zero_extenddfdi2): New pattern.
4694 * gcc.target/aarch64/fix_extend1.c: New testcase.
4696 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4700 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4701 and BUILT_IN_STRNCMP_EQ.
4702 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4703 BUILT_IN_STRNCMP_EQ.
4704 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4705 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4706 (gimple_fold_builtin): Likewise.
4707 * tree-ssa-strlen.c (compute_string_length): New function.
4708 (determine_min_obsize): New function.
4709 (handle_builtin_string_cmp): New function to handle calls to
4710 string compare functions.
4711 (strlen_optimize_stmt): Add handling to builtin string compare
4713 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4714 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4715 * tree.c (build_common_builtin_nodes): Add new defines of
4716 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4718 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4721 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4722 functions with naked attribute.
4724 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4726 * config/i386/sse.md (avx_vec_concat<mode>):
4727 Substitute concat_tg_mode mode attribute with xtg_mode.
4728 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4729 (concat_tg_mode): Remove mode attribute.
4731 2018-05-31 Martin Sebor <msebor@redhat.com>
4734 * calls.c (alloc_max_size): Correct a logic error/typo.
4735 Treat excessive arguments as infinite. Warn for invalid arguments.
4736 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4738 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4741 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4742 and movx for Haswell.
4744 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4745 Cesar Philippidis <cesar@codesourcery.com>
4748 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4749 when emitting error on private/firstprivate reductions.
4750 * omp-low.c (lower_omp_target): Avoid reference-type processing
4751 on pointers for firstprivate clause.
4753 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4755 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4758 * config/aarch64/aarch64-simd.md
4759 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4760 (aarch64_ld1_x3_<mode>): Likewise
4761 (aarch64_st1x2<VALLDIF:mode>): Likewise
4762 (aarch64_st1_x2_<mode>): Likewise
4763 (aarch64_st1x3<VALLDIF:mode>): Likewise
4764 (aarch64_st1_x3_<mode>): Likewise
4765 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4766 (vld1_s8_x3): Likewise.
4767 (vld1_u16_x3): Likewise.
4768 (vld1_s16_x3): Likewise.
4769 (vld1_u32_x3): Likewise.
4770 (vld1_s32_x3): Likewise.
4771 (vld1_u64_x3): Likewise.
4772 (vld1_s64_x3): Likewise.
4773 (vld1_f16_x3): Likewise.
4774 (vld1_f32_x3): Likewise.
4775 (vld1_f64_x3): Likewise.
4776 (vld1_p8_x3): Likewise.
4777 (vld1_p16_x3): Likewise.
4778 (vld1_p64_x3): Likewise.
4779 (vld1q_u8_x3): Likewise.
4780 (vld1q_s8_x3): Likewise.
4781 (vld1q_u16_x3): Likewise.
4782 (vld1q_s16_x3): Likewise.
4783 (vld1q_u32_x3): Likewise.
4784 (vld1q_s32_x3): Likewise.
4785 (vld1q_u64_x3): Likewise.
4786 (vld1q_s64_x3): Likewise.
4787 (vld1q_f16_x3): Likewise.
4788 (vld1q_f32_x3): Likewise.
4789 (vld1q_f64_x3): Likewise.
4790 (vld1q_p8_x3): Likewise.
4791 (vld1q_p16_x3): Likewise.
4792 (vld1q_p64_x3): Likewise.
4793 (vst1_s64_x2): Likewise.
4794 (vst1_u64_x2): Likewise.
4795 (vst1_f64_x2): Likewise.
4796 (vst1_s8_x2): Likewise.
4797 (vst1_p8_x2): Likewise.
4798 (vst1_s16_x2): Likewise.
4799 (vst1_p16_x2): Likewise.
4800 (vst1_s32_x2): Likewise.
4801 (vst1_u8_x2): Likewise.
4802 (vst1_u16_x2): Likewise.
4803 (vst1_u32_x2): Likewise.
4804 (vst1_f16_x2): Likewise.
4805 (vst1_f32_x2): Likewise.
4806 (vst1_p64_x2): Likewise.
4807 (vst1q_s8_x2): Likewise.
4808 (vst1q_p8_x2): Likewise.
4809 (vst1q_s16_x2): Likewise.
4810 (vst1q_p16_x2): Likewise.
4811 (vst1q_s32_x2): Likewise.
4812 (vst1q_s64_x2): Likewise.
4813 (vst1q_u8_x2): Likewise.
4814 (vst1q_u16_x2): Likewise.
4815 (vst1q_u32_x2): Likewise.
4816 (vst1q_u64_x2): Likewise.
4817 (vst1q_f16_x2): Likewise.
4818 (vst1q_f32_x2): Likewise.
4819 (vst1q_f64_x2): Likewise.
4820 (vst1q_p64_x2): Likewise.
4821 (vst1_s64_x3): Likewise.
4822 (vst1_u64_x3): Likewise.
4823 (vst1_f64_x3): Likewise.
4824 (vst1_s8_x3): Likewise.
4825 (vst1_p8_x3): Likewise.
4826 (vst1_s16_x3): Likewise.
4827 (vst1_p16_x3): Likewise.
4828 (vst1_s32_x3): Likewise.
4829 (vst1_u8_x3): Likewise.
4830 (vst1_u16_x3): Likewise.
4831 (vst1_u32_x3): Likewise.
4832 (vst1_f16_x3): Likewise.
4833 (vst1_f32_x3): Likewise.
4834 (vst1_p64_x3): Likewise.
4835 (vst1q_s8_x3): Likewise.
4836 (vst1q_p8_x3): Likewise.
4837 (vst1q_s16_x3): Likewise.
4838 (vst1q_p16_x3): Likewise.
4839 (vst1q_s32_x3): Likewise.
4840 (vst1q_s64_x3): Likewise.
4841 (vst1q_u8_x3): Likewise.
4842 (vst1q_u16_x3): Likewise.
4843 (vst1q_u32_x3): Likewise.
4844 (vst1q_u64_x3): Likewise.
4845 (vst1q_f16_x3): Likewise.
4846 (vst1q_f32_x3): Likewise.
4847 (vst1q_f64_x3): Likewise.
4848 (vst1q_p64_x3): Likewise.
4850 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4852 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4853 user_label_prefix to name.
4855 * tree-core.h: Update comment about the format of NAME string
4856 passed to handler in attribute_spec.
4858 * config/msp430/msp430.md: Remove erroneous subreg expression from
4859 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4860 zero_extend{q,h}isi2.
4862 2018-05-30 Borislav Petkov <bp@suse.de>
4864 * doc/extend.texi: Document some architecture specific
4865 constraints and sort entries.
4867 2018-05-30 Martin Sebor <msebor@redhat.com>
4870 * builtins.c (expand_builtin_stpcpy_1): New function.
4871 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4872 only if the former succeeds.
4874 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4876 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4879 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4881 * doc/invoke.texi (-flinker-output): Document
4883 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4885 * passes.c (ipa_write_summaries): Only modify statements if body
4887 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4888 incrementally linking.
4889 (ipa_passes): Likewise.
4890 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4891 pass down resolution info.
4892 * common.opt (flag_incremental_link): Update info.
4893 * gcc.c (plugin specs): Turn flinker-output=* to
4894 -plugin-opt=-linker-output-known
4895 * toplev.c (compile_file): Also cut compilation when doing incremental
4897 * flag-types. (enum lto_partition_model): Add
4898 LTO_LINKER_OUTPUT_NOLTOREL.
4899 (invoke.texi): Add -flinker-output docs.
4900 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4901 link same way as WPA; do not stream in dead initializers.
4903 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4904 darf2out_register_external_decl): Support incremental link.
4906 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4908 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4910 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4912 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4913 it down to simple_object_copy_lto_debug_sections.
4914 (run_gcc): Determine incremental LTO link time and configure
4915 lto1 into non-wpa mode, disable renaming of debug sections.
4917 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4919 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4920 descriptions of various incorrectly documented functions.
4922 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4925 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4926 address check not strict.
4928 2018-05-30 Richard Biener <rguenther@suse.de>
4930 PR tree-optimization/85964
4931 * tracer.c (better_p): Drop initialized count check, we only
4932 call the function with initialized counts now.
4933 (find_best_successor): Do find a best edge if one
4934 has uninitialized count.
4935 (find_best_predecessor): Likewise. Do BB frequency check only
4936 if count is initialized.
4938 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4940 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4941 (aarch64_ldrstr_offset_compare): New.
4942 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4943 load/store orderings.
4944 (aarch64_gen_adjusted_ldpstp): Likewise.
4946 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4948 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4949 Check for subset of GENERAL_REGS and FP_REGS.
4950 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4953 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4955 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4956 and wi::to_poly_offset. Add the current offset and then check
4957 whether the sum fits, rather than using an unchecked addition of
4958 a checked term. Check for a shwi rather than a uhwi.
4959 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4960 (store_constructor): Use poly_int_tree_p.
4961 (expand_expr_real_1): Likewise.
4962 * function.c (assign_temp): Likewise.
4963 * fold-const.c (const_binop): Use poly_int_tree_p and
4965 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4967 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4968 to_poly_offset for MEM offsets.
4969 * ipa-icf.c (sem_variable::equals): Likewise.
4970 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4971 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4972 wi::to_poly_offset for BIT_FIELD_REF offsets.
4973 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4975 * var-tracking.c (emit_note_insn_var_location): Use
4976 tree_to_poly_uint64.
4978 2018-05-29 Jim Wilson <jimw@sifive.com>
4980 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4982 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4985 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4986 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4988 (sse4_1_round<mode>2): Use nonimmediate_operand
4989 for operand 1 predicate.
4991 2018-05-29 Martin Sebor <msebor@redhat.com>
4992 Richard Biener <rguenther@suse.de>
4995 * calls.c (get_size_range): Call determine_value_range instead
4996 of get_value_range..
4997 * tree-vrp.h (determine_value_range): Declared new function.
4998 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5000 2018-05-29 Richard Biener <rguenther@suse.de>
5002 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5003 sure to use non-pattern stmts for get_earlier_stmt arguments.
5004 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5005 called on pattern stmts.
5006 (get_later_stmt): Likewise.
5008 2018-05-29 Martin Liska <mliska@suse.cz>
5010 PR gcov-profile/85759
5011 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5014 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5016 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5018 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5022 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5023 VEC_PACK_FLOAT_EXPR): New tree codes.
5024 * tree-pretty-print.c (op_code_prio): Handle
5025 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5026 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5027 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5028 * tree-inline.c (estimate_operator_cost): Likewise.
5029 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5030 * fold-const.c (const_binop): Likewise.
5031 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5032 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5033 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5034 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5035 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5036 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5037 * expr.c (expand_expr_real_2): Likewise.
5038 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5039 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5040 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5042 * optabs.c (expand_widen_pattern_expr): For
5043 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5044 sign from result type rather than operand's type.
5045 (expand_binop_directly): For vec_packu_float_optab and
5046 vec_packs_float_optab allow result type to be different from operand's
5048 * optabs-tree.c (optab_for_tree_code): Handle
5049 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5050 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5051 * tree-vect-generic.c (expand_vector_operations_1): Handle
5052 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5053 VEC_PACK_FLOAT_EXPR.
5054 * tree-vect-stmts.c (supportable_widening_operation): Handle
5056 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5057 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5058 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5059 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5060 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5062 (vec_pack<floatprefix>_float_<mode>): New expander.
5063 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5065 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5066 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5067 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5068 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5069 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5071 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5072 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5073 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5074 VEC_PACK_FLOAT_EXPR): Document.
5076 2018-05-29 Richard Biener <rguenther@suse.de>
5078 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5080 (stmt_vec_info_vec): Make pointer.
5081 (init_stmt_vec_info_vec): Remove.
5082 (free_stmt_vec_info_vec): Likewise.
5083 (set_stmt_vec_info_vec): New function.
5084 (free_stmt_vec_infos): Likewise.
5085 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5086 (set_vinfo_for_stmt): Likewise.
5087 (get_earlier_stmt): Likewise.
5088 (get_later_stmt): Likewise.
5089 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5090 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5091 (vec_info::~vec_info): Free stmt_vec_infos.
5092 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5093 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5094 (pass_slp_vectorize::execute): Likewise.
5095 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5096 (free_stmt_vec_info_vec): Likewise.
5097 (set_stmt_vec_info_vec): New function.
5098 (free_stmt_vec_infos): Likewise.
5099 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5100 the global stmt_vec_info_vec.
5101 * tree-parloops.c (gather_scalar_reductions): Use
5102 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5105 2018-05-29 Richard Biener <rguenther@suse.de>
5107 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5109 2018-05-29 Martin Liska <mliska@suse.cz>
5110 David Malcolm <dmalcolm@redhat.com>
5112 * vec.c (test_reverse): New.
5113 (vec_c_tests): Add new test.
5114 * vec.h (vl_ptr>::reverse): New function.
5116 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5118 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5120 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5123 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5125 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5127 2018-05-28 Richard Biener <rguenther@suse.de>
5129 PR tree-optimization/85933
5130 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5131 look at stmts marked as vectorizable.
5133 2018-05-28 Richard Biener <rguenther@suse.de>
5135 PR tree-optimization/85934
5136 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5137 vector boolean check before scalar optimization.
5139 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5141 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5144 2018-05-28 Mark Wielaard <mark@klomp.org>
5146 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5147 if it is an expression containing a minus sign.
5149 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5151 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5153 2018-05-27 Paul Koning <ni1d@arrl.net>
5155 * config/pdp11/pdp11.md (truncsihi2): Remove.
5157 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5158 Chung-Ju Wu <jasonwucj@gmail.com>
5160 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5162 (unaligned_store_dw): Ditto.
5163 * config/nds32/nds32-memory-manipulation.c
5164 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5165 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5166 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5167 (emit_setmem_word_loop): Rename to ...
5168 (emit_setmem_doubleword_loop): ... this.
5169 (nds32_gen_dup_4_byte_to_word_value): New function.
5170 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5171 (nds32_expand_setmem_loop): Refine implementation.
5172 (nds32_expand_setmem_loop_v3m): Ditto.
5173 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5176 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5178 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5180 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5182 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5183 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5184 machine->attr_no_prologue_p.
5185 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5186 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5187 (nds32_expand_epilogue): Consider attr_naked_p.
5188 (nds32_expand_epilogue_v3pop): Likewise.
5189 (nds32_can_use_return_insn): Likewise.
5190 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5191 attr_no_prologue_p fields.
5192 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5194 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5197 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5199 * config/i386/sse.md
5200 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5202 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5204 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5206 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5208 (*<floatsuffix>floatv2div2sf2): Rename to ...
5209 (*float<floatunssuffix>v2div2sf2): ... this.
5210 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5211 (float<floatunssuffix>v2div2sf2_mask): ... this.
5212 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5213 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5214 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5216 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5218 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5220 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5222 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5224 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5226 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5227 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5228 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5229 gen_ufix_truncv8dfv8si2.
5230 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5231 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5232 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5233 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5234 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5235 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5236 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5237 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5239 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5243 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5245 2018-05-25 Jim Wilson <jimw@sifive.com>
5247 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5248 * config/riscv/riscv.c (struct machine_function): Add
5249 interrupt_handler_p and attribute_checked_p fields.
5250 (riscv_attribute_table): Add interrupt.
5251 (riscv_interrupt_type_p): New.
5252 (riscv_save_reg_p): Save extra regs for interrupt handler.
5253 (riscv_use_save_libcall): Return false for interrupt handler.
5254 (riscv_first_stack_step): Add forward declaration.
5255 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5256 for interrupt handler with large frame. Use it for saved reg list.
5257 (riscv_expand_prologue): Move flag_stack_usage_info support to
5258 eliminate duplication.
5259 (riscv_expand_epilogue): Generate mret for interrupt handler.
5260 (riscv_epilogue_uses): New.
5261 (riscv_can_use_return_insn): Return false for interrupt handler.
5262 (riscv_function_ok_for_sibcall): Likewise.
5263 (riscv_set_current_function): Add interrupt handler support.
5264 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5265 * config/riscv/riscv.md (UNSPECV_MRET): New.
5267 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5269 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5271 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5273 PR tree-optimization/85712
5274 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5275 this candidate has already been replaced in-situ by a copy.
5277 2018-05-25 Jason Merrill <jason@redhat.com>
5279 PR c++/80485 - inline function non-zero address.
5280 * symtab.c (nonzero_address): Check DECL_COMDAT.
5282 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5285 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5286 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5287 extension of SImode operation. Use const123_operand predicate.
5288 (*saddsi_1): Remove.
5289 (*saddl_se_1): Ditto.
5291 (*ssubl_se_1): Ditto.
5292 * config/alpha/predicates.md (const123_operand): New predicate.
5293 * config/alpha/constraints.md (P): Use IN_RANGE.
5295 2018-05-25 Richard Biener <rguenther@suse.de>
5297 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5299 (ref_maybe_used_by_stmt_p): Likewise.
5300 (stmt_may_clobber_ref_p): Likewise.
5301 (stmt_may_clobber_ref_p_1): Likewise.
5302 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5304 (ref_maybe_used_by_stmt_p): Likewise.
5305 (stmt_may_clobber_ref_p): Likewise.
5306 (stmt_may_clobber_ref_p_1): Likewise.
5307 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5308 the alias oracle to disambiguate DRs with stmts DR analysis
5310 (vect_analyze_data_refs): Do not give up on not analyzable
5311 DRs for BB vectorization. Remove code truncating the dataref
5314 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5317 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5318 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5319 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5321 2018-05-25 Richard Biener <rguenther@suse.de>
5323 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5324 function, combining stmt data ref gathering and fatal analysis
5326 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5327 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5328 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5329 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5330 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5332 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5334 PR tree-optimization/85720
5335 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5336 SCC if all partitions are builtins.
5337 (version_loop_by_alias_check): New parameter. Generate cancelable
5338 runtime alias check if all partitions are builtins.
5339 (distribute_loop): Update call to above function.
5341 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5343 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5344 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5345 (parm_default_def_partition_arg): Ditto.
5346 (set_parm_default_def_partition): Ditto.
5347 (get_parm_default_def_partitions): Ditto and make it static.
5348 (get_undefined_value_partitions): Ditto and make it static.
5349 (remove_ssa_form): Refactor call to init_var_map here.
5350 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5351 computation for loop region.
5352 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5353 (register_default_def): Delete.
5354 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5355 (parm_default_def_partition_arg): Ditto.
5356 (set_parm_default_def_partition): Ditto.
5357 (get_parm_default_def_partitions): Ditto and make it static.
5358 (get_undefined_value_partitions): Ditto and make it static.
5359 (coalesce_with_default, coalesce_with_default): Update comment.
5360 (create_coalesce_list_for_region): New func factored out from
5361 create_outofssa_var_map.
5362 (populate_coalesce_list_for_outofssa): New func factored out from
5363 create_outofssa_var_map and coalesce_ssa_name.
5364 (create_outofssa_var_map): Delete.
5365 (coalesce_ssa_name): Refactor to support live range computation.
5366 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5367 (get_parm_default_def_partitions): Delete.
5368 (get_undefined_value_partitions): Ditto.
5369 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5370 computation for loop region.
5371 (new_tree_live_info, loe_visit_block): Ditto.
5372 (live_worklist, set_var_live_on_entry): Ditto.
5373 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5374 * tree-ssa-live.h (struct _var_map): New fields.
5375 (init_var_map): Change decl.
5376 (region_contains_p): New.
5378 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5380 * tree-ssa-live.h (live_merge_and_clear): Delete.
5382 2018-05-25 Richard Biener <rguenther@suse.de>
5385 * tree-dump.c (dequeue_and_dump): Remove access to removed
5386 operand 2 of a SWITCH_EXPR.
5388 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5390 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5391 multiplication and division.
5392 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5393 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5394 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5395 (cond_udiv_optab, cond_umod_optab): New optabs.
5396 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5397 (IFN_COND_RDIV): New internal functions.
5398 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5399 TRUNC_MOD_EXPR and RDIV_EXPR.
5400 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5401 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5403 (SVE_INT_BINARY): Include mult.
5404 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5405 (optab, sve_int_op): Handle mult.
5406 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5408 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5409 for SVE_INT_BINARY_SD.
5411 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5413 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5414 (optab, sve_int_op): Handle div and udiv.
5415 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5416 for SVE_INT_BINARY_SD.
5417 (*<optab><mode>3): New insn for the same.
5419 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5421 * tree-vect-patterns.c: Include predict.h.
5422 (vect_recog_divmod_pattern): Restrict check for division support
5423 to when optimizing for size.
5425 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5427 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5428 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5429 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5430 (gimple_match_op::set_op): Likewise.
5431 (gimple_resimplify4): Declare.
5432 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5433 (expr::gen_transform): Likewise.
5434 (decision_tree::gen): Generate a simplification routine for 4 operands.
5435 * gimple-match-head.c (gimple_simplify): Add an overload for
5436 4 operands. In the top-level function, handle up to 4 call
5437 arguments and call gimple_resimplify4.
5438 (gimple_resimplify4): New function.
5439 (build_call_internal): Pass a fourth operand.
5440 (maybe_push_to_seq): Likewise.
5441 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5442 Fold VEC_COND_EXPRs of an operation and a default value into
5443 an IFN_COND_* function if possible.
5444 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5446 (SVE_COND_FP_BINARY): Include them.
5447 (optab, sve_fp_op): Handle them.
5448 (SVE_INT_BINARY_REV): New code iterator.
5449 (SVE_COND_FP_BINARY_REV): New int iterator.
5450 (commutative): New int attribute.
5451 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5453 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5455 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5456 (*cond_<optab><mode>): New patterns for reversed operands.
5458 2018-05-25 Richard Biener <rguenther@suse.de>
5460 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5461 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5462 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5463 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5464 * tree-vect-data-refs.c (everywhere): Adjust users.
5465 * tree-vect-loop.c (everywhere): Likewise.
5466 * tree-vect-slp.c (everywhere): Likewise.
5467 * tree-vect-stmts.c (everywhere): Likewise.
5468 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5470 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5472 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5474 (gcc_cv_as_section_exclude): ... this.
5475 Try Solaris as #exclude syntax.
5476 * configure: Regenerate.
5477 * config.in: Regenerate.
5478 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5480 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5481 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5483 * varasm.c (default_elf_asm_named_section): Don't check if
5484 HAVE_GAS_SECTION_EXCLUDE is defined.
5486 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5488 * doc/md.texi: Update the documentation of the cond_* optabs
5489 to mention the new final operand. Fix GET_MODE_NUNITS call.
5490 Describe the scalar case too.
5491 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5492 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5494 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5495 (get_conditional_internal_fn): Update comment.
5496 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5497 accumulator value as a final argument to conditional functions.
5498 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5499 a define_expand and add an "else" operand. Assert for now that
5500 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5501 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5502 (*cond_<optab><mode>): New patterns.
5503 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5504 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5505 (UNSPEC_COND_EOR): Delete.
5506 (optab): Remove associated mappings.
5507 (SVE_INT_BINARY): New code iterator.
5508 (sve_int_op): Remove int attribute and add "minus" to the code
5510 (SVE_COND_INT_OP): Delete.
5511 (SVE_COND_FP_OP): Rename to...
5512 (SVE_COND_FP_BINARY): ...this.
5514 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5516 * optabs.c (can_reuse_operands_p): New function.
5517 (maybe_legitimize_operands): Try to reuse the results for
5520 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5522 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5523 Add {q} suffix to insn mnemonic.
5525 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5527 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5528 (msp430_warn_func_return): New.
5530 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5532 * fold-const.c (tree_nonzero_bits): New function.
5533 * fold-const.h (tree_nonzero_bits): Likewise.
5534 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5535 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5537 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5541 * varasm.c (assemble_alias): Check ifunc_resolver only on
5544 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5547 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5548 when memory input operand is handled.
5550 2018-05-24 Luis Machado <luis.machado@linaro.org>
5552 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5554 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5556 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5558 * match.pd: Delay FMA folds until after vectorization.
5560 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5563 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5564 address check not strict.
5566 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5568 * gimple-match.h (gimple_match_op): New class.
5569 (mprts_hook): Replace parameters with a gimple_match_op *.
5570 (maybe_build_generic_op): Likewise.
5571 (gimple_simplified_result_is_gimple_val): Replace parameters with
5572 a const gimple_match_op *.
5573 (gimple_simplify): Replace code_helper * and tree * parameters with
5574 a gimple_match_op * parameter.
5575 (gimple_resimplify1): Replace code_helper *, tree and tree *
5576 parameters with a gimple_match_op * parameter.
5577 (gimple_resimplify2): Likewise.
5578 (gimple_resimplify3): Likewise.
5579 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5580 parameters with a gimple_match_op * parameter.
5581 * gimple-match-head.c (gimple_simplify): Change prototypes of
5582 auto-generated functions to take a gimple_match_op * instead of
5583 separate code_helper * and tree * parameters. Make the same
5584 change in the top-level overload and update calls to the
5585 gimple_resimplify routines. Update calls to the auto-generated
5586 functions and to maybe_push_res_to_seq in the publicly-facing
5587 operation-specific gimple_simplify overloads.
5588 (gimple_match_op::MAX_NUM_OPS): Define.
5589 (gimple_resimplify1): Replace rcode and ops with a single res_op
5590 parameter. Update call to gimple_simplify.
5591 (gimple_resimplify2): Likewise.
5592 (gimple_resimplify3): Likewise.
5593 (mprts_hook): Replace parameters with a gimple_match_op *.
5594 (maybe_build_generic_op): Likewise.
5595 (build_call_internal): Replace type, nargs and ops with
5596 a gimple_match_op *.
5597 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5598 with a single gimple_match_op *. Update calls to mprts_hook,
5599 build_call_internal and gimple_simplified_result_is_gimple_val.
5600 Factor out code that is common to the tree_code and combined_fn cases.
5601 * genmatch.c (expr::gen_transform): Replace tem_code and
5602 tem_ops with a gimple_match_op called tem_op. Update calls
5603 to the gimple_resimplify functions and maybe_push_res_to_seq.
5604 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5605 res_ops. Update call to the gimple_resimplify functions.
5606 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5607 (decision_tree::gen): Make the functions take a gimple_match_op *
5608 called res_op instead of separate res_code and res_ops parameters.
5609 Update call accordingly.
5610 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5611 and ops with a single res_op parameter. Update calls to
5612 maybe_build_generic_op and maybe_push_res_to_seq.
5613 (fold_stmt_1): Update calls to gimple_simplify and
5614 replace_stmt_with_simplification.
5615 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5616 and gimple_simplified_result_is_gimple_val.
5617 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5619 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5620 with a gimple_match_op *.
5621 (vn_nary_build_or_lookup): Likewise. Update call to
5622 vn_nary_build_or_lookup_1.
5623 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5624 gimple_match_op *. Update calls to the gimple_resimplify routines
5625 and to gimple_simplified_result_is_gimple_val.
5626 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5627 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5628 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5629 (visit_nary_op): Likewise.
5630 (visit_reference_op_load): Likewise.
5632 2018-05-23 Luis Machado <luis.machado@linaro.org>
5634 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5635 modifier for printing the step amount.
5637 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5640 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5643 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5645 * doc/sourcebuild.texi (Endianness): New subsubsection.
5647 2018-05-23 Luis Machado <luis.machado@linaro.org>
5649 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5650 <prefetch_dynamic_strides>: New const bool field.
5651 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5652 prefetch_dynamic_strides.
5653 (exynosm1_prefetch_tune): Likewise.
5654 (thunderxt88_prefetch_tune): Likewise.
5655 (thunderx_prefetch_tune): Likewise.
5656 (thunderx2t99_prefetch_tune): Likewise.
5657 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5659 (aarch64_override_options_internal): Update to set
5660 PARAM_PREFETCH_DYNAMIC_STRIDES.
5661 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5662 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5663 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5664 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5665 prefetch-dynamic-strides setting.
5667 2018-05-23 Luis Machado <luis.machado@linaro.org>
5669 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5670 <minimum_stride>: New const int field.
5671 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5672 minimum_stride field defaulting to -1.
5673 (exynosm1_prefetch_tune): Likewise.
5674 (thunderxt88_prefetch_tune): Likewise.
5675 (thunderx_prefetch_tune): Likewise.
5676 (thunderx2t99_prefetch_tune): Likewise.
5677 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5678 <default_opt_level>: Set to 3.
5679 (aarch64_override_options_internal): Update to set
5680 PARAM_PREFETCH_MINIMUM_STRIDE.
5681 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5682 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5683 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5684 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5685 stride is constant and is below the minimum stride threshold.
5687 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5689 * config/arm/arm-cpus.in (mode26): Delete.
5690 (armv4): Delete mode26 reference.
5691 * config/arm/arm.c (arm_configure_build_target): Delete use of
5694 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5696 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5698 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5699 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5701 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5702 (floatunsdidf2): Ditto.
5704 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5706 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5707 (fixuns_trunc<mode>si2_avx512f): Ditto.
5708 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5709 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5710 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5712 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5714 PR rtl-optimization/79985
5715 * df-scan.c (df_insn_refs_collect): Remove special case for
5716 global registers and asm statements.
5718 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5720 * extend.texi (Global Register Variables): Rewrite the bullet list.
5721 Note that the register is available for allocation. Note that access
5722 via inline asm must use constraints. Add note about async-signal
5723 handlers. Remove paragraph about automagic register selection.
5725 2018-05-23 Richard Biener <rguenther@suse.de>
5727 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5728 of fixed offset from memset VN.
5730 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5732 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5734 (alloc_cand_and_find_basis): Initialize first_interp field.
5735 (slsr_process_mul): Modify first_interp field.
5736 (slsr_process_add): Likewise.
5737 (slsr_process_cast): Modify first_interp field for each new
5739 (slsr_process_copy): Likewise.
5740 (dump_candidate): Dump first_interp field.
5741 (replace_mult_candidate): Process all interpretations, not just
5743 (replace_rhs_if_not_dup): Likewise.
5744 (replace_one_candidate): Likewise.
5746 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5748 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5750 (aarch64_needs_frame_chain): New function.
5751 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5753 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5756 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5757 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5758 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5759 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5760 as true for strict-align.
5761 (aarch64_can_inline_p): Perform checks even when callee has no
5762 attributes to check for strict alignment.
5763 * doc/extend.texi (AArch64 Function Attributes): Document
5765 * doc/invoke.texi: (AArch64 Options): Likewise.
5767 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5769 PR tree-optimization/85853
5770 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5771 the handling of the root of the node to...
5772 (vect_slp_analyze_node_operations_1): ...this new function,
5773 and run the whole thing with the child nodes' def types
5774 set according to their SLP node's def type.
5776 2018-05-23 Richard Biener <rguenther@suse.de>
5779 * tree-data-ref.c (create_runtime_alias_checks): Defer
5780 and ignore overflow warnings.
5782 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5784 PR tree-optimization/85822
5785 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5788 2018-05-23 Richard Biener <rguenther@suse.de>
5790 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5791 memset constants via native_interpret_expr.
5793 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5796 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5798 (cgraph_node::create_alias): Likewise.
5799 (cgraph_node::get_availability): Check ifunc_resolver instead
5800 of looking up ifunc attribute.
5801 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5802 * varasm.c (do_assemble_alias): Likewise.
5803 (assemble_alias): Likewise.
5804 (default_binds_local_p_3): Likewise.
5805 * cgraph.h (cgraph_node): Add ifunc_resolver.
5806 (cgraph_node::only_called_directly_or_aliased_p): Return false
5808 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5810 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5811 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5812 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5813 instead of looking up ifunc attribute.
5815 2018-05-22 Luis Machado <luis.machado@linaro.org>
5817 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5819 2018-05-22 Martin Sebor <msebor@redhat.com>
5822 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5823 only when expasion succeeds.
5824 (expand_builtin_strcmp): Same.
5825 (expand_builtin_strncmp): Same.
5827 2018-05-22 Martin Sebor <msebor@redhat.com>
5829 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5831 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5832 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5834 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5835 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5837 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5838 Add check for legitimate_address.
5839 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5840 (aarch64_swap_ldrstr_operands): New.
5841 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5844 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5845 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5847 * config/aarch64/aarch64.md: New patterns to generate stp
5849 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5850 single words and double words.
5851 (load_pair_sw, load_pair_dw): Likewise.
5852 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5854 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5856 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5857 for different mode ldpstp and add peephole for merged zero stores.
5859 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5861 (aarch64_gen_store_pair): Rename calls to match new patterns.
5862 (aarch64_gen_load_pair): Rename calls to match new patterns.
5863 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5864 (load_pair<DREG:mode><DREG2:mode>): ... This.
5865 (store_pair<mode>): Rename to...
5866 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5867 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5869 (V_INT_EQUIV): Handle SImode.
5870 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5873 2018-05-22 Martin Sebor <msebor@redhat.com>
5876 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5877 or ajust the presumed bound on an operation to avoid unnecessary
5880 2018-05-22 Martin Sebor <msebor@redhat.com>
5882 PR tree-optimization/85826
5883 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5884 assuming that a DECL necesarily has a constant size.
5886 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5889 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5891 2018-05-22 Richard Biener <rguenther@suse.de>
5893 PR tree-optimization/85834
5894 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5895 non-constant and non-zero memset arguments.
5897 2018-05-22 Martin Liska <mliska@suse.cz>
5900 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5902 2018-05-22 Richard Biener <rguenther@suse.de>
5904 PR tree-optimization/85863
5905 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5906 comparisons when vectype is specified.
5907 (vectorizable_condition): Do not specify vectype for
5908 vect_is_simple_cond when SLP vectorizing.
5910 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5913 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5914 define __ibm128 as long double.
5915 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5916 as a distinct type when IEEE 128-bit support is enabled.
5917 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5918 128-bit types to use the correct functions.
5919 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5920 convert between 128-bit floating point types that have different
5921 modes but the same representation, instead of using gen_lowpart to
5923 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5925 (IFKF_reg): New attributes to give the register constraints for
5927 (extend<mode>tf2_internal): New insns to mark an explicit
5928 conversion between 128-bit floating point types that have a
5929 different mode but share the same representation.
5931 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5933 PR tree-optimization/85814
5934 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5935 a null return from get_strinfo when unsharing the next
5936 strinfo in the chain.
5938 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5941 * varasm.c (weak_finish): Clean up weak_decls.
5943 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5945 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5946 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5947 UNSPEC_UADALP values.
5948 * config/aarch64/iterators.md (ABAL): New int iterator.
5951 (sur): Add mappings for the above.
5952 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5954 (aarch64_<sur>abal<mode>_4): Likewise.
5955 (aarch64_<sur>adalp<mode>_3): Likewise.
5956 (<sur>sadv16qi): New define_expand.
5958 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5960 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5961 (*movdf_internal): Ditto.
5962 (*rcpsf2_sse): Ditto.
5963 (*rsqrtsf2_sse): Ditto.
5964 (*sqrt<mode>2_sse): Ditto.
5966 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5968 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5970 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5971 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5972 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5973 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5975 * config/aarch64/arm_neon.h: Likewise.
5976 * config/aarch64/iterators.md (VQ_I): New.
5978 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5980 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5982 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5985 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5987 * config/nds32/constraints.md (S): New constraint.
5988 * config/nds32/nds32.md (call_internal): Use constraint S.
5989 (call_value_internal): Likewise.
5990 (sibcall_internal): Likewise.
5991 (sibcall_value_internal): Likewise.
5993 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5994 Chung-Ju Wu <jasonwucj@gmail.com>
5996 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5999 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6000 Chung-Ju Wu <jasonwucj@gmail.com>
6002 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6003 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6004 (nds32_rtx_costs_impl): Simplify.
6005 (nds32_address_cost_impl): Simplify.
6006 (nds32_init_rtx_costs): New function.
6007 (nds32_rtx_costs_speed_prefer): Likewise.
6008 (nds32_rtx_costs_size_prefer): Likewise.
6009 (nds32_address_cost_speed_prefer): Likewise.
6010 (nds32_address_cost_speed_fwprop): Likewise.
6011 (nds32_address_cost_size_prefer): Likewise.
6012 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6013 * config/nds32/nds32.c (nds32_option_override): Use
6014 nds32_init_rtx_costs function.
6016 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6018 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6019 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6020 (TARGET_PIPELINE_N8): Likewise.
6021 (TARGET_PIPELINE_N10): Likewise.
6022 (TARGET_PIPELINE_N13): Likewise.
6023 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6025 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6027 * config/nds32/nds32-fpu.md: Update copyright year.
6029 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6031 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6033 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6035 * config/nds32/nds32.c
6036 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6037 * config/nds32/nds32.opt (minline-asm-r15): New option.
6039 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6041 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6043 * config/nds32/nds32.md (abssi2): New pattern.
6045 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6047 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6048 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6049 Merge insn pattern from sse_cvtsi2ss<round_name> and
6050 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6051 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6052 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6053 using SWI48 mode iterator.
6054 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6055 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6056 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6057 pattern from sse_cvttss2si<round_saeonly_name>
6058 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6059 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6060 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6061 using SWI48 mode iterator.
6062 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6063 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6064 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6065 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6066 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6067 using SWI48 mode iterator.
6068 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6069 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6070 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6071 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6072 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6073 SWI48 mode iterator.
6074 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6075 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6076 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6077 pattern from sse_cvttsd2si<round_saeonly_name>
6078 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6080 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6082 * config/nds32/nds32-md-auxiliary.c
6083 (nds32_valid_smw_lwm_base_p): Refine.
6084 (nds32_output_smw_single_word): Refine.
6085 (nds32_output_smw_double_word): New.
6086 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6088 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6090 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6091 (nds32_output_stack_pop): Refine.
6092 (nds32_expand_unaligned_load): Refine.
6093 (nds32_expand_unaligned_store): Refine.
6095 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6096 Chung-Ju Wu <jasonwucj@gmail.com>
6098 * config/nds32/constants.md: Add TP_REGNUM constant.
6099 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6100 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6102 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6103 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6104 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6105 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6106 and PIC code generation.
6107 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6109 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6111 * config/nds32/nds32.md: Support TLS and PIC.
6112 * config/nds32/nds32.c: Support TLS and PIC.
6113 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6114 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6117 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6119 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6120 mode with E_ prefix.
6122 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6123 Chung-Ju Wu <jasonwucj@gmail.com>
6125 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6126 * config/nds32/nds32-md-auxiliary.c
6127 (symbolic_reference_mentioned_p): New.
6128 (nds32_legitimize_ict_address): New.
6129 (nds32_expand_ict_move): New.
6130 (nds32_indirect_call_referenced_p): New.
6131 (nds32_symbol_binds_local_p): Delete.
6132 (nds32_long_call_p): Modify.
6133 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6134 * config/nds32/nds32-protos.h
6135 (symbolic_reference_mentioned_p): Declare.
6136 (nds32_legitimize_ict_address): Declare.
6137 (nds32_expand_ict_move): Declare.
6138 (nds32_indirect_call_referenced_p): Declare.
6139 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6140 (nds32_relax_group): Use nds32_ict_const_p as condition.
6141 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6142 (nds32_asm_file_start): Output ict_model directive in asm code.
6143 (nds32_legitimate_address_p): Consider indirect call.
6144 (nds32_print_operand): Consider indirect call.
6145 (nds32_print_operand_address): Consider indirect call.
6146 (nds32_insert_attributes): Handle "indirect_call" attribute.
6147 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6148 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6149 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6150 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6151 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6152 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6153 (TARGET_ICT_MODEL_SMALL): Define.
6154 (TARGET_ICT_MODEL_LARGE): Define.
6155 * config/nds32/nds32.md (movsi): Consider ict model.
6156 (call, call_value): Consider ict model.
6157 (sibcall, sibcall_value): Consider ict model.
6158 * config/nds32/nds32.opt (mict-model): New option.
6159 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6162 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6163 Monk Chiang <sh.chiang04@gmail.com>
6164 Jim Wilson <jimw@sifive.com>
6166 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6167 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6168 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6169 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6170 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6171 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6172 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6173 compute save_libcall_adjustment properly.
6174 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6175 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6176 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6177 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6178 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6179 (ABI_SPEC): Handle mabi=ilp32e.
6180 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6181 (RVE): Add RVE mask.
6182 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6183 <-march>: Add rv32e as an example.
6185 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6188 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6189 (intra_create_variable_infos): Handle C++ constructors.
6191 2018-05-18 Martin Liska <mliska@suse.cz>
6193 * passes.def: Remove a redundant pass.
6195 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6198 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6200 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6202 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6204 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6206 (armv2, armv2a, armv3, armv3m): Delete architectures.
6207 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6208 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6209 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6211 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6212 (*mulsidi3adddi): Likewise.
6213 (mulsidi3): Likewise.
6214 (*mulsidi3_nov6): Likewise.
6215 (umulsidi3): Likewise.
6216 (umulsidi3_nov6): Likewise.
6217 (umaddsidi4): Likewise.
6218 (*umulsidi3adddi): Likewise.
6219 (smulsi3_highpart): Likewise.
6220 (*smulsi3_highpart_nov6): Likewise.
6221 (umulsi3_highpart): Likewise.
6222 (*umulsi3_highpart_nov6): Likewise.
6223 * config/arm/arm.h (arm_arch3m): Delete.
6224 * config/arm/arm.c (arm_arch3m): Delete.
6225 (arm_option_override_internal): Update armv3-related comment.
6226 (arm_configure_build_target): Delete use of isa_bit_mode32.
6227 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6228 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6229 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6231 (mulusa3): Likewise.
6232 * config/arm/arm-protos.h (arm_arch3m): Delete.
6233 * config/arm/arm-tables.opt: Regenerate.
6234 * config/arm/arm-tune.md: Likewise.
6235 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6236 deleted architectures.
6238 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6240 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6241 (armv5t, armv5te): New features.
6242 (ARMv5, ARMv5e): Delete fgroups.
6243 (ARMv5t, ARMv5te): Adjust for above changes.
6245 (armv5, armv5e): Delete arches.
6246 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6248 (*call_reg_arm): Likewise.
6249 (*call_value_reg_armv5): Likewise.
6250 (*call_value_reg_arm): Likewise.
6251 (*call_symbol): Likewise.
6252 (*call_value_symbol): Likewise.
6253 (*sibcall_insn): Likewise.
6254 (*sibcall_value_insn): Likewise.
6256 (prefetch): Likewise.
6257 (define_split and define_peephole2 dependent on arm_arch5):
6259 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6261 (TARGET_ARM_QBIT): Likewise.
6262 (TARGET_DSP_MULTIPLY): Likewise.
6263 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6264 (arm_arch5, arm_arch5e): Delete.
6265 (arm_arch5t, arm_arch5te): Declare.
6266 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6267 (arm_arch5t): Declare.
6268 (arm_option_reconfigure_globals): Update for the above.
6269 (arm_options_perform_arch_sanity_checks): Update comment, replace
6270 use of arm_arch5 with arm_arch5t.
6271 (use_return_insn): Likewise.
6272 (arm_emit_call_insn): Likewise.
6273 (output_return_instruction): Likewise.
6274 (arm_final_prescan_insn): Likewise.
6275 (arm_coproc_builtin_available): Likewise.
6276 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6277 arm_arch5e with arm_arch5t and arm_arch5te.
6278 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6279 (arm_arch5t, arm_arch5te): Declare.
6280 * config/arm/arm-tables.opt: Regenerate.
6281 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6282 * config/arm/t-multilib: Likewise.
6283 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6284 instead of arm_arch5.
6285 (*call_reg_thumb1): Likewise.
6286 (*call_value_reg_thumb1_v5): Likewise.
6287 (*call_value_reg_thumb1): Likewise.
6288 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6290 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6292 2018-05-18 Martin Liska <mliska@suse.cz>
6294 PR gcov-profile/84846
6295 * doc/gcov.texi: Document -t option of gcov tool.
6297 2018-05-18 Martin Liska <mliska@suse.cz>
6299 PR gcov-profile/84846
6300 * gcov.c (print_usage): Add new -t option.
6301 (process_args): Handle the option.
6302 (generate_results): Use stdout as output when requested by
6305 2018-05-18 Martin Liska <mliska@suse.cz>
6307 PR gcov-profile/84846
6308 * coverage.c (coverage_init): Write PWD to .gcno file.
6309 * doc/gcov.texi: Document how working directory is printed.
6310 * gcov-dump.c (dump_gcov_file): Print PWD.
6311 * gcov.c (output_intermediate_file): Likewise.
6312 (read_graph_file): Read PWD string.
6313 (output_lines): Print PWD.
6315 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6318 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6319 for retval and return false if all args to phi are zero.
6321 2018-05-18 Richard Biener <rguenther@suse.de>
6323 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6325 (evrp_dom_walker::before_dom_children): Call it.
6327 2018-05-18 Richard Biener <rguenther@suse.de>
6329 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6330 results when processing array refs with variable index.
6332 2018-05-18 Toon Moene <toon@moene.org>
6334 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6335 directly after that of -floop-interchange. Indicate that both
6336 options are enabled by default when specifying -O3.
6338 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6340 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6341 iterator. Delete separate integer-mode vec_set<mode> expander.
6342 (aarch64_simd_vec_setv2di): Delete.
6343 (vec_setv2di): Delete.
6344 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6345 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6346 the "w, r" alternative.
6348 2018-05-18 Martin Liska <mliska@suse.cz>
6350 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6351 * tree-pass.h (make_pass_lower_switch_O0): New function.
6352 * tree-switch-conversion.c (node_has_low_bound): Remove.
6353 (node_has_high_bound): Likewise.
6354 (node_is_bounded): Likewise.
6355 (class pass_lower_switch): Make it a template type and create
6357 (pass_lower_switch::execute): Add template argument.
6358 (make_pass_lower_switch): New function.
6359 (make_pass_lower_switch_O0): New function.
6360 (do_jump_if_equal): Remove.
6361 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6362 all the hard work to tree optimization passes.
6364 2018-05-18 Martin Liska <mliska@suse.cz>
6366 * dbgcnt.c (limit_low): Renamed from limit.
6367 (limit_high): New variable.
6368 (dbg_cnt_is_enabled): Check for upper limit.
6369 (dbg_cnt): Adjust dumping.
6370 (dbg_cnt_set_limit_by_index): Add new argument for high
6372 (dbg_cnt_set_limit_by_name): Likewise.
6373 (dbg_cnt_process_single_pair): Parse new format.
6374 (dbg_cnt_process_opt): Use strtok.
6375 (dbg_cnt_list_all_counters): Remove 'value' and add
6377 * doc/invoke.texi: Document changes.
6379 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6381 * doc/sourcebuild.texi (scalar_all_fma): Document.
6382 * tree.def (FMA_EXPR): Delete.
6383 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6384 * internal-fn.c (ternary_direct): New macro.
6385 (expand_ternary_optab_fn): Likewise.
6386 (direct_ternary_optab_supported_p): Likewise.
6387 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6388 * builtins.c (fold_builtin_fma): Delete.
6389 (fold_builtin_3): Don't call it.
6390 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6391 * expr.c (expand_expr_real_2): Likewise.
6392 * fold-const.c (operand_equal_p): Likewise.
6393 (fold_ternary_loc): Likewise.
6394 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6395 * gimple.c (DEFTREECODE): Likewise.
6396 * gimplify.c (gimplify_expr): Likewise.
6397 * optabs-tree.c (optab_for_tree_code): Likewise.
6398 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6399 * tree-eh.c (operation_could_trap_p): Likewise.
6400 (stmt_could_throw_1_p): Likewise.
6401 * tree-inline.c (estimate_operator_cost): Likewise.
6402 * tree-pretty-print.c (dump_generic_node): Likewise.
6403 (op_code_prio): Likewise.
6404 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6405 * tree-ssa-operands.c (get_expr_operands): Likewise.
6406 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6407 * fold-const-call.h (fold_fma): Delete.
6408 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6409 CFN_FNMA and CFN_FNMS.
6411 * genmatch.c (combined_fn): New enum.
6412 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6413 (commutative_op): New function.
6414 (commutate): Use it. Handle more than 2 operands.
6415 (dt_operand::gen_gimple_expr): Use commutative_op.
6416 (parser::parse_expr): Allow :c to be used with non-binary
6417 operators if the commutative operand is known.
6418 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6419 CFN_FMS, CFN_FNMA and CFN_FNMS.
6420 (backprop::process_assign_use): Remove FMA_EXPR handling.
6421 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6422 (gen_hsa_fma): New function.
6423 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6424 IFN_FNMA and IFN_FNMS.
6425 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6426 * gimple-fold.h (follow_all_ssa_edges): Declare.
6427 * gimple-fold.c (follow_all_ssa_edges): New function.
6428 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6429 gimple_build interface and use follow_all_ssa_edges to fold the result.
6430 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6431 instead of checking for optabs directly.
6432 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6433 rather than FMA_EXPRs.
6434 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6435 call to IFN_FMA instead of an FMA_EXPR.
6437 2018-05-17 Jim Wilson <jimw@sifive.com>
6439 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6440 sign extended promoted subreg, and the range does not have the sign bit
6441 set, then do a sign extend.
6443 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6444 test, check for sign extended subreg and/or constant operands, and
6445 do a sign extend in that case.
6447 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6449 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6450 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6452 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6453 Change logics_shift_reg to logics_shift_imm.
6454 (thunderx2t99_fp_loadpair_basic): Delete.
6455 (thunderx2t99_fp_storepair_basic): Delete.
6456 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6457 (thunderx2t99_asimd_polynomial): Delete.
6458 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6459 and neon_fp_mul_d_scalar_q.
6460 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6461 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6462 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6463 (thunderx2t99_asimd_lut): Add missing tbl types.
6464 (thunderx2t99_asimd_ext): Delete.
6465 (thunderx2t99_asimd_load1_1_mult): Delete.
6466 (thunderx2t99_asimd_load1_2_mult): Delete.
6467 (thunderx2t99_asimd_load1_ldp): New.
6468 (thunderx2t99_asimd_load1): New.
6469 (thunderx2t99_asimd_load2): Add missing *load2* types.
6470 (thunderx2t99_asimd_load3): New.
6471 (thunderx2t99_asimd_load4): New.
6472 (thunderx2t99_asimd_store1_1_mult): Delete.
6473 (thunderx2t99_asimd_store1_2_mult): Delete.
6474 (thunderx2t99_asimd_store2_mult): Delete.
6475 (thunderx2t99_asimd_store2_onelane): Delete.
6476 (thunderx2t99_asimd_store_stp): New.
6477 (thunderx2t99_asimd_store1): New.
6478 (thunderx2t99_asimd_store2): New.
6479 (thunderx2t99_asimd_store3): New.
6480 (thunderx2t99_asimd_store4): New.
6482 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6484 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6485 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6487 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6488 Segher Boessenkool <segher@kernel.crashing.org>
6491 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6494 2018-05-17 Richard Biener <rguenther@suse.de>
6496 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6497 for pruning loop and prune defs feeding only already visited PHIs.
6499 2018-05-17 Richard Biener <rguenther@suse.de>
6501 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6503 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6504 Richard Biener <rguenther@suse.de>
6506 PR tree-optimization/85793
6507 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6508 for VMAT_ELEMENTWISE.
6510 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6512 * internal-fn.h (lookup_internal_fn): Declare
6513 * internal-fn.c (lookup_internal_fn): New function.
6514 * gimple.c (gimple_build_call_from_tree): Handle calls to
6516 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6517 internal function names.
6518 * tree-pretty-print.c (dump_generic_node): Likewise.
6519 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6521 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6523 * gimple-fold.h (gimple_build): Make the function forms take
6524 combined_fn rather than built_in_function.
6525 (gimple_simplify): Likewise.
6526 * gimple-match-head.c (gimple_simplify): Likewise.
6527 * gimple-fold.c (gimple_build): Likewise.
6528 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6529 rather than gimple_build_call_internal.
6530 (get_initial_defs_for_reduction): Likewise.
6531 (vect_create_epilog_for_reduction): Likewise.
6532 (vectorizable_live_operation): Likewise.
6534 2018-05-17 Martin Liska <mliska@suse.cz>
6536 * gimple-ssa-sprintf.c (format_directive): Do not use
6537 space in between 'G_' and '('.
6539 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6542 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6543 even if the mask is not all ones.
6546 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6548 (ix86_gimple_fold_builtin): Likewise.
6551 * config/i386/i386.c: Include tree-vector-builder.h.
6552 (ix86_vector_shift_count): New function.
6553 (ix86_fold_builtin): Fold shift builtins by scalar count.
6554 (ix86_gimple_fold_builtin): Likewise.
6556 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6557 _mm512_setzero): New intrinsics.
6559 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6560 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6562 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6563 code generation for cases where splatting a value is not useful.
6564 * simplify-rtx.c (simplify_ternary_operation): Simplify
6565 vec_merge across a vec_duplicate and a paradoxical subreg forming
6566 a vector mode to a vec_concat.
6568 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6570 * config.gcc: Support "goldmont-plus".
6571 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6573 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6574 PROCESSOR_GOLDMONT_PLUS.
6575 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6576 (processor_target_table): Add "goldmont-plus".
6577 (PTA_GOLDMONT_PLUS): Define.
6578 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6579 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6580 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6581 (fold_builtin_cpu): Add "goldmont-plus".
6582 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6583 (ix86_option_override_internal): Add "goldmont-plus".
6584 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6585 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6586 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6587 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6589 2018-05-17 Richard Biener <rguenther@suse.de>
6591 PR tree-optimization/85757
6592 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6593 remove defs that only feed that PHI from further processing.
6595 2018-05-16 Jim Wilson <jimw@sifive.com>
6597 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6599 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6601 2018-05-16 Mark Wielaard <mark@klomp.org>
6603 * dwarf2out.c (count_index_strings): New function.
6604 (output_indirect_strings): Call count_index_strings and generate
6605 header for dwarf_version >= 5.
6607 2018-05-16 Mark Wielaard <mark@klomp.org>
6609 * dwarf2out.c (dwarf_FORM): New function.
6610 (set_indirect_string): Use dwarf_FORM.
6611 (reset_indirect_string): Likewise.
6612 (size_of_die): Likewise.
6613 (value_format): Likewise.
6614 (output_die): Likewise.
6615 (add_skeleton_AT_string): Likewise.
6616 (output_macinfo_op): Likewise.
6617 (index_string): Likewise.
6618 (output_index_string_offset): Likewise.
6619 (output_index_string): Likewise.
6620 (count_index_strings): Likewise.
6622 2018-05-16 Carl Love <cel@us.ibm.com>
6624 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6625 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6627 2018-05-16 Martin Jambor <mjambor@suse.cz>
6629 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6630 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6632 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6634 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6635 (fnma<mode>4): Likewise.
6636 (fms<mode>4): Likewise.
6637 (fnms<mode>4): Likewise.
6638 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6639 (aarch64_fnma<mode>4): Likewise.
6640 (aarch64_fms<mode>4): Likewise.
6641 (aarch64_fnms<mode>4): Likewise.
6642 (aarch64_fnmadd<mode>4): Likewise.
6644 2018-05-16 Jason Merrill <jason@redhat.com>
6646 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6648 2018-05-16 Richard Biener <rguenther@suse.de>
6650 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6651 (dump_stmt_cost): Declare.
6652 (add_stmt_cost): Dump cost we add.
6653 (add_stmt_costs): New function.
6654 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6656 (vect_analyze_stmt): Adjust prototype.
6657 (vectorizable_condition): Likewise.
6658 (vectorizable_live_operation): Likewise.
6659 (vectorizable_reduction): Likewise.
6660 (vectorizable_induction): Likewise.
6661 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6662 cost vector to pass to vectorizable_ and record afterwards.
6663 (vect_model_reduction_cost): Take cost vector argument and adjust.
6664 (vect_model_induction_cost): Likewise.
6665 (vectorizable_reduction): Likewise.
6666 (vectorizable_induction): Likewise.
6667 (vectorizable_live_operation): Likewise.
6668 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6669 SLP_TREE_NUMBER_OF_VEC_STMTS.
6670 (vect_analyze_slp_cost_1): Remove.
6671 (vect_analyze_slp_cost): Likewise.
6672 (vect_slp_analyze_node_operations): Take visited args and
6673 a target cost vector. Avoid processing already visited stmt sets.
6674 (vect_slp_analyze_operations): Use a local cost vector to gather
6675 costs and register those of non-discarded instances.
6676 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6677 (vect_schedule_slp_instance): Remove copying of
6678 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6680 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6681 adding cost. Record cost entry location.
6682 (vect_prologue_cost_for_slp_op): Function to compute cost of
6683 a constant or invariant generated for SLP vect in the prologue,
6684 split out from vect_analyze_slp_cost_1.
6685 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6686 (vect_model_promotion_demotion_cost): Likewise.
6687 (vect_model_store_cost): Likewise, make static.
6688 (vect_model_load_cost): Likewise.
6689 (vectorizable_bswap): Add cost vector arg and adjust.
6690 (vectorizable_call): Likewise.
6691 (vectorizable_simd_clone_call): Likewise.
6692 (vectorizable_conversion): Likewise.
6693 (vectorizable_assignment): Likewise.
6694 (vectorizable_shift): Likewise.
6695 (vectorizable_operation): Likewise.
6696 (vectorizable_store): Likewise.
6697 (vectorizable_load): Likewise.
6698 (vectorizable_condition): Likewise.
6699 (vectorizable_comparison): Likewise.
6700 (can_vectorize_live_stmts): Likewise.
6701 (vect_analyze_stmt): Likewise.
6702 (vect_transform_stmt): Adjust calls to vectorizable_*.
6703 * tree-vectorizer.c: Include gimple-pretty-print.h.
6704 (dump_stmt_cost): New function.
6706 2018-05-16 Richard Biener <rguenther@suse.de>
6708 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6709 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6710 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6711 (check_name): New callback.
6712 (dse_classify_store): Track cycles via a visited bitmap of PHI
6713 defs and simplify handling of in-loop and across loop dead stores
6714 and properly fail for loop-variant refs. Handle byte-tracking with
6715 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6718 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6720 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6721 (vect_get_mask_type_for_stmt): Likewise.
6722 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6724 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6725 to determine the statement's vector type and the vector type that
6726 should be used for calculating nunits. Deal with cases in which
6727 the type has to be deferred.
6728 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6729 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6730 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6731 (vect_determine_vf_for_stmt): New functions, split out from...
6732 (vect_determine_vectorization_factor): ...here.
6733 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6734 (vect_get_mask_type_for_stmt): New functions, split out from
6735 vect_determine_vectorization_factor.
6737 2018-05-16 Richard Biener <rguenther@suse.de>
6739 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6740 verify the [VEC_]COND_EXPR embedded comparison.
6742 2018-05-15 Martin Sebor <msebor@redhat.com>
6744 PR tree-optimization/85753
6745 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6746 RECORD_TYPE in addition to ARRAY_TYPE.
6748 2018-05-15 Martin Sebor <msebor@redhat.com>
6751 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6753 2018-05-15 Richard Biener <rguenther@suse.de>
6755 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6756 add by_clobber_p one. Change algorithm to collect all defs
6757 representing uses we need to walk and try reducing them to
6758 a single one before failing.
6759 (dse_dom_walker::dse_optimize_stmt): Adjust.
6761 2018-05-13 Mark Wielaard <mark@klomp.org>
6763 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6764 (size_of_loc_descr): Likewise.
6765 (output_loc_operands): Likewise.
6766 (output_loc_operands_raw): Likewise.
6767 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6768 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6769 (hash_loc_operands): Likewise.
6770 (compare_loc_operands): Likewise.
6772 2018-05-14 Mark Wielaard <mark@klomp.org>
6774 * dwarf2out.c (count_index_addrs): New function.
6775 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6777 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6779 PR tree-optimization/83648
6780 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6781 return value as malloc candidate.
6783 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6786 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6787 param as true in call to suggest_attribute.
6789 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6791 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6792 -mreadonly-in-sdata.
6794 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6796 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6798 (aarch64_crypto_aesd_fused): Likewise.
6800 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6802 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6803 (movsi_aarch64): Likewise.
6804 (load_pairsi): Likewise.
6805 (load_pairdi): Likewise.
6806 (store_pairsi): Likewise.
6807 (store_pairdi): Likewise.
6808 (load_pairsf): Likewise.
6809 (load_pairdf): Likewise.
6810 (store_pairsf): Likewise.
6811 (store_pairdf): Likewise.
6812 (zero_extend): Likewise.
6813 (trunc): Swap alternatives.
6814 (fcvt_target): Add '?' to prefer w over r.
6816 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6819 * config/i386/i386.md: Disallow non-commutative arithmetics in
6820 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6821 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6822 in the peephole2 before it.
6824 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6826 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6827 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6828 (ix86_handle_option): Handle -mcldemote.
6829 * config.gcc: New header.
6830 * config/i386/cldemoteintrin.h: New file.
6831 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6832 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6834 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6835 OPTION_MASK_ISA_CLDEMOTE.
6836 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6837 (ix86_valid_target_attribute_inner_p): Ditto.
6838 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6839 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6840 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6841 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6842 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6844 * config/i386/i386.opt: Add -mcldemote.
6845 * config/i386/x86intrin.h: New header.
6846 * doc/invoke.texi: Add -mcldemote.
6848 2018-05-14 Richard Biener <rguenther@suse.de>
6850 * doc/match-and-simplify.texi: Adjust :s documentation.
6852 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6854 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6855 intended memcpy size.
6856 (REORDER_45): Likewise.
6858 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6860 * sort.cc: New file.
6861 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6862 * vec.c (qsort_chk): Use gcc_qsort.
6863 * Makefile.in (OBJS-libcommon): Add sort.o.
6864 (build/sort.o): New target. Use it...
6865 (BUILD_RTL): ... here, and...
6866 (build/gencfn-macros): ... here, and...
6867 (build/genmatch): ... here.
6869 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6870 Chung-Ju Wu <jasonwucj@gmail.com>
6872 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6873 * config/nds32/nds32-graywolf.md: New file.
6874 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6875 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6877 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6878 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6879 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6880 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6881 * config/nds32/pipelines.md: Include n15 settings.
6883 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6884 Chung-Ju Wu <jasonwucj@gmail.com>
6886 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6887 * config/nds32/nds32-n13.md: New file.
6888 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6889 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6891 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6892 * config/nds32/nds32.md (pipeline_model): Add n13.
6893 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6894 * config/nds32/pipelines.md: Include n13 settings.
6896 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6897 Chung-Ju Wu <jasonwucj@gmail.com>
6899 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6900 * config/nds32/nds32-n10.md: New file.
6901 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6902 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6904 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6905 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6906 * config/nds32/nds32.md (pipeline_model): Add n10.
6907 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6908 * config/nds32/pipelines.md: Include n10 settings.
6910 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6911 Kito Cheng <kito.cheng@gmail.com>
6912 Chung-Ju Wu <jasonwucj@gmail.com>
6914 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6915 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6916 Add enum values for DSP extension instructions.
6917 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6919 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6920 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6922 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6923 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6924 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6925 * config/nds32/nds32-intrinsic.md: Likewise.
6926 * config/nds32/nds32_intrinsic.h: Likewise.
6927 * config/nds32/nds32-md-auxiliary.c: Likewise.
6928 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6929 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6930 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6931 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6932 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6933 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6934 TYPE_DMAC in switch statement.
6935 * config/nds32/nds32.c: New checking and implementation for DSP
6936 extension instructions.
6937 * config/nds32/nds32.h: Likewise.
6938 * config/nds32/nds32.md: Likewise.
6939 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6940 * config/nds32/predicates.md: Implement new predicates for DSP
6943 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6945 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6946 Reformat alternatives and attributes so it is easier to identify
6947 which constraints/attributes go with which instruction.
6948 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6949 (mov<mode>_softfloat32, FMOVE64): Likewise.
6950 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6951 (mov<mode>_softfloat64, FMOVE64): Likewise.
6953 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6955 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6957 (Basic PowerPC Built-in Functions): The new name of the
6958 subsection previously known as "PowerPC Built-in Functions".
6959 (Basic PowerPC Built-in Functions Available on all Configurations):
6961 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6962 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6963 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6964 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6966 2018-05-11 Martin Jambor <mjambor@suse.cz>
6969 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6972 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6975 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6977 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6979 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6980 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6981 (ix86_handle_option): Handle -mwaitpkg.
6982 * config.gcc: New header.
6983 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6984 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6985 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6987 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6988 OPTION_MASK_ISA_WAITPKG.
6989 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6990 (ix86_option_override_internal): Add PTA_WAITPKG.
6991 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6992 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6993 IX86_BUILTIN_TPAUSE.
6994 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6995 __builtin_ia32_umwait and __builtin_ia32_tpause.
6996 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6997 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6998 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6999 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7000 UNSPECV_TPAUSE): New.
7001 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7002 * config/i386/i386.opt: Add -mwaitpkg.
7003 * config/i386/waitpkgintrin.h: New file.
7004 * config/i386/x86intrin.h: New header.
7005 * doc/invoke.texi: Add -mwaitpkg.
7007 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7010 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7012 (cortex-m0): Use armv6s-m isa.
7013 (cortex-m0plus): Likewise.
7014 (cortex-m1): Likewise.
7015 (cortex-m0.small-multiply): Likewise.
7016 (cortex-m0plus.small-multiply): Likewise.
7017 (cortex-m1.small-multiply): Likewise.
7019 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7020 Jakub Jelinek <jakub@redhat.com>
7022 PR tree-optimization/85692
7023 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7024 source permute as well.
7026 2018-05-11 Martin Liska <mliska@suse.cz>
7029 * doc/extend.texi: Document LLVM style format for no_sanitize
7032 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7034 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7035 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7036 (mode_supports_vsx_dform_quad): Likewise.
7037 (mode_supports_vmx_dform): Move these functions to be next to the
7038 other mode_supports functions.
7039 (mode_supports_dq_form): Likewise.
7040 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7041 mode_supports_dq_form.
7042 (reg_offset_addressing_ok_p): Likewise.
7043 (offsettable_ok_by_alignment): Likewise.
7044 (rs6000_legitimate_offset_address_p): Likewise.
7045 (legitimate_lo_sum_address_p): Likewise.
7046 (rs6000_legitimize_address): Likewise.
7047 (rs6000_legitimize_reload_address): Likewise.
7048 (rs6000_secondary_reload_inner): Likewise.
7049 (rs6000_preferred_reload_class): Likewise.
7050 (rs6000_output_move_128bit): Likewise.
7052 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7054 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7055 Generate SImode target register for null target.
7056 <case IX86_BUILTIN_XGETBV>: Ditto.
7057 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7058 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7060 2018-05-10 Carl Love <cel@us.ibm.com>
7062 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7063 dcbtt and dcbtstt if operands[2] is 0.
7065 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7068 * config/i386/sse.md (usadv64qi): New expander.
7070 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7072 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7073 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7074 -maltivec=be support.
7075 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7076 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7077 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7078 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7079 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7080 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7081 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7082 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7083 altivec_vsumsws): Adjust.
7084 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7085 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7086 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7088 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7089 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7090 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7091 (altivec_lve<VI_char>x): Delete expand.
7092 (*altivec_lve<VI_char>x_internal): Rename to...
7093 (altivec_lve<VI_char>x): ... this.
7094 (altivec_lvxl_<mode>): Delete expand.
7095 (*altivec_lvxl_<mode>_internal): Rename to ...
7096 (altivec_lvxl_<mode>): ... this.
7097 (altivec_stvxl_<mode>): Delete expand.
7098 (*altivec_stvxl_<mode>_internal): Rename to ...
7099 (altivec_stvxl_<mode>): ... this.
7100 (altivec_stve<VI_char>x): Delete expand.
7101 (*altivec_stve<VI_char>x_internal): Rename to ...
7102 (altivec_stve<VI_char>x): ... this.
7103 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7104 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7105 reduc_plus_scal_<mode>): Adjust.
7106 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7108 (rs6000_cpu_cpp_builtins): Adjust.
7109 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7110 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7111 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7112 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7113 -maltivec=be support.
7114 (rs6000_split_vec_extract_var): Adjust.
7115 (rs6000_split_v4si_init): Adjust.
7116 (swap_selector_for_mode): Delete.
7117 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7118 altivec_expand_stvex_be): Delete.
7119 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7120 -maltivec=be support.
7121 (rs6000_gimple_fold_builtin): Ditto.
7122 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7124 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7125 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7126 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7127 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7128 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7129 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7130 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7131 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7132 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7133 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7134 anonymous split): Adjust.
7135 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7136 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7138 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7140 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7141 when --with-gxx-include-dir is also specified.
7142 * configure: Regenerate.
7144 2018-05-09 Jim Wilson <jimw@sifive.com>
7147 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7148 * config/riscv/t-withmultilib: New.
7149 * config/riscv/withmultilib.h: New.
7150 * doc/install.texi: Document RISC-V --with-multilib-list support.
7152 2018-05-09 Richard Biener <rguenther@suse.de>
7154 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7156 (vect_bb_vectorization_profitable_p): Adjust. Compute
7157 actual scalar cost using the cost vector and the add_stmt_cost
7160 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7162 PR rtl-optimization/85645
7163 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7164 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7166 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7168 PR rtl-optimization/85645
7169 * shrink-wrap.c (spread_components): Return a boolean saying if
7170 anything was changed.
7171 (try_shrink_wrapping_separate): Iterate spread_components until
7172 nothing changes anymore.
7174 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7176 PR rtl-optimization/85645
7177 * regrename.c (build_def_use): Also kill the chains that include the
7178 destination of a REG_CFA_REGISTER note.
7180 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7182 PR rtl-optimization/85645
7183 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7184 insn that has a REG_CFA_REGISTER note.
7186 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7188 * cfgexpand.c (expand_clobber): New function.
7189 (expand_gimple_stmt_1): Use it.
7190 * tree-vect-stmts.c (vect_clobber_variable): New function,
7192 (vectorizable_simd_clone_call): ...here.
7193 (vectorizable_store): Emit a clobber either side of an
7194 IFN_STORE_LANES sequence.
7195 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7197 2018-05-09 Tom de Vries <tom@codesourcery.com>
7200 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7201 (define_insn "trap_if_false"): Add exit after trap.
7203 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7205 PR rtl-optimization/85638
7206 * bb-reorder.c: Include common/common-target.h.
7207 (create_forwarder_block): New function extracted from...
7208 (fix_up_crossing_landing_pad): ...here. Rename into...
7209 (dw2_fix_up_crossing_landing_pad): ...this.
7210 (sjlj_fix_up_crossing_landing_pad): New function.
7211 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7212 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7213 from both partitions and exit the loop after one iteration.
7215 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7218 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7220 (Basic PowerPC Built-in Functions): The new name of the
7221 subsection previously known as "PowerPC Built-in Functions".
7222 (Basic PowerPC Built-in Functions Available on all Configurations):
7224 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7226 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7228 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7230 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7233 2018-05-08 Jim Wilson <jimw@sifive.com>
7235 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7236 (LD_EMUL_SUFFIX): New.
7237 (LINK_SPEC): Use it.
7239 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7241 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7243 (Basic PowerPC Built-in Functions): The new name of the
7244 subsection previously known as "PowerPC Built-in Functions".
7245 (Basic PowerPC Built-in Functions Available on all Configurations):
7247 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7249 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7251 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7253 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7256 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7259 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7260 after cmpelim optimization.
7262 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7264 * config.gcc: Support "goldmont".
7265 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7266 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7268 * config/i386/i386.c (m_GOLDMONT): Define.
7269 (processor_target_table): Add "goldmont".
7270 (PTA_GOLDMONT): Define.
7271 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7272 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7273 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7274 (fold_builtin_cpu): Add "goldmont".
7275 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7276 (ix86_option_override_internal): Add "goldmont".
7277 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7278 (processor_type): Add PROCESSOR_GOLDMONT.
7279 * config/i386/i386.md: Add CPU "glm".
7280 * config/i386/glm.md: New file.
7281 * config/i386/x86-tune.def: Add m_GOLDMONT.
7282 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7284 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7287 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7289 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7290 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7291 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7294 * config/i386/i386.c (ix86_fold_builtin): Handle
7295 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7298 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7299 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7301 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7304 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7305 (check_arch): Likewise.
7306 (check_fpu): Return the result rather than printing it.
7307 (end arch): Fix operator precedence.
7308 (end cpu): Likewise.
7309 (END): Print the result from check_fpu.
7311 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7312 Alan Hayward <alan.hayward@arm.com>
7313 David Sherwood <david.sherwood@arm.com>
7315 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7316 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7317 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7318 (*fcmuo<mode>_and): New patterns.
7320 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7322 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7323 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7324 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7325 (cmp_op, sve_imm_con): New code attributes.
7326 (SVE_COND_INT_CMP, imm_con): Delete.
7327 (cmp_op): Remove above unspecs from int attribute.
7328 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7330 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7331 comparison-specific unspecs.
7332 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7333 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7334 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7335 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7336 (*vec_fcm<cmp_op><mode>): Rename to...
7337 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7338 (*vec_fcmuo<mode>): Rename to...
7339 (*fcmuo<mode>): ...this and adjust likewise.
7340 (*pred_fcm<cmp_op><mode>): New pattern.
7341 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7342 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7344 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7346 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7347 (aarch64_emit_sve_predicated_cond): New function.
7348 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7349 (aarch64_emit_unspec_cond_or): Replace with...
7350 (aarch64_emit_sve_or_conds): ...this new function. Use
7351 aarch64_emit_sve_ptrue_op for the individual comparisons and
7352 aarch64_emit_binop to OR them together.
7353 (aarch64_emit_inverted_unspec_cond): Replace with...
7354 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7355 aarch64_emit_sve_ptrue_op for the comparison and
7356 aarch64_emit_unop to invert the result.
7357 (aarch64_expand_sve_vec_cmp_float): Update after the above
7358 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7360 2018-05-07 Nathan Sidwell <nathan@acm.org>
7362 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7363 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7364 (Backwards Compatibility): Likewise.
7366 2018-05-07 Luis Machado <luis.machado@linaro.org>
7370 2018-05-07 Luis Machado <luis.machado@linaro.org>
7372 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7373 <prefetch_dynamic_strides>: New const bool field.
7374 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7375 prefetch_dynamic_strides.
7376 (exynosm1_prefetch_tune): Likewise.
7377 (thunderxt88_prefetch_tune): Likewise.
7378 (thunderx_prefetch_tune): Likewise.
7379 (thunderx2t99_prefetch_tune): Likewise.
7380 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7382 (aarch64_override_options_internal): Update to set
7383 PARAM_PREFETCH_DYNAMIC_STRIDES.
7384 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7385 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7386 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7387 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7388 prefetch-dynamic-strides setting.
7390 2018-05-07 Luis Machado <luis.machado@linaro.org>
7392 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7393 <minimum_stride>: New const int field.
7394 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7395 minimum_stride field.
7396 (exynosm1_prefetch_tune): Likewise.
7397 (thunderxt88_prefetch_tune): Likewise.
7398 (thunderx_prefetch_tune): Likewise.
7399 (thunderx2t99_prefetch_tune): Likewise.
7400 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7401 (aarch64_override_options_internal): Update to set
7402 PARAM_PREFETCH_MINIMUM_STRIDE.
7403 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7404 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7405 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7406 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7407 stride is constant and is below the minimum stride threshold.
7409 2018-05-07 Luis Machado <luis.machado@linaro.org>
7411 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7414 2018-05-07 Luis Machado <luis.machado@linaro.org>
7416 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7417 <prefetch_dynamic_strides>: New const bool field.
7418 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7419 prefetch_dynamic_strides.
7420 (exynosm1_prefetch_tune): Likewise.
7421 (thunderxt88_prefetch_tune): Likewise.
7422 (thunderx_prefetch_tune): Likewise.
7423 (thunderx2t99_prefetch_tune): Likewise.
7424 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7426 (aarch64_override_options_internal): Update to set
7427 PARAM_PREFETCH_DYNAMIC_STRIDES.
7428 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7429 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7430 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7431 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7432 prefetch-dynamic-strides setting.
7434 2018-05-07 Luis Machado <luis.machado@linaro.org>
7436 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7437 <minimum_stride>: New const int field.
7438 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7439 minimum_stride field.
7440 (exynosm1_prefetch_tune): Likewise.
7441 (thunderxt88_prefetch_tune): Likewise.
7442 (thunderx_prefetch_tune): Likewise.
7443 (thunderx2t99_prefetch_tune): Likewise.
7444 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7445 (aarch64_override_options_internal): Update to set
7446 PARAM_PREFETCH_MINIMUM_STRIDE.
7447 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7448 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7449 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7450 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7451 stride is constant and is below the minimum stride threshold.
7453 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7456 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7457 the type is addressable. Don't force op into register if it has
7460 2018-05-05 Roland McGrath <mcgrathr@google.com>
7463 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7464 any section for which we don't know a specific type it should have,
7465 regardless of name. Previously this was done only for the exact
7466 names ".init_array", ".fini_array", and ".preinit_array".
7467 (default_elf_asm_named_section): Add comment about
7468 relationship with default_section_type_flags and SECTION_NOTYPE.
7469 (get_section): Don't consider it a type conflict if one side has
7470 SECTION_NOTYPE and the other doesn't, as long as neither has the
7471 SECTION_BSS et al used in the default_section_type_flags logic.
7473 2018-05-05 Tom de Vries <tom@codesourcery.com>
7476 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7477 (workaround_barsyncs): New function.
7478 (nvptx_reorg): Use workaround_barsyncs.
7479 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7480 (define_expand "nvptx_membar_cta"): New define_expand.
7481 (define_insn "*nvptx_membar_cta"): New insn.
7483 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7485 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7486 To improve optimization opportunities.
7487 * builtin-types.def: The new needed builtin types for the above.
7489 2018-05-04 Richard Biener <rguenther@suse.de>
7491 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7492 * gimple-ssa-store-merging.c
7493 (imm_store_chain_info::output_merged_store): Remove redundant create,
7494 release split_store vector contents on failure.
7495 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7496 scalar stmt vector on cache hit.
7498 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7500 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7502 * config.gcc (powerpc-xilinx-eabi*): Remove.
7503 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7505 (fusion_addis_mem_combo_load): Ditto.
7506 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7508 (rs6000_cpu_cpp_builtins): Ditto.
7509 * config/rs6000/rs6000-linux.c
7510 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7511 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7512 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7514 (rs6000_setup_reg_addr_masks): Ditto.
7515 (rs6000_init_hard_regno_mode_ok): Ditto.
7516 (rs6000_option_override_internal): Ditto.
7517 (legitimate_lo_sum_address_p): Ditto.
7518 (rs6000_legitimize_address): Ditto.
7519 (rs6000_legitimize_reload_address): Ditto.
7520 (rs6000_legitimate_address_p): Ditto.
7521 (abi_v4_pass_in_fpr): Ditto.
7522 (setup_incoming_varargs): Ditto.
7523 (rs6000_gimplify_va_arg): Ditto.
7524 (rs6000_split_multireg_move): Ditto.
7525 (rs6000_savres_strategy): Ditto.
7526 (rs6000_emit_prologue_components): Ditto.
7527 (rs6000_emit_epilogue_components): Ditto.
7528 (rs6000_emit_prologue): Ditto.
7529 (rs6000_emit_epilogue): Ditto.
7530 (rs6000_elf_file_end): Ditto.
7531 (rs6000_function_value): Ditto.
7532 (rs6000_libcall_value): Ditto.
7533 * config/rs6000/rs6000.h: Ditto.
7534 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7535 (TARGET_MINMAX): ... this. New.
7536 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7537 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7538 (*movsi_internal1_single): Delete.
7539 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7540 mfpu=, mxilinx-fpu): Delete.
7541 * config/rs6000/singlefp.h: Delete.
7542 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7543 * config/rs6000/t-rs6000: Ditto.
7544 * config/rs6000/t-xilinx: Delete.
7545 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7546 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7547 (VStype_simple): Delete.
7548 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7549 * config/rs6000/xfpu.h: Delete.
7550 * config/rs6000/xfpu.md: Delete.
7551 * config/rs6000/xilinx.h: Delete.
7552 * config/rs6000/xilinx.opt: Delete.
7553 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7554 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7556 2018-05-04 Tom de Vries <tom@codesourcery.com>
7559 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7562 2018-05-04 Richard Biener <rguenther@suse.de>
7565 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7566 (expand_complex_div_wide): Likewise.
7567 (expand_complex_operations_1): Likewise.
7568 (expand_complex_libcall): Preserve EH info of the original stmt.
7569 (tree_lower_complex): Handle removed blocks.
7570 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7571 on complex multiplication and division libcall builtins.
7573 2018-05-04 Richard Biener <rguenther@suse.de>
7576 * fold-const.c (negate_expr_p): Restrict negation of operand
7577 zero of a division to when we know that can happen without
7579 (fold_negate_expr_1): Likewise.
7581 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7584 * real.h (real_nextafter): Declare.
7585 * real.c (real_nextafter): New function.
7586 * fold-const-call.c (fold_const_nextafter): New function.
7587 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7588 CASE_CFN_NEXTTOWARD.
7589 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7590 even when arg1_mode is different from arg0_mode.
7592 2018-05-03 Nathan Sidwell <nathan@acm.org>
7594 * doc/extend.texi (Deprecated Features): Remove
7596 (Backwards Compatibility): Likewise.
7597 * doc/invoke.texi (C++ Language Options): Likewise.
7598 (C++ Dialect Options): Likewise.
7600 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7603 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7604 _mm512_mask_mullox_epi64): New intrinsics.
7606 2018-05-03 Tom de Vries <tom@codesourcery.com>
7609 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7610 dump files): Add offload-tree.
7612 2018-05-03 Richard Biener <rguenther@suse.de>
7614 PR tree-optimization/85615
7615 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7616 to loops not nested in BBs loop father to avoid creating multi-entry
7619 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7621 PR tree-optimization/70291
7622 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7623 arguments. Change return type to tree. Emit libcall as a new
7624 statement rather than replacing existing one when inplace_p is true.
7625 (expand_complex_multiplication_components): New function.
7626 (expand_complex_multiplication): Expand floating-point complex
7627 multiplication using the above.
7628 (expand_complex_division): Rename inner_type parameter to type.
7629 Update expand_complex_libcall call-site.
7630 (expand_complex_operations_1): Update expand_complex_multiplication
7631 and expand_complex_division call-sites.
7633 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7636 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7637 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7638 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7639 the highest significant bit of the shift count mask is clear. In
7640 check whether and[sq]i3 is needed verify that all significant bits
7641 of the shift count other than the highest are set.
7643 2018-05-02 Tom de Vries <tom@codesourcery.com>
7646 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7647 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7648 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7649 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7650 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7651 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7652 __builtin_goacc_parlevel_size.
7654 2018-05-02 Richard Biener <rguenther@suse.de>
7656 PR tree-optimization/85597
7657 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7658 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7661 2018-05-02 Tom de Vries <tom@codesourcery.com>
7664 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7665 dump files): Add ltrans-tree.
7667 2018-05-02 Tom de Vries <tom@codesourcery.com>
7670 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7671 dump files): Add wpa-ipa.
7673 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7675 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7676 powerpc*-*-linux*paired* target.
7677 * config/rs6000/750cl.h: Delete.
7678 * config/rs6000/paired.h: Delete.
7679 * config/rs6000/paired.md: Delete.
7680 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7682 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7683 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7684 comment. Remove paired float support.
7685 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7686 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7688 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7689 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7691 * config/rs6000/rs6000.c: Remove paired float support.
7692 (paired_expand_vector_init, paired_expand_vector_move,
7693 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7694 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7695 paired_expand_builtin, paired_expand_predicate_builtin,
7696 paired_init_builtins): Delete.
7697 * config/rs6000/rs6000.h: Remove paired float support.
7698 * config/rs6000/rs6000.md: Remove paired float support.
7699 (move_from_CR_ov_bit): Delete.
7700 * config/rs6000/rs6000.opt (mpaired): Delete.
7701 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7702 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7704 2018-05-02 Richard Biener <rguenther@suse.de>
7707 * gimplify.c (gimplify_save_expr): When in SSA form allow
7708 SAVE_EXPRs to compute to SSA vars.
7710 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7713 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7714 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7715 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7716 clobber operands[2], instead use a new pseudo. Formatting fixes.
7718 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7720 PR tree-optimization/85586
7721 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7722 exit early for statements in the same group if the accesses are
7725 2018-05-02 Tom de Vries <tom@codesourcery.com>
7728 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7731 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7733 PR tree-optimization/85143
7734 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7736 2018-05-01 Tom de Vries <tom@codesourcery.com>
7739 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7740 not found" error message.
7742 2018-05-01 Tom de Vries <tom@codesourcery.com>
7745 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7746 * vec.c (test_ordered_remove_if): New function.
7747 (vec_c_tests): Call test_ordered_remove_if.
7748 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7749 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7750 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7751 VEC_ORDERED_REMOVE_IF.
7753 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7755 PR tree-optimization/82665
7756 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7757 pointer subtraction where arguments come from a memchr call.
7759 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7761 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7762 --push-state --as-needed and --pop-state instead of --as-needed and
7763 --no-as-needed if ld supports it.
7764 * configure: Regenerated.
7767 * doc/install.texi2html: Replace _002d with - and _002a with * in
7768 generated html files using sed.
7770 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7773 * gcc-rich-location.c (blank_line_before_p): New function.
7774 (use_new_line): New function.
7775 (gcc_rich_location::add_fixit_insert_formatted): New function.
7776 * gcc-rich-location.h
7777 (gcc_rich_location::add_fixit_insert_formatted): New function.
7779 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7781 * selftest.c (assert_streq): Rename "expected" and "actual" to
7782 "val1" and "val2". Extend NULL-handling to cover both inputs
7783 symmetrically, while still requiring both to be non-NULL for a pass.
7784 * selftest.h (assert_streq): Rename "expected" and "actual" to
7786 (ASSERT_EQ): Likewise.
7787 (ASSERT_EQ_AT): Likewise.
7788 (ASSERT_KNOWN_EQ): Likewise.
7789 (ASSERT_KNOWN_EQ_AT): Likewise.
7790 (ASSERT_NE): Likewise.
7791 (ASSERT_MAYBE_NE): Likewise.
7792 (ASSERT_MAYBE_NE_AT): Likewise.
7793 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7794 the assertion to pass.
7795 (ASSERT_STREQ_AT): Likewise.
7797 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7799 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7800 interaction with -pie.
7802 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7804 * selftest.h: Fix alphabetization of per-source-file selftest
7807 2018-04-30 Jason Merrill <jason@redhat.com>
7809 PR c++/61982 - dead stores to destroyed objects.
7810 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7813 2018-04-30 Jason Merrill <jason@redhat.com>
7815 * tree.c (build_clobber): New.
7816 * tree.h: Declare it.
7817 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7819 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7821 * diagnostic-show-locus.c (layout::layout): Update for
7822 location_get_source_line returning a char_span.
7823 (struct char_span): Move to input.h.
7824 (struct correction): Update for fields in char_span becoming
7826 (struct source_line): Update for location_get_source_line
7827 returning a char_span.
7828 (layout::print_line): Likewise.
7829 * edit-context.c (edited_file::print_content): Likewise.
7830 (edited_file::print_diff_hunk): Likewise.
7831 (edited_file::print_run_of_changed_lines): Likewise.
7832 (edited_file::get_num_lines): Likewise.
7833 (edited_line::edited_line): Likewise.
7834 * final.c (asm_show_source): Likewise.
7835 * input.c (location_get_source_line): Convert return type
7836 from const char * to char_span, losing the final "line_len"
7838 (dump_location_info): Update for the above.
7839 (get_substring_ranges_for_loc): Likewise. Use a char_span
7840 when handling the literal within the line.
7841 (test_reading_source_line): Update for location_get_source_line
7842 returning a char_span.
7843 * input.h (class char_span): Move here from
7844 diagnostic-show-locus.c, converting from a struct to a class.
7845 Make data members private.
7846 (char_span::operator bool): New.
7847 (char_span::length): New.
7848 (char_span::get_buffer): New.
7849 (char_span::operator[]): New.
7850 (char_span::subspan): Make const.
7851 (char_span::xstrdup): New.
7852 (location_get_source_line): Convert return type from const char *
7853 to char_span, losing the final "line_size" param.
7855 2018-04-30 Jan Hubicka <jh@suse.cz>
7857 * lto-wrapper.c (ltrans_priorities): New static var.
7858 (cmp_priority): New.
7859 (run_gcc): Read priorities and if doing parallel build order
7860 the Makefile by them.
7862 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7864 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7866 2018-04-30 Richard Biener <rguenther@suse.de>
7868 * tree-cfg.c (verify_address): Remove base argument, add
7869 flag whether to check TREE_ADDRESSABLE and do that.
7870 (verify_expr): Remove.
7871 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7872 (verify_gimple_assign_single): Likewise.
7873 (verify_gimple_switch): Likewise.
7874 (verify_expr_location_1): Dereference tp once. Add (disabled)
7875 piece from verify_expr.
7876 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7878 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7880 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7882 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7884 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7885 (small_data_pattern): Likewise.
7886 (arc_rewrite_small_data): Likewise.
7887 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7888 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7889 (get_symbol_alignment): New function.
7890 (legitimate_small_data_address_p): Likewise.
7891 (legitimate_scaled_address): Update, call
7892 legitimate_small_data_address_p.
7893 (output_sdata): New static variable.
7894 (arc_print_operand): Update how we handle small data operands.
7895 (arc_print_operand_address): Likewise.
7896 (arc_legitimate_address_p): Update, use
7897 legitimate_small_data_address_p.
7898 (arc_rewrite_small_data_p): Remove.
7899 (arc_rewrite_small_data_1): Likewise.
7900 (arc_rewrite_small_data): Likewise.
7901 (small_data_pattern): Likewise.
7902 (compact_sda_memory_operand): Update to use
7903 legitimate_small_data_address_p and get_symbol_alignment.
7904 (prepare_move_operands): Don't rewite sdata pattern.
7905 (prepare_extend_operands): Remove.
7906 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7908 (zero_extendqisi2): Likewise.
7909 (zero_extendhisi2): Likewise.
7910 (extendqihi2): Likewise.
7911 (extendqisi2): Likewise.
7912 (extendhisi2): Likewise.
7916 * config/arc/constraints.md (Usd): Change it to memory constraint.
7918 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7920 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7921 as source of std instructions.
7922 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7923 allow 6-bit constants as source for store instructions.
7924 (movdi_insn): Update instruction pattern to allow 6-bit constants
7925 as source for store instructions.
7927 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7929 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7931 2018-04-30 Nathan Sidwell <nathan@acm.org>
7932 Sandra Loosemore <sandra@codesourcery.com>
7934 * dumpfile.c (dump_open): Allow '-' for stdout.
7935 * doc/invoke.texi (Developer Options): Document dump filename
7936 determination early. Document stdin/stdout selection.
7938 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7940 Microblaze Target: PIC data text relative
7942 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7943 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7945 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7946 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7947 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7948 New addressing mode for data-text relative position indepenedent code.
7949 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7950 'ADDRESS_SYMBOLIC_TXT_REL'.
7951 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7952 (microblaze_legitimate_pic_operand): Exclude function calls from
7953 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7954 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7956 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7957 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7958 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7959 for 'address + offset'.
7960 (microblaze_expand_prologue): Add new function prologue call for
7962 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7963 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7964 table in case of TARGET_PIC_DATA_TEXT_REL.
7965 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7966 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7967 Add new macros 'UNSPEC_TEXT',
7968 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7969 + exclude function calls from 'UNSPEC_PLT' in case of data text
7971 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7972 new target hook for generating address diff vector tables in case of
7974 * doc/tm.texi : Regenerate.
7975 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7976 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7977 of addr diff vector generation.
7978 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7979 target hook definition.
7980 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7981 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7982 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7983 option for data text relative.
7985 2018-04-30 Richard Biener <rguenther@suse.de>
7987 * tree-chrec.h (evolution_function_is_constant_p): Remove
7989 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7991 2018-04-30 Richard Biener <rguenther@suse.de>
7994 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7996 2018-04-30 Richard Biener <rguenther@suse.de>
7998 PR tree-optimization/28364
7999 PR tree-optimization/85275
8000 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8001 copying first exit test.
8003 2018-04-28 Mark Wielaard <mark@klomp.org>
8005 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8007 (dwarf_AT): Handle DW_AT_addr_base.
8008 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8010 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8013 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8014 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8015 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8016 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8018 2018-04-28 Richard Biener <rguenther@suse.de>
8020 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8021 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8022 to reflect use. Only add interesting stmts.
8024 2018-04-27 Martin Jambor <mjambor@suse.cz>
8027 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8028 the jump function allows for passing through aggregate values.
8030 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8032 * input.h (in_system_header_at): Convert from macro to inline
8034 (from_macro_expansion_at): Likewise.
8035 (from_macro_definition_at): Likewise.
8037 2018-04-27 Jeff Law <law@redhat.com>
8039 * config.gcc: Mark tile* targets as deprecated/obsolete.
8041 2018-04-27 Richard Biener <rguenther@suse.de>
8043 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8046 2018-04-27 Richard Biener <rguenther@suse.de>
8048 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8050 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8052 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8053 with Yd constraint. Set "preferred_for_speed" attribute from
8054 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8056 (*movdi_internal): Ditto.
8057 (movti_interunit splitters): Remove
8058 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8059 (movdi_interunit splitters): Ditto.
8060 * config/i386/constraints.md (Ye): Remove.
8061 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8063 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8066 * config/aarch64/constraints.md (Usg): Limit to 31.
8069 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8071 PR tree-optimization/85529
8072 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8073 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8074 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8075 zero extension or masking of the MSB bit.
8076 (optimize_range_tests): Add FIRST_BB argument, pass it through
8077 to optimize_range_tests_var_bound.
8078 (maybe_optimize_range_tests, reassociate_bb): Adjust
8079 optimize_range_tests callers.
8081 2018-04-26 Richard Biener <rguenther@suse.de>
8082 Jakub Jelinek <jakub@redhat.com>
8084 * cgraph.h (symbol_table): Just declare debug method here.
8085 * symtab.c (symbol_table::debug): Define.
8087 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8089 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8091 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8093 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8094 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8095 (*movdi_internal): Substitute Yi and Yj constraint with x
8096 and Ym and Yn constraint with y constraint. Update "isa"
8097 attribute and set "preferred_for_speed" attribute from
8098 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8099 (*movsi_internal): Ditto.
8100 (*movdf_internal): Ditto.
8101 (*movsf_internal): Ditto.
8102 (*zero_extendsidi2): Ditto.
8103 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8104 (sse2_loadld): Ditto.
8105 (*vec_extract<ssevecmodelower>_0): Ditto.
8106 (*vec_extractv4si_0_zext_sse4): Ditto.
8107 (vec_concatv2di): Ditto.
8108 (*vec_dup<mode>): Ditto.
8109 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8110 * config/i386/constraints.md (Yi): Remove.
8115 2018-04-26 Nathan Sidwell <nathan@acm.org>
8117 * dumpfile.c (dump_open): New.
8118 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8119 (dump_finish): Detect stdio/stderr by value not name.
8121 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8123 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8125 2018-04-26 Tom de Vries <tom@codesourcery.com>
8128 * config/nvptx/nvptx.c (verify_neutering_jumps)
8129 (verify_neutering_labels): New function
8130 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8132 2018-04-26 Tom de Vries <tom@codesourcery.com>
8135 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8136 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8139 2018-04-26 Richard Biener <rguenther@suse.de>
8140 Tom de Vries <tom@codesourcery.com>
8143 * lto-streamer-out.c (output_function): Fixup loops if required to match
8144 discovery done in the reader.
8146 2018-04-26 Richard Biener <rguenther@suse.de>
8148 PR tree-optimization/85116
8149 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8150 have a loop exit from the single latch predecessor. Remove
8151 case of header with just condition.
8152 (ch_base::copy_headers): Exclude infinite loops from any
8154 (pass_ch::execute): Record exits.
8156 2018-04-26 Richard Biener <rguenther@suse.de>
8158 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8159 prologue cost vector and pass it to vect_get_load_cost.
8160 (vect_get_peeling_costs_all_drs): Likewise.
8161 (vect_peeling_hash_get_lowest_cost): Likewise.
8162 (vect_enhance_data_refs_alignment): Likewise.
8164 2018-04-26 Richard Biener <rguenther@suse.de>
8167 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8168 checking of integer<->pointer conversions.
8169 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8170 sign-/zero-extending pointer types.
8171 (expand_omp_for_static_chunk): Likewise.
8173 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8174 Jean Lee <xiaoyur347@gmail.com>
8176 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8177 (TARGET_ASAN_SHADOW_OFFSET): Define.
8178 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8179 true for -fsanitize=address.
8181 2018-04-25 Mark Wielaard <mark@klomp.org>
8183 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8186 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8188 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8189 than "alu", remove explicit "memory" and "imm_disp" attributes.
8190 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8193 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8194 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8195 gen_lowpart_no_emit.
8197 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8200 * config/i386/i386.c (ix86_expand_builtin): Change memory
8201 operand to XI, extend p0 to Pmode.
8202 * config/i386/i386.md: Change unspec volatile and operand
8203 1 mode to XI, change operand 0 mode to P.
8205 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8207 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8208 GET_MODE_MASK before any checking.
8209 (nds32_can_use_bset_p): Likewise.
8210 (nds32_can_use_btgl_p): Likewise.
8212 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8214 * config/nds32/nds32-doubleword.md: New define_split pattern for
8215 illegal register number.
8217 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8219 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8221 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8223 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8225 2018-04-25 Richard Biener <rguenther@suse.de>
8227 * lto-streamer.h (LTO_major_version): Bump to 8.
8229 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8231 * BASE-VER: Set to 9.0.0.
8233 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8235 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8236 in __abskf2 and __powikf2.
8238 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8241 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8242 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8243 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8244 Use the above on operand 2. Reindent.
8245 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8247 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8250 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8252 * config/i386/i386.opt (mcet): Removed.
8253 * doc/install.texi: Remove -mcet documentation.
8254 * doc/invoke.texi: Likewise.
8256 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8259 * doc/install.texi: Remove -mcet from bootstrap-cet.
8261 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8264 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8265 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8269 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8270 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8271 containing a CONST_VECTOR.
8273 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8275 * doc/install.texi: Update newlib dependency for nvptx.
8277 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8280 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8281 instead of INTVAL when shifting x left.
8283 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8285 PR tree-optimization/85478
8286 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8287 vect_grouped_store_supported for single element vectors.
8289 2018-04-24 Richard Biener <rguenther@suse.de>
8292 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8293 load cost increase to the case of non-constant step.
8295 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8298 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8299 destination if any_malformed_asm.
8301 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8304 * expr.c (store_field): In the bitfield case, if the value comes from
8305 a function call and is returned in registers by means of a PARALLEL,
8306 do not change the mode of the temporary unless BLKmode and VOIDmode.
8308 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8310 PR rtl-optimization/85423
8311 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8312 dependencies to debug insns when the previous insn is non-debug.
8314 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8316 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8317 enums into a single definition.
8318 (fls): Fix predicates and printing.
8321 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8323 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8324 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8325 and short u6 immediate.
8326 (check_if_valid_sleep_operand): Remove.
8327 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8329 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8331 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8332 flag_always_save_lp condition.
8333 * config/nds32/nds32.opt (malways-save-lp): New option.
8335 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8337 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8338 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8339 * config/nds32/nds32.h
8340 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8341 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8343 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8345 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8346 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8348 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8349 Chung-Ju Wu <jasonwucj@gmail.com>
8351 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8353 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8354 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8356 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8358 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8360 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8362 * config/nds32/nds32-protos.h (nds32_data_alignment,
8363 nds32_local_alignment): Declare.
8364 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8365 nds32_local_alignment): New functions.
8366 (TARGET_CONSTANT_ALIGNMENT): Define.
8367 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8369 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8371 * config/nds32/nds32.c
8372 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8373 (TARGET_MODES_TIEABLE_P): Likewise.
8375 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8377 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8380 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8381 Chung-Ju Wu <jasonwucj@gmail.com>
8383 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8384 for unaligned access.
8385 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8387 * config/nds32/nds32-intrinsic.md: Likewise.
8388 * config/nds32/nds32_intrinsic.h: Likewise.
8389 * config/nds32/nds32.h (nds32_builtins): Likewise.
8390 * config/nds32/nds32.opt (munaligned-access): New option.
8391 * config/nds32/nds32.c (nds32_asm_file_start): Display
8392 flag_unaligned_access status.
8394 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8396 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8397 -mno-relax is present.
8398 * config/riscv/linux.h (LINK_SPEC): Ditto.
8400 2018-04-20 Martin Sebor <msebor@redhat.com>
8403 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8405 (gimple_fold_builtin_stxcpy_chk): Same.
8406 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8408 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8411 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8412 __powikf2 when long double is IEEE 128-bit.
8414 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8416 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8417 step to make sure stack always aligned.
8419 2018-04-20 Carl Love <cel@us.ibm.com>
8422 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8423 size check for arg0.
8425 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8426 Tom de Vries <tom@codesourcery.com>
8429 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8430 Emit insns for calls too.
8431 (nvptx_find_par): Always look for worker-level predecessor insn.
8432 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8434 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8435 (nvptx_process_pars): Propagate frames for calls.
8437 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8440 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8442 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8443 (ix86_handle_option): Don't handle OPT_mibt.
8444 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8446 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8448 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8449 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8450 (ix86_target_macros): Define __CET__ with flag_cf_protection
8451 for -fcf-protection.
8452 * config/i386/i386.c (isa2_opts): Remove -mibt.
8453 * config/i386/i386.h (TARGET_IBT): Removed.
8454 (TARGET_IBT_P): Likewise.
8455 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8456 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8457 * config/i386/i386.opt (mcet): Update help message.
8460 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8461 -mcet as an alias for -mshstk.
8463 2018-04-20 Richard Biener <rguenther@suse.de>
8466 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8467 complexity by forcing a single use of the multiply operand.
8469 2018-04-20 Martin Jambor <mjambor@suse.cz>
8472 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8473 recursion dependency to only apply to non-clones.
8475 2018-04-20 Martin Jambor <mjambor@suse.cz>
8478 * ipa-cp.c (create_specialized_node): Check that clones of
8479 self-recursive edges exist during IPA-CP.
8481 2018-04-19 Toon Moene <toon@moene.org>
8483 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8486 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8488 PR tree-optimization/85467
8489 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8490 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8491 VECTOR_CST element to type.
8493 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8496 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8497 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8498 (builtin_longjmp): Likewise.
8499 (save_stack_nonlocal): New pattern.
8500 (restore_stack_nonlocal): Likewise.
8502 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8505 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8506 Replace ASM_OUTPUT_LABEL with fprintf.
8508 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8511 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8512 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8513 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8514 define __IBT__ and __SHSTK__ for -fcf-protection.
8515 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8517 (ix86_trampoline_init): Likewise.
8518 (x86_output_mi_thunk): Likewise.
8519 (ix86_notrack_prefixed_insn_p): Likewise.
8520 (ix86_option_override_internal): Don't disallow -fcf-protection.
8521 * config/i386/i386.md (rdssp<mode>): Also enable for
8523 (incssp<mode>): Likewise.
8524 (nop_endbr): Likewise.
8525 * config/i386/i386.opt (mcet): Change help message to built-in
8529 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8530 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8531 enable CET built-in functions.
8533 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8535 * common/config/i386/i386-common.c
8536 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8537 OPTION_MASK_ISA_MOVDIRI_UNSET,
8538 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8539 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8540 * config.gcc (movdirintrin.h): New header.
8541 * config/i386/cpuid.h (bit_MOVDIRI,
8542 bit_MOVDIR64B): New bits.
8543 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8545 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8546 (VOID, PVOID, PCVOID)): New function types.
8547 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8548 __builtin_ia32_directstoreu_u64,
8549 __builtin_ia32_movdir64b): New builtins.
8550 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8551 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8553 (ix86_valid_target_attribute_inner_p): Ditto.
8554 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8555 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8556 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8557 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8558 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8559 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8560 (movdiri<mode>, movdir64b_<mode>): New.
8561 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8562 * config/i386/immintrin.h: Include movdirintrin.h.
8563 * config/i386/movdirintrin.h: New file.
8564 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8566 2018-04-19 Richard Biener <rguenther@suse.de>
8569 * cfg.c (clear_bb_flags): When loop state says we have
8570 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8572 2018-04-19 Richard Biener <rguenther@suse.de>
8574 PR tree-optimization/84737
8575 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8576 copying restrict info.
8577 (vect_setup_realignment): Use it.
8578 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8579 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8580 the first DR to all generated stores.
8581 (vectorizable_load): Likewise for loads.
8583 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8585 PR tree-optimization/85446
8586 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8587 the integral and pointer types to have the same precision.
8589 * doc/install.texi: Document --disable-cet being the default and
8592 2018-04-18 Martin Liska <mliska@suse.cz>
8594 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8597 2018-04-18 Martin Liska <mliska@suse.cz>
8600 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8603 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8604 arguments if they are comparable.
8606 2018-04-18 Martin Liska <mliska@suse.cz>
8609 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8612 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8615 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8618 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8619 ENDBR after calling __morestack.
8621 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8624 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8625 by using gcc_base_ver to generate a gcc_driver_version, and use
8626 it when generating GCC_DRIVER_NAME.
8627 * configure: Regenerate.
8629 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8632 * config.gcc: Obsolete powerpc*-*-*spe*.
8634 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8637 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8638 (stabstr_D): Change type of unum from unsigned int to
8639 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8642 2018-04-17 Jim Wilson <jimw@sifive.com>
8645 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8646 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8647 Set arg_pointer_offset after using pretend_args_size.
8649 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8651 PR rtl-optimization/85431
8652 * dse.c (record_store): Ignore zero width stores.
8655 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8656 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8657 __builtin_stack_restore rather than after it.
8658 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8659 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8660 argument instead of virtual_dynamic_stack_rtx.
8662 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8664 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8666 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8667 Add note to error message to explain internal mapping of overloaded
8668 built-in function name to non-overloaded built-in function name.
8669 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8672 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8675 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8676 where the inputs overlap with the output.
8678 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8680 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8681 (=v, v) alternative and explicit "memory" attribute.
8682 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8683 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8685 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8686 "sselog1" type instead of "sselog".
8687 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8688 "sselog". Remove explicit "memory" attribute.
8689 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8690 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8692 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8693 "sselog1" type instead of "sselog". Remove explicit "memory"
8695 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8696 use "sselog1" type instead of "sselog". Remove explicit "memory"
8698 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8699 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8701 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8702 "sselog1" type instead of "sselog". Remove explicit "memory"
8704 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8705 use "sselog1" type instead of "sselog". Remove explicit "memory"
8709 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8712 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8715 2018-04-17 Martin Jambor <mjambor@suse.cz>
8718 * ipa-cp.c (create_specialized_node): Call
8719 expand_all_artificial_thunks if necessary.
8721 2018-04-17 Martin Liska <mliska@suse.cz>
8724 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8725 in message, remote space in between '_G' and '('.
8727 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8730 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8731 avx512f_vmcmp<mode>3<round_saeonly_name>,
8732 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8733 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8734 avx512f_rndscale<mode><round_saeonly_name>,
8735 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8736 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8737 Use %<iptr>2 instead of %2 for -masm=intel.
8738 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8739 avx512f_vcvttss2usi<round_saeonly_name>,
8740 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8742 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8743 avx512f_vcvttsd2usi<round_saeonly_name>,
8744 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8745 Use %q1 instead of %1 for -masm=intel.
8746 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8747 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8748 of %3 for -masm=intel.
8749 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8751 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8753 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8755 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8756 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8758 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8760 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8761 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8762 %g1 and one with %0 and %1.
8763 (avx512er_vmrcp28<mode><round_saeonly_name>,
8764 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8766 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8767 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8768 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8769 of %0 and %{%4%} for -masm=intel.
8770 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8771 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8772 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8773 order of %0 and %{%5%}%{z%} for -masm=intel.
8775 2018-04-17 Jan Hubicka <jh@suse.cz>
8778 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8780 2018-04-17 Martin Liska <mliska@suse.cz>
8783 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8784 for target_clone error message. Make default implementation
8785 clone to be a local declaration.
8786 (separate_attrs): Add new argument and check for an empty
8788 (expand_target_clones): Handle it.
8789 (ipa_target_clone): Make redirection just for target_clones
8792 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8793 Tom de Vries <tom@codesourcery.com>
8796 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8797 tiled basic blocks without omp continue statements.
8799 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8802 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8803 vec_extract expression as having side effects to make sure it gets
8806 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8809 * config/i386/i386.c (get_builtin_code_for_version): Check
8812 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8815 * gcc/config.gcc: Support "skylake".
8816 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8818 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8819 (processor_target_table): Add "skylake".
8820 (ix86_option_override_internal): Add "skylake".
8821 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8822 PROCESSOR_CANNONLAKE.
8823 (get_builtin_code_for_version): Fix priority for
8824 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8825 PROCESSOR_SKYLAKE-AVX512.
8826 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8827 (processor_type): Add PROCESSOR_SKYLAKE.
8829 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8830 Jason Merrill <jason@redhat.com>
8833 * convert.c (convert_to_integer_1): Use direct recursion for
8834 enumeral types and types with a precision less than the number
8835 of bits in their mode.
8837 2018-04-16 Julia Koval <julia.koval@intel.com>
8840 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8841 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8843 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8846 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8847 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8848 and -mno-direct-move.
8850 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8853 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8854 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8855 Ensure negative shifts result in {0}.
8857 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8859 PR rtl-optimization/79916
8860 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8861 regs (if any) to define how to gnerate SD moves when LRA is in
8864 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8866 PR rtl-optimization/85393
8867 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8868 * except.c (expand_dw2_landing_pad_for_region): Make static.
8869 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8870 a label and unconditional jump to old_bb, rather than
8871 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8874 PR rtl-optimization/85376
8875 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8876 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8877 instead of a specific value.
8879 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8880 Bin Cheng <bin.cheng@arm.com>
8882 PR tree-optimization/82965
8883 PR tree-optimization/83991
8884 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8885 by_profile_only parameter.
8886 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8887 information if the loop was predicted to iterate too many times.
8888 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8890 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8893 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8896 2018-04-13 Martin Liska <mliska@suse.cz>
8897 Jakub Jelinek <jakub@redhat.com>
8900 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8901 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8902 * builtins.c (expand_builtin_memory_copy_args): Use
8903 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8904 handle dest_addr == pc_rtx.
8906 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8909 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8910 asked to not generate direct moves.
8911 (fix_trunc<mode>si2_stfiwx): Similar.
8912 (fix_trunc<mode>si2_internal): Similar.
8914 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8917 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8918 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8919 lookup if dest in some wider mode is known to be const0_rtx and
8920 if so, record permanent equivalence for it to be ZERO_EXTEND of
8921 the narrower mode destination.
8923 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8925 * lto-streamer-out.c (output_function): Revert 259346.
8926 * omp-expand.c (expand_oacc_for): Likewise.
8928 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8930 PR rtl-optimization/85354
8931 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8932 * sel-sched.c (sel_global_init): ... here.
8934 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8937 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8938 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8939 mode for PE-COFF targets.
8940 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8941 (i386_pe_asm_lto_end): Likewise.
8942 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8943 (TARGET_ASM_LTO_END): Likewise.
8944 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8945 (i386_pe_asm_lto_start): New function.
8946 (i386_pe_asm_lto_end): Likewise.
8948 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8949 Richard Biener <rguenther@suse.de>
8952 * lto-streamer-out.c (output_function): Fix CFG loop state before
8954 * omp-expand.c (expand_oacc_for): Handle calls to internal
8955 functions like regular functions.
8957 2018-04-12 Richard Biener <rguenther@suse.de>
8960 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8961 for the early LTO debug to properly generate references to it
8962 during DIE emission. Do not re-use that for the skeleton for
8964 (dwarf2out_early_finish): Likewise.
8966 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8969 * config/i386/sse.md
8970 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8971 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8972 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8973 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8974 and output is a reg, avoid creating invalid lowpart subreg, but
8975 instead split into a 512-bit move. Don't split if not AVX512VL,
8976 input is xmm16+ reg and output is a mem.
8977 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8978 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8979 xmm16+ reg and output is a mem.
8981 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8983 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8984 also for flag_dwarf2_cfi_asm.
8986 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8988 PR rtl-optimization/85342
8989 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8990 a bool scalar var inside of the loop instead. Don't try to update
8991 recog_data.operand after failed apply_change_group.
8993 2018-04-12 Tom de Vries <tom@codesourcery.com>
8996 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8997 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8998 array with flexible array member as array without given dimension.
8999 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9000 argument for undefined param to true.
9002 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9005 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9006 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9007 from PowerPC section.
9008 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9009 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9010 help text that is too long.
9011 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9012 help text that is too long.
9013 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9014 help text that is too long.
9016 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9018 * config/alpha/alpha.md (stack_probe_internal): Rename
9019 from "probe_stack". Update all callers.
9021 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9023 PR rtl-optimization/84566
9024 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9025 sched_macro_fuse_insns.
9027 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9030 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9031 (compute_block_dependences): ... from here.
9033 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9035 PR tree-optimization/85331
9036 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9037 from int to HOST_WIDE_INT.
9039 2018-04-11 Martin Jambor <mjambor@suse.cz>
9042 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9043 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9044 not the same as the source val.
9045 (cgraph_edge_brings_value_p): New parameter.
9046 (gather_edges_for_value): Pass destination value to
9047 cgraph_edge_brings_value_p.
9048 (perhaps_add_new_callers): Likewise.
9049 (get_info_about_necessary_edges): Likewise and exclude values brought
9050 only by self-recursive edges.
9051 (create_specialized_node): Redirect only clones of self-calling edges.
9052 (+self_recursive_pass_through_p): New function.
9053 (find_more_scalar_values_for_callers_subset): Use it.
9054 (find_aggregate_values_for_callers_subset): Likewise.
9055 (known_aggs_to_agg_replacement_list): Removed.
9056 (decide_whether_version_node): Re-calculate known constants for all
9057 remaining context clones.
9059 2018-04-11 Richard Biener <rguenther@suse.de>
9062 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9063 from early DWARF output.
9064 (dwarf2out_early_finish): Output line info unconditionally into
9065 early DWARF and add reference to it.
9067 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9070 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9071 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9072 other than V2DFmode using iptr mode attribute.
9073 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9075 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9077 PR rtl-optimization/84659
9078 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9080 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9083 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9085 (output_loc_list): Pass address of a dummy size variable even in the
9086 locview handling loop.
9087 (index_location_lists): Add comment on why skip_loc_list_entry can't
9090 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9093 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9096 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9099 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9100 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9101 and -mstring-compare-inline-limit.
9103 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9106 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9107 for stack clash protection in a register whenever we need it to be in
9110 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9112 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9113 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9115 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9118 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9120 (mlong-double-): Ditto.
9121 * config/rs6000/sysv4.opt (msdata=): Ditto.
9122 (mtls-size=): Ditto.
9124 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9126 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9127 erroneous entries for
9128 "vector int vec_ldl (int, long int *)", and
9129 "vector unsigned int vec_ldl (int, unsigned long int *)".
9130 Add comments and entries for
9131 "vector bool char vec_ldl (int, bool char *)",
9132 "vector bool short vec_ldl (int, bool short *)",
9133 "vector bool int vec_ldl (int, bool int *)",
9134 "vector bool long long vec_ldl (int, bool long long *)",
9135 "vector pixel vec_ldl (int, pixel *)",
9136 "vector long long vec_ldl (int, long long *)",
9137 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9138 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9139 type tree bool_long_long_type_node and correct definition of
9140 bool_V2DI_type_node to make reference to this new type tree.
9141 (rs6000_mangle_type): Replace erroneous reference to
9142 bool_long_type_node with bool_long_long_type_node.
9143 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9144 comments to emphasize sign distinctions for char and int types and
9145 replace RS6000_BTI_bool_long constant with
9146 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9147 use of RS6000_BTI_pixel.
9148 (bool_long_type_node): Remove this macro definition.
9149 (bool_long_long_type_node): New macro definition
9151 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9153 PR rtl-optimization/85300
9154 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9155 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9156 simplify_unary_operation fails.
9158 2018-04-10 Martin Liska <mliska@suse.cz>
9160 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9161 cgraph_edge and ipa_ref.
9163 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9167 * config/i386/sse.md
9168 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9169 computation of the VEC_MERGE selector from mask.
9170 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9171 Fix decoding of the VEC_MERGE selector into mask.
9173 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9175 PR tree-optimization/85286
9176 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9178 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9180 * final.c (final_1): Set insn_last_address as well as
9181 insn_current_address.
9183 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9186 * explow.c (emit_stack_probe): Call validize_mem on memory location
9187 before passing it to gen_probe_stack. Create address operand and
9188 legitimize it for the probe_stack_address case.
9190 2018-04-09 Jan Hubicka <jh@suse.cz>
9193 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9194 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9195 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9197 (free_lang_data): Rebuild type inheritance graph.
9199 2018-04-09 Martin Sebor <msebor@redhat.com>
9201 * invoke.texi (-finline-small-functions): Mention other optimization
9203 (-findirect-inlining, -fpartial-inlining): Same.
9204 (-finline-functions-called-once): Same.
9205 (-freorder-blocks-and-partition): Same.
9207 2018-04-09 Jan Hubicka <jh@suse.cz>
9210 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9211 jumps; choose last target that matches the criteria (i.e.
9212 no partition changes for non-crossing jumps).
9213 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9214 support for redirecting crossing jumps to non-crossing.
9216 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9218 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9219 also for naked functions.
9221 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9223 * config/arc/arc.md (add_shift): New pattern.
9224 (add_shift2): Likewise.
9225 (sub_shift): Likewise.
9226 (sub_shift_cmp0_noout): Likewise.
9227 (compare_si_ashiftsi): Likewise.
9228 (xbfu_cmp0_noout): New combine pattern.
9229 (xbfu_cmp0"): Likewise.
9230 (movsi_set_cc_insn): Place the predicable variant first.
9231 (commutative_binary_cmp0_noout): Remove clobber.
9232 (commutative_binary_cmp0): New pattern.
9233 (noncommutative_binary_cmp0): Likewise.
9234 (noncommutative_binary_cmp0_noout): Likewise.
9235 (noncommutative_binary_comparison_result_used): Removed.
9236 (rsub_cmp0): New pattern.
9237 (rsub_cmp0_noout): Likewise.
9238 (extzvsi): Changed, keep only meaningful variants.
9239 (SQH, SEZ): New iterators.
9240 (SQH_postfix): New mode attribute.
9241 (SEZ_prefix): New code attribute.
9242 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9243 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9244 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9246 (noncommutative_operator): Check the availability of barrel
9249 2018-04-09 Richard Biener <rguenther@suse.de>
9251 PR tree-optimization/85284
9252 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9253 Only use the niter constraining form of simple_iv when the exit
9256 2018-04-09 Tom de Vries <tom@codesourcery.com>
9259 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9260 (define_expand "*memory_barrier"): New define_expand.
9261 (define_insn "memory_barrier"): New insn.
9263 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9265 PR rtl-optimization/80463
9266 PR rtl-optimization/83972
9267 PR rtl-optimization/83480
9269 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9270 correct producer for the insn.
9271 (tidy_control_flow): Fixup seqnos in case of debug insns.
9273 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9275 PR rtl-optimization/83913
9277 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9278 different sched-times when merging exprs.
9280 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9282 PR rtl-optimization/83962
9284 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9285 tidy_fallthru_edge and tidy_control_flow.
9287 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9289 PR rtl-optimization/83530
9291 * sel-sched.c (force_next_insn): New global variable.
9292 (remove_insn_for_debug): When force_next_insn is true, also leave only
9293 next insn in the ready list.
9294 (sel_sched_region): When the region wasn't scheduled, make another pass
9295 over it with force_next_insn set to 1.
9297 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9299 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9301 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9302 for interrupt control.
9303 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9304 functions for interrupt control.
9305 * config/nds32/nds32-intrinsic.md: Likewise.
9306 * config/nds32/nds32_intrinsic.h: Likewise.
9307 * config/nds32/nds32.h (nds32_builtins): Likewise.
9309 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9311 * config/nds32/nds32.c (nds32_init_machine_status,
9312 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9313 strict_aligned_p field.
9314 (nds32_expand_to_rtl_hook): New function.
9315 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9316 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9318 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9319 Chung-Ju Wu <jasonwucj@gmail.com>
9321 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9322 * config/nds32/nds32-n7.md: New file.
9323 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9324 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9326 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9327 * config/nds32/nds32.md (pipeline_model): Add n7.
9328 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9329 * config/nds32/pipelines.md: Include n7 settings.
9331 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9332 Chung-Ju Wu <jasonwucj@gmail.com>
9334 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9335 * config/nds32/nds32-e8.md: New file.
9336 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9337 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9339 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9340 * config/nds32/nds32.md (pipeline_model): Add e8.
9341 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9342 * config/nds32/pipelines.md: Include e8 settings.
9344 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9345 Chung-Ju Wu <jasonwucj@gmail.com>
9347 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9348 * config/nds32/nds32-n8.md: New file.
9349 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9350 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9352 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9353 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9354 * config/nds32/nds32.md (pipeline_model): Add n8.
9355 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9356 * config/nds32/pipelines.md: Include n8 settings.
9358 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9359 Chung-Ju Wu <jasonwucj@gmail.com>
9361 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9362 * config/nds32/nds32-n9-2r1w.md: New file.
9363 * config/nds32/nds32-n9-3r2w.md: New file.
9364 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9365 nds32_register_ports): New or modify for cpu n9.
9366 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9368 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9369 * config/nds32/nds32-utils.c: New file.
9370 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9371 TARGET_MUL_SLOW): Define.
9372 * config/nds32/nds32.md (pipeline_model): New attribute.
9373 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9374 New options that support cpu n9.
9375 * config/nds32/pipelines.md: Include n9 settings.
9376 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9378 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9380 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9381 information if necessary.
9382 (output_cond_branch_compare_zero): Likewise.
9383 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9384 (nds32_target_alignment): Refine for alignment.
9385 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9386 (FUNCTION_BOUNDARY): Modify.
9387 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9389 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9391 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9393 * config/nds32/constants.md (unspec_volatile_element): Add values for
9394 TLB operation and data prefetch.
9395 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9396 functions for TLB operation and data prefetch.
9397 * config/nds32/nds32-intrinsic.md: Likewise.
9398 * config/nds32/nds32_intrinsic.h: Likewise.
9399 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9400 (nds32_print_operand): Likewise.
9401 * config/nds32/nds32.h (nds32_builtins): Likewise.
9403 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9404 Andrew Pinski <pinsika@gcc.gnu.org>
9407 * match.pd: Use constant_boolean_node of correct type instead of
9408 boolean_true_node or boolean_false_node for simplifying
9409 pointer comparisons to zero.
9411 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9413 PR tree-optimization/80021
9414 * tree.c (verify_type_variant): Make error call in verify_variant_match
9415 translatable and remove final full stop.
9417 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9419 * config/nds32/constants.md (unspec_volatile_element): Add
9420 UNSPEC_VOLATILE_EH_RETURN.
9421 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9422 nds32_output_stack_pop): Support dwarf exception handling process.
9423 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9424 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9425 exception handling process.
9426 (nds32_compute_stack_frame): Likewise.
9427 (nds32_return_addr_rtx): Likewise.
9428 (nds32_initial_elimination_offset): Likewise.
9429 (nds32_expand_prologue): Likewise.
9430 (nds32_expand_epilogue): Likewise.
9431 (nds32_dynamic_chain_address): New function.
9432 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9434 (DYNAMIC_CHAIN_ADDRESS): Define.
9435 (EH_RETURN_DATA_REGNO): Define.
9436 (EH_RETURN_STACKADJ_RTX): Define.
9437 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9438 patterns for dwarf exception handling.
9440 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9442 * config/nds32/nds32.h: Clean up obsolete macros.
9444 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9446 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9447 Add enum values for particular instructions.
9448 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9449 particular intrinsic functions.
9450 * config/nds32/nds32-intrinsic.md: Likewise.
9451 * config/nds32/nds32_intrinsic.h: Likewise.
9452 * config/nds32/nds32.h (nds32_builtins): Likewise.
9453 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9454 (btst, ave): New patterns for particular instructions.
9456 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9458 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9459 Add enum values for atomic load/store and memory sync.
9460 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9462 * config/nds32/nds32-intrinsic.md: Likewise.
9463 * config/nds32/nds32_intrinsic.h: Likewise.
9464 * config/nds32/nds32.h (nds32_builtins): Likewise.
9466 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9468 PR tree-optimization/85257
9469 * fold-const.c (native_encode_vector): If not all elts could fit
9470 and off is -1, return 0 rather than offset.
9471 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9472 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9473 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9474 adjust buffer in native_interpret_expr call.
9476 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9478 * config/nds32/constants.md (unspec_volatile_element): Add cache
9479 control enum values.
9480 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9481 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9482 * config/nds32/nds32.c (nds32_cctl_names): New.
9483 (nds32_print_operand): Handle cache control register names.
9484 * config/nds32/nds32.h (nds32_builtins): New enum values.
9485 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9487 * config/nds32/nds32.md (type): Add mmu.
9488 * config/nds32/pipelines.md (simple_insn): Add mmu.
9490 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9492 * config/nds32/nds32.md (type): Remove call.
9493 * config/nds32/pipelines.md (simple_insn): Likewise.
9495 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9497 * config/nds32/constants.md (unspec_volatile_element): Add
9498 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9499 UNSPEC_VOLATILE_FMFCFG.
9500 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9501 description for fmfcfg and fmfcsr.
9502 (bdesc_1arg): Add fmtcsr.
9503 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9504 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9505 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9506 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9507 unspec_fmfcfg): New patterns.
9508 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9509 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9510 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9511 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9512 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9513 __nds32__fmfcfg): Define.
9515 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9517 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9518 intrinsic register names.
9519 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9520 intrinsic register enum values and macros.
9522 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9524 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9525 for load/store addressing form.
9526 (nds32_print_operand_address): Likewise.
9528 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9531 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9532 based on LABEL_REF. Remove useless assertion.
9533 (pic_address_needs_scratch): Fix formatting.
9534 (sparc_legitimize_pic_address): Minor tweaks.
9535 (sparc_delegitimize_address): Adjust assertion accordingly.
9536 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9537 into symbolic_operand.
9538 (movsi_high_pic_label_ref): Likewise.
9539 (movsi_lo_sum_pic_label_ref): Likewise.
9540 (movdi_pic_label_ref): Likewise.
9541 (movdi_high_pic_label_ref): Likewise.
9542 (movdi_lo_sum_pic_label_ref): Likewise.
9544 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9546 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9547 custom LIB_SPEC setup.
9549 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9550 Kito Cheng <kito.cheng@gmail.com>
9552 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9553 * config/riscv/freebsd.h: New.
9555 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9557 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9558 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9561 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9562 Kito Cheng <kito.cheng@gmail.com>
9564 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9565 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9566 * config/nds32/nds32-protos.h (nds32_output_call,
9567 nds32_output_return): Declare.
9568 * config/nds32/nds32.md: Refine all the call and return patterns.
9570 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9573 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9574 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9576 PR rtl-optimization/84872
9577 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9578 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9581 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9583 * expr.c (copy_blkmode_to_reg): Revert 254862.
9584 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9586 2018-04-06 Richard Biener <rguenther@suse.de>
9589 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9590 after seeing a component reference with an adjacent field. Treat
9591 refs to arrays at struct end of external decls similar to
9592 refs to unconstrained commons.
9594 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9597 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9598 look through SAVE_EXPRs with non-side-effects argument. Adjust
9600 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9603 2018-04-06 Richard Biener <rguenther@suse.de>
9606 * alias.c (find_base_term): New wrapper around find_base_term
9607 unwinding CSELIB_VAL_PTR changes.
9608 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9611 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9613 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9615 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9616 constant definitions.
9617 ("nop"): lr 0,0 -> nopr r0
9618 ("nop_lr0", "nop_lr1"): New insn definitions.
9620 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9622 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9623 NDS32_V3PUSH_AVAILABLE_P macro.
9625 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9626 Chung-Ju Wu <jasonwucj@gmail.com>
9628 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9629 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9630 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9631 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9632 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9633 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9634 * config/nds32/constraints.md: New constraints and checking for hard
9635 float configuration.
9636 * config/nds32/iterators.md: New mode iterator and attribute for hard
9637 float configuration.
9638 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9640 * config/nds32/nds32-fpu.md: New file.
9641 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9642 deal with hard float code generation.
9643 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9645 (abi_type, float_reg_number): New enum type.
9646 * config/nds32/nds32-predicates.c: New predicates for hard float.
9647 * config/nds32/nds32-protos.h: Declare functions for hard float.
9648 * config/nds32/nds32.c: Implementation for hard float configuration.
9649 * config/nds32/nds32.h: Definitions for hard float configuration.
9650 * config/nds32/nds32.md: Include hard float machine description and
9651 modify patterns for hard float configuration.
9652 * config/nds32/nds32.opt: New options for hard float configuration.
9653 * config/nds32/predicates.md: New predicates for hard float
9656 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9658 * common/config/nds32/nds32-common.c
9659 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9661 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9664 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9665 CONSTRUCTOR_ELT (ctor, ...)->value.
9667 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9670 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9672 2018-04-05 Tom de Vries <tom@codesourcery.com>
9675 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9678 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9679 Kito Cheng <kito.cheng@gmail.com>
9681 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9682 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9683 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9684 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9686 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9687 Kito Cheng <kito.cheng@gmail.com>
9689 * config/nds32/constraints.md (Ufe): New memory constraint.
9690 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9691 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9692 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9694 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9695 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9697 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9699 * config/nds32/nds32.md: Use optimize_size in the condition for
9700 alu-shift instructions.
9702 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9704 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9706 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9708 * config/nds32/nds32.md (negsi2): Refine pattern.
9710 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9711 Chung-Ju Wu <jasonwucj@gmail.com>
9713 * config/nds32/iterators.md (shift_rotate): New code iterator.
9714 (shift): New code attribute.
9715 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9716 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9717 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9718 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9719 bit-wise operations.
9720 (andsi3, *andsi3): Ditto.
9721 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9722 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9723 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9724 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9725 nds32_ior_operand, nds32_xor_operand): New predicates.
9727 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9729 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9730 (addsi3, subsi3): ... this.
9732 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9734 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9736 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9738 * config/nds32/nds32.md: Adjust indention.
9740 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9742 * config/nds32/nds32.md (feature): New attribute.
9744 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9746 * config/nds32/nds32.md (subtype): New attribute.
9748 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9751 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9752 expansion to perform a bitwise AND of the argument followed by a
9753 boolean negation of the result.
9755 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9757 PR rtl-optimization/84878
9758 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9759 the basic block. Assert the use reference is not artificial and that
9760 it has an associated insn.
9762 2018-04-04 Michael Matz <matz@suse.de>
9764 * builtins.c (compute_objsize): Pass correct operand
9765 to array_at_struct_end_p.
9767 2018-04-04 Richard Biener <rguenther@suse.de>
9770 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9771 from contexts for DINFO_LEVEL_TERSE and below.
9773 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9775 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9776 resiter_operand condition.
9777 * config/nds32/nds32.md (*move<mode>): Ditto.
9779 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9780 Monk Chiang <sh.chiang04@gmail.com>
9782 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9784 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9786 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9788 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9789 Kito Cheng <kito.cheng@gmail.com>
9791 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9792 nds32_cond_code_str, output_cond_branch,
9793 output_cond_branch_compare_zero, nds32_expand_cbranch,
9794 nds32_expand_cstore, nds32_expand_movcc,
9795 nds32_output_cbranchsi4_equality_zero,
9796 nds32_output_cbranchsi4_equality_reg,
9797 nds32_output_cbranchsi4_equality_reg_or_const_int,
9798 nds32_output_cbranchsi4_greater_less_zero: New functions.
9799 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9800 nds32_expand_cstore, nds32_expand_movcc,
9801 nds32_output_cbranchsi4_equality_zero,
9802 nds32_output_cbranchsi4_equality_reg,
9803 nds32_output_cbranchsi4_equality_reg_or_const_int,
9804 nds32_output_cbranchsi4_greater_less_zero): Declare.
9805 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9806 nds32_rimm11s_operand): New predicates.
9807 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9808 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9811 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9813 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9814 * config/nds32/nds32.md: Ditto.
9815 * config/nds32/pipelines.md: Ditto.
9817 2018-04-04 Richard Biener <rguenther@suse.de>
9819 PR tree-optimization/85168
9820 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9821 propagating abnormals.
9823 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9825 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9827 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9828 Kito Cheng <kito.cheng@gmail.com>
9830 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9831 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9832 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9833 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9834 * config/nds32/nds32.md (sibcall_internal): New.
9835 (sibcall_register): Remove.
9836 (sibcall_immediate): Remove.
9837 (sibcall_value_internal): New.
9838 (sibcall_value_register): Remove.
9839 (sibcall_value_immediate): Remove.
9840 * config/nds32/predicates.md (nds32_general_register_operand): New.
9841 (nds32_call_address_operand): New.
9843 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9845 PR rtl-optimization/85167
9846 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9847 bb_defs if *split_p, instead preinitialize it to NULL.
9849 PR tree-optimization/85156
9850 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9851 evaluating the argument multiple times.
9853 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9855 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9857 (_mm_cvtpd_ps): Likewise.
9858 (_mm_cvttpd_epi32): Likewise.
9859 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9860 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9861 vector, pixel, and bool following altivec.h include.
9863 2018-04-03 Martin Sebor <msebor@redhat.com>
9865 * doc/extend.texi (Common Function Attributes): Clarify.
9866 (const attribute): Likewise.
9867 (pure attribute): Likewise.
9869 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9872 * config/i386/i386.c (ix86_expand_vector_set): Use
9873 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9875 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9877 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9878 instructions when changing rounding bits to preserve precision bits
9879 in the x87 control word.
9881 2018-04-03 Martin Liska <mliska@suse.cz>
9883 PR tree-optimization/82491
9884 * rtl.h (strip_offset_and_add): Replace += suboffset with
9885 poly_uint64 () + suboffset.
9887 2018-03-29 Martin Liska <mliska@suse.cz>
9888 Martin Jambor <mjambor@suse.cz>
9891 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9892 param_type is not an integral or pointer type.
9894 2018-04-03 Richard Biener <rguenther@suse.de>
9896 * sese.h (recompute_all_dominators): Remove.
9898 2018-04-02 Martin Sebor <msebor@redhat.com>
9900 * doc/invoke.texi (-Wrestrict): Fix typos.
9902 2018-04-02 Jim Wilson <jimw@sifive.com>
9904 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9905 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9906 (<optab>di3, <optab>si3_extend): Likewise.
9907 (<optab>si3_mask, <optab>si3_mask_1): New.
9908 (<optab>di3_mask, <optab>di3_mask_1): New.
9909 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9910 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9911 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9913 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9915 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9918 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9920 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9921 (nds32_canonicalize_comparison): New function.
9923 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9924 Kito Cheng <kito.cheng@gmail.com>
9925 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9927 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9928 * config/nds32/constants.md (unspec_volatile_element): Add
9929 UNSPEC_VOLATILE_RELAX_GROUP.
9930 * config/nds32/nds32-relax-opt.c: New file.
9931 * config/nds32/nds32-predicates.c
9932 (nds32_symbol_load_store_p): New function.
9933 * config/nds32/nds32-protos.h
9934 (nds32_symbol_load_store_p): Declare function.
9935 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9936 * config/nds32/nds32.c
9937 (nds32_register_pass): New function to register pass.
9938 (nds32_register_passes): New function to register passes.
9939 * config/nds32/nds32.md (relax_group): New pattern.
9940 * config/nds32/nds32.opt (mrelax-hint): New option.
9941 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9943 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9945 * config/nds32/t-nds32: Modify files dependency.
9947 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9949 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9950 (PROFILE_HOOK): Define its implementation.
9952 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9954 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9955 type and 32-bit size.
9957 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9960 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9961 (V_128_256): New mode iterator.
9962 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9963 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9964 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9966 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9967 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9969 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9972 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9973 NaN inputs correctly.
9975 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9978 * config/rs6000/vsx.md (??r): New mode attribute.
9979 (*vsx_mov<mode>_64bit): Use it.
9980 (*vsx_mov<mode>_32bit): Likewise.
9982 2018-03-30 Martin Sebor <msebor@redhat.com>
9984 PR tree-optimization/84818
9985 * builtins.c (check_access): Use warning_n.
9987 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9990 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9992 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9995 2018-03-30 Julia Koval <julia.koval@intel.com>
9998 * x86-tune.def (movx, partial_reg_dependency): Enable for
10001 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10003 PR inline-asm/84985
10004 * lra-constraints.c (process_alt_operands): Move setting
10005 this_alternative_matches below.
10007 2018-03-29 Martin Liska <mliska@suse.cz>
10010 * doc/invoke.texi: Document how LTO works with debug info.
10011 Describe auto-load support of binutils. Mention 'x86-64'
10012 as valid option value of -march option.
10014 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10016 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10019 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10020 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10021 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10024 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10027 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10028 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10029 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10030 for RS6000_BTM_POWERPC64.
10031 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10032 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10033 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10036 (DIVDEU): Likewise.
10038 2018-03-28 Carl Love <cel@us.ibm.com>
10041 2017-09-27 Carl Love <cel@us.ibm.com>
10043 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10044 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10045 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10048 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10050 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10051 instead of __vector bool.
10052 (_mm_max_pu8): Likewise.
10053 (_mm_min_pi16): Likewise.
10055 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10058 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10059 (DIVWEUO): Likewise.
10060 (DIVDEO): Likewise.
10061 (DIVDEUO): Likewise.
10062 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10063 DIVWEUO and DIVDEUO.
10064 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10065 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10066 (div_extend): Likewise.
10067 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10069 (__builtin_divweuo): Likewise.
10070 (__builtin_divdeo): Likewise.
10071 (__builtin_divdeuo): Likewise.
10073 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10076 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10077 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10079 PR tree-optimization/82004
10080 * gimple-match-head.c (optimize_pow_to_exp): New function.
10081 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10082 Don't fold to exp if optimize_pow_to_exp is false.
10084 2018-03-28 Martin Liska <mliska@suse.cz>
10087 * calls.c (initialize_argument_information): Fix trailing space.
10088 * common.opt: Fix typo and provide better explanation for
10089 -fsanitize-coverage option.
10090 * config/i386/i386.opt: Fix typo.
10092 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10093 Martin Liska <mliska@suse.cz>
10096 * gimplify.c (asan_poison_variable): Don't do the check for
10097 gimplify_omp_ctxp here.
10098 (gimplify_decl_expr): Do it here.
10099 (gimplify_target_expr): Likewise.
10101 2018-03-28 Martin Liska <mliska@suse.cz>
10104 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10105 chkp_type_bounds_count if MPX is not enabled.
10107 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10109 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10111 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10114 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10115 function to create the function decl for complex long double
10116 multiply and divide for -mabi=ieeelongdouble.
10117 (init_float128_ieee): Call it.
10119 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10122 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10123 -fcf-protection=branch -mibt.
10124 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10126 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10129 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10130 arm_word_relocations.
10132 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10135 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10136 extern array declarations.
10138 2018-03-27 Richard Biener <rguenther@suse.de>
10140 PR middle-end/84067
10141 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10142 explicit single_use checks.
10144 2018-03-27 Richard Biener <rguenther@suse.de>
10146 PR tree-optimization/85082
10147 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10150 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10152 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10153 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10154 Turn on fasynchronous-unwind-tables and funwind-tables.
10156 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10159 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10160 (*bmi_blsr_<mode>_ccz): Ditto.
10162 2018-03-26 Tom de Vries <tom@codesourcery.com>
10164 PR tree-optimization/85063
10165 * omp-general.c (offloading_function_p): New function. Factor out
10167 * omp-offload.c (pass_omp_target_link::gate): ... here.
10168 * omp-general.h (offloading_function_p): Declare.
10169 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10170 with attribute omp declare target for offloading functions.
10172 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10174 PR tree-optimization/84005
10175 * tree-data-ref.h (get_base_for_alignment): Declare.
10176 * tree-data-ref.c (get_base_for_alignment_1): New function.
10177 (get_base_for_alignment): Likewise.
10178 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10179 get_base_for_alignment to find a suitable base object, instead
10180 of always using drb->base_address.
10182 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10184 PR inline-asm/85022
10185 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10186 known size by default.
10188 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10190 PR inline-asm/85030
10191 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10192 and non BLKmode operands.
10194 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10197 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10198 Clean up attributes.
10200 2018-03-23 Richard Biener <rguenther@suse.de>
10203 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10204 we are going to emit early debug for LTO.
10206 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10208 PR inline-asm/85034
10209 * function.c (match_asm_constraints_1): Don't optimize if input
10210 doesn't satisfy general_operand predicate for output's mode.
10212 PR inline-asm/85022
10213 * alias.c (write_dependence_p): Don't require for x_canonicalized
10214 non-VOIDmode if x has VOIDmode.
10217 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10218 just don't try to optimize it rather than assert it never happens.
10220 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10222 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10223 macro expansions for definition of ST_INTERNAL_<mode> and
10224 LD_INTERNAL_<mode> builtins.
10225 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10227 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10229 (altivec_expand_st_builtin): Likewise.
10230 (altivec_expand_builtin): Remove calls to deleted functions.
10231 (rs6000_address_for_altivec): Delete this function.
10232 * config/rs6000/vector.md: Remove expands for
10233 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10235 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10238 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10239 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10240 re-computing once computed.
10241 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10242 (arm_init_machine_status): Initialize
10243 machine->static_chain_stack_bytes.
10245 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10248 * doc/extend.texi: Add four new prototypes for vec_ld.
10249 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10250 definitions for more logical presentation.
10251 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10252 entries for V1TI variants of __builtin_altivec_ld builtin.
10253 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10254 handling of V1TI variant of LVX icode pattern.
10255 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10256 (rs6000_gimple_fold_builtin): Likewise.
10257 (altivec_init_builtins): Add code to define
10258 __builtin_altivec_lvx_v1ti function.
10260 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10262 PR inline-asm/84941
10263 * function.c (match_asm_constraints_1): Don't do the optimization
10264 if input isn't a REG, SUBREG, MEM or constant.
10266 2018-03-22 Tom de Vries <tom@codesourcery.com>
10268 PR tree-optimization/84956
10269 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10270 bb_has_abnormal_pred.
10272 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10275 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10276 DECL_INITIAL (decl) to decl at the end.
10277 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10278 adjust the comment.
10280 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10282 * doc/extend.texi (__builtin_tgmath): Document when complex
10283 integer types are treated as _Complex _Float64.
10285 2018-03-21 Tom de Vries <tom@codesourcery.com>
10287 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10289 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10291 PR tree-optimization/84960
10292 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10293 if it is ENTRY block, move them into single succ of ENTRY in that case.
10295 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10297 PR tree-optimization/84811
10298 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10299 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10300 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10301 (known_subrange_p): Update accordingly. Cast each value involved
10302 in the size comparison, rather than casting the result of the
10305 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10307 PR tree-optimization/84982
10308 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10309 by flipping the least significant bit rather than all bits from
10310 bitpos to bitpos + bitsize - 1.
10312 2018-03-21 Nathan Sidwell <nathan@acm.org>
10314 * doc/extend.texi (Deprecated Features): Remove mention of
10315 long-deleted deprecations.
10317 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10320 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10321 * configure: Regenerate.
10323 2018-03-21 Tom de Vries <tom@codesourcery.com>
10325 PR tree-optimization/83126
10326 * tree-parloops.c (num_phis): New function.
10327 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10329 2018-03-21 Nathan Sidwell <nathan@acm.org>
10331 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10332 mention anon-struct/union members and trailing attributes.
10334 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10336 PR tree-optimization/84969
10337 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10338 builtin memset partitions if they set different rhs values.
10340 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10342 PR rtl-optimization/84989
10343 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10344 VEC_DUPLICATE with scalar result mode.
10346 2018-03-21 Martin Liska <mliska@suse.cz>
10349 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10350 not intended return statement.
10352 2018-03-21 Martin Liska <mliska@suse.cz>
10355 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10356 (chkp_find_bound_slots_1): Limit number of iterations.
10358 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10361 * Minor grammar fixes for x86 options.
10363 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10366 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10367 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10369 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10372 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10373 (altivec_lvx_<mode>_1op): Likewise.
10374 (altivec_stvx_<mode>_2op): Likewise.
10375 (altivec_stvx_<mode>_1op): Likewise.
10376 (altivec_lvx_<VM2:mode>): New define_expand.
10377 (altivec_stvx_<VM2:mode>): Likewise.
10378 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10379 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10380 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10381 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10382 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10383 (rs6000_gen_lvx): Likewise.
10384 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10385 (altivec_expand_stv_builtin): Likewise.
10386 (altivec_expand_builtin): Likewise.
10387 * config/rs6000/vector.md: Likewise.
10389 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10392 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10395 2018-03-20 Richard Biener <rguenther@suse.de>
10398 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10399 sign-conversions as zero, fall back to standard scalar_stmt
10402 2018-03-20 Martin Liska <mliska@suse.cz>
10405 * predict.c (rebuild_frequencies): Handle case when we have
10406 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10408 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10411 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10412 flag_section_anchors.
10413 * varasm.c (use_blocks_for_decl_p): Remove hack for
10414 dw2_force_const_mem.
10417 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10419 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10420 be created, use lowpart_subreg of operands[0] rather than operands[0]
10422 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10423 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10424 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10425 and n constraint instead of aarch64_shift_imm_di and Usd.
10426 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10427 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10429 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10432 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10433 to favor GPR over NEON registers.
10434 (<shift>di3_neon): Likewise.
10436 2018-03-20 Tom de Vries <tom@codesourcery.com>
10439 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10440 (nvptx_process_pars): Emit bar.sync asap and alap.
10442 2018-03-20 Tom de Vries <tom@codesourcery.com>
10445 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10446 seen_label if seen_label is already set.
10448 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10451 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10452 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10453 Use 1U instead of 1. Formatting fixes.
10456 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10457 instead of TREE_TYPE (s1) for the return value.
10459 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10461 PR tree-optimization/84946
10462 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10463 bitsize + bitsize in poly_uint64 rather than poly_int64.
10466 * dwarf2asm.c: Include fold-const.c.
10467 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10468 of decl rather than decl itself.
10470 PR rtl-optimization/84643
10471 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10473 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10476 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10477 calling assemble_variable.
10479 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10482 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10483 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10485 2018-03-19 Jim Wilson <jimw@sifive.com>
10488 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10489 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10490 (riscv_first_stack_step): Likewise.
10491 (riscv_option_override): Use STACK_BOUNDARY instead of
10492 MIN_STACK_BOUNDARY.
10493 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10494 MIN_STACK_BOUNDARY.
10495 (BIGGEST_ALIGNMENT): Set to 128.
10496 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10497 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10500 2018-03-19 Richard Biener <rguenther@suse.de>
10502 PR tree-optimization/84933
10503 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10504 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10506 2018-03-19 Richard Biener <rguenther@suse.de>
10508 PR tree-optimization/84859
10509 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10510 (cond_if_else_store_replacement): Perform sinking operation on
10511 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10512 Generalize what a BB with a single eligible store is.
10514 2018-03-19 Richard Biener <rguenther@suse.de>
10516 PR tree-optimization/84929
10517 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10518 chrec_is_positive against non-chrec arg.
10520 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10523 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10525 2018-03-18 Martin Liska <mliska@suse.cz>
10527 PR rtl-optimization/84635
10528 * regrename.c (build_def_use): Use matches_mode only when
10531 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10533 PR tree-optimization/84913
10534 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10535 vectorize chains of COND_EXPRs.
10537 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10539 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10541 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10543 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10545 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10547 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10549 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10550 Kito Cheng <kito.cheng@gmail.com>
10552 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10553 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10554 (nds32_adjust_reg_alloc_order): New function.
10555 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10557 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10559 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10560 nds32_print_operand, nds32_print_operand_address): Use
10561 HOST_WIDE_INT_PRINT_DEC instead.
10563 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10565 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10567 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10570 * config/i386/i386.c (initial_ix86_tune_features,
10571 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10572 unsigned long long.
10573 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10574 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10575 rather than 1u << ix86_tune. Formatting fix.
10576 (ix86_option_override_internal): Change ix86_arch_mask from
10577 unsigned int to unsigned HOST_WIDE_INT, initialize to
10578 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10579 (ix86_function_specific_restore): Likewise.
10581 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10584 * postreload.c (reload_combine_recognize_pattern): Perform
10585 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10586 truncate_int_for_mode the result for the destination's mode.
10589 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10590 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10592 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10595 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10598 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10599 regno_allocno_class_array and sorted_pseudos.
10600 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10601 insns where regno is used.
10603 2018-03-16 Martin Liska <mliska@suse.cz>
10606 * multiple_target.c (create_dispatcher_calls): Redirect
10607 reference in the symbol table.
10609 2018-03-16 Martin Liska <mliska@suse.cz>
10612 * multiple_target.c (create_dispatcher_calls): Redirect also
10615 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10619 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10620 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10621 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10623 2018-03-16 Julia Koval <julia.koval@intel.com>
10625 * doc/invoke.texi (Skylake Server): Add CLWB.
10626 Cannonlake): Remove CLWB.
10628 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10630 PR tree-optimization/84841
10631 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10633 (FLOAT_ONE_CONST_TYPE): Define.
10634 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10635 (sort_by_operand_rank): Put entries with higher constant_type last
10636 rather than first to match comments.
10638 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10640 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10643 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10646 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10647 check and use address of target always.
10649 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10652 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10653 (indirect_thunk_bnd_needed): Likewise.
10654 (indirect_thunks_used): Likewise.
10655 (indirect_thunks_bnd_used): Likewise.
10656 (indirect_return_needed): New.
10657 (indirect_return_bnd_needed): Likewise.
10658 (output_indirect_thunk_function): Add a bool argument for
10660 (output_indirect_thunk_function): Don't generate alias for
10661 function return thunk.
10662 (ix86_code_end): Call output_indirect_thunk_function to generate
10663 function return thunks.
10664 (ix86_output_function_return): Set indirect_return_bnd_needed
10665 and indirect_return_needed instead of indirect_thunk_bnd_needed
10666 and indirect_thunk_needed.
10668 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10670 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10671 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10672 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10674 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10675 Paul Hua <paul.hua.gm@gmail.com>
10678 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10680 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10682 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10683 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10684 resp. SFmode cases.
10686 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10689 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10690 instead of GET_MODE_SIZE when comparing Units.
10692 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10695 * varasm.c (hash_section): Return an unchangeble hash value
10696 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10697 Return !aarch64_can_use_per_function_literal_pools_p ().
10699 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10702 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10703 rather than address of cmode as last argument to prepare_cmp_insn.
10705 2018-03-15 Julia Koval <julia.koval@intel.com>
10707 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10708 F_AVX512VNNI, F_AVX512BITALG): New.
10710 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10713 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10714 insn for floating-point loads and stores.
10716 2018-03-14 Carl Love <cel@us.ibm.com>
10718 * config/rs6000/rs6000-c.c: Add macro definitions for
10719 ALTIVEC_BUILTIN_VEC_PERMXOR.
10720 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10721 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10722 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10723 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10725 * config/doc/extend.texi: Add prototypes for vec_permxor.
10727 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10730 * diagnostic-show-locus.c (class layout_point): Convert m_line
10731 from int to linenum_type.
10732 (line_span::comparator): Use linenum "compare" function when
10733 comparing line numbers.
10734 (test_line_span): New function.
10735 (layout_range::contains_point): Convert param "row" from int to
10737 (layout_range::intersects_line_p): Likewise.
10738 (layout::will_show_line_p): Likewise.
10739 (layout::print_source_line): Likewise.
10740 (layout::should_print_annotation_line_p): Likewise.
10741 (layout::print_annotation_line): Likewise.
10742 (layout::print_leading_fixits): Likewise.
10743 (layout::annotation_line_showed_range_p): Likewise.
10744 (struct line_corrections): Likewise for field m_row.
10745 (line_corrections::line_corrections): Likewise for param "row".
10746 (layout::print_trailing_fixits): Likewise.
10747 (layout::get_state_at_point): Likewise.
10748 (layout::get_x_bound_for_row): Likewise.
10749 (layout::print_line): Likewise.
10750 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10751 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10752 * input.c (selftest::test_linenum_comparisons): New function.
10753 (selftest::input_c_tests): Call it.
10754 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10755 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10756 * selftest.h (ASSERT_GT): New macro.
10757 (ASSERT_GT_AT): New macro.
10758 (ASSERT_LT): New macro.
10759 (ASSERT_LT_AT): New macro.
10761 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10763 PR rtl-optimization/84780
10764 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10766 2018-03-14 Martin Liska <mliska@suse.cz>
10768 * tree.c (record_node_allocation_statistics): Use
10769 get_stats_node_kind.
10770 (get_stats_node_kind): New function extracted from
10771 record_node_allocation_statistics.
10772 (free_node): Use get_stats_node_kind.
10774 2018-03-14 Richard Biener <rguenther@suse.de>
10776 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10777 that the value-set of ANTIC_IN doesn't grow.
10780 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10782 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10783 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10785 2018-03-14 Julia Koval <julia.koval@intel.com>
10787 * config.gcc (icelake-client, icelake-server): New.
10789 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10790 (initial_ix86_arch_features): Ditto.
10791 (PTA_SKYLAKE): Add SGX.
10792 (PTA_ICELAKE): Remove.
10793 (PTA_ICELAKE_CLIENT): New.
10794 (PTA_ICELAKE_SERVER): New.
10795 (ix86_option_override_internal): Split up icelake on icelake client and
10797 (get_builtin_code_for_version): Ditto.
10798 (fold_builtin_cpu): Ditto.
10799 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10800 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10801 * config/i386/i386.h (processor_type): Ditto.
10802 * doc/invoke.texi: Ditto.
10804 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10807 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10808 INTEGER_CST offset, add it together with bitpos / 8 and
10809 sign extend based on POINTER_SIZE.
10813 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10816 * config/i386/constraints.md (Yc): New register constraint.
10817 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10818 Use Yc constraint for alternative 2 of operand 0. Remove
10819 preferred_for_speed attribute.
10821 2018-03-14 Richard Biener <rguenther@suse.de>
10823 PR tree-optimization/84830
10824 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10825 with the old one to avoid oscillations.
10827 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10830 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10832 (assign_by_spills): Return a flag of reload assignment failure.
10833 Do not process the reload assignment failures. Do not spill other
10834 reload pseudos if they has the same reg class. Update n if
10836 (lra_assign): Add a return arg. Set up from the result of
10837 assign_by_spills call.
10838 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10839 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10840 usage_insns if it is not NULL.
10841 (spill_hard_reg_in_range): New function.
10842 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10843 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10844 function prototypes.
10845 (lra_assign): Change prototype.
10846 * lra.c (lra): Add code to deal with fails by splitting hard reg
10849 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10851 * config/riscv/riscv.opt (mrelax): New option.
10852 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10853 "norelax" when riscv_mrelax is disabled.
10854 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10856 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10859 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10860 reassociation for int modes.
10862 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10864 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10865 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10867 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10868 * config/aarch64/aarch64-sve.md
10869 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10870 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10871 (*extend<mode><Vwide>2): Rename to...
10872 (aarch64_sve_extend<mode><Vwide>2): ...this.
10873 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10874 renaming the old pattern to...
10875 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10877 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10878 define_expand, renaming the old pattern to...
10879 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10880 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10881 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10882 account when deciding which SVE instruction the optab should use.
10883 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10885 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10887 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10888 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10889 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10890 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10891 (tlsdesc_small_<mode>): Turn a define_expand and use
10892 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10893 (tlsdesc_small_advsimd_<mode>): ...this.
10894 (tlsdesc_small_sve_<mode>): New pattern.
10896 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10898 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10899 (UNSPEC_UMUL_HIGHPART): New constants.
10900 (MUL_HIGHPART): New int iteraor.
10901 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10902 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10904 (*<su>mul<mode>3_highpart): New define_insn.
10906 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10909 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10912 2018-03-13 Martin Liska <mliska@suse.cz>
10915 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10917 (sem_item_optimizer::~sem_item_optimizer): Release it.
10918 (sem_item_optimizer::merge_classes): Register variable aliases.
10919 (sem_item_optimizer::fixup_pt_set): New function.
10920 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10921 * ipa-icf.h: Declare new variables and functions.
10923 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10925 PR middle-end/84834
10926 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10927 integer_pow2p@2 and test integer_pow2p in condition.
10928 (A < 0 ? C : 0): Similarly for @1.
10930 PR middle-end/84831
10931 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10932 characters starting at p contain '\0' character, don't look beyond
10936 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10937 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10940 * reg-stack.c (change_stack): Change update_end var from int to
10941 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10942 also call set_block_for_insn on the newly added insns and rescan.
10945 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10946 on the last operand.
10949 * tree.c (stabilize_reference_1): Return save_expr (e) for
10950 STATEMENT_LIST even if it doesn't have side-effects.
10952 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10954 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10956 2018-03-12 Renlin Li <renlin.li@arm.com>
10958 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10959 aarch64_output_scalar_simd_mov_immediate.
10961 2018-03-12 Martin Sebor <msebor@redhat.com>
10963 PR tree-optimization/83456
10964 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10965 for perfectly overlapping calls to memcpy.
10966 (gimple_fold_builtin_memory_chk): Same.
10967 (gimple_fold_builtin_strcpy): Handle no-warning.
10968 (gimple_fold_builtin_stxcpy_chk): Same.
10969 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10971 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10973 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10974 parameter. Use it for SFmode.
10975 (rs6000_function_arg_advance_1): Adjust.
10976 (rs6000_function_arg): Adjust.
10977 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10979 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10981 PR rtl-optimization/84169
10982 PR rtl-optimization/84780
10983 * combine.c (can_combine_p): Check for a 2-insn combination whether
10984 the destination register is used between the two insns, too.
10986 2018-03-12 Richard Biener <rguenther@suse.de>
10988 PR tree-optimization/84803
10989 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10990 for refs DR analysis didn't process.
10992 2018-03-12 Richard Biener <rguenther@suse.de>
10994 PR tree-optimization/84777
10995 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10996 force-vectorize loops ignore whether we are optimizing for size.
10998 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11000 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11001 (TARGET_MD_ASM_ADJUST): Define.
11003 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11004 Kito Cheng <kito.cheng@gmail.com>
11005 Chung-Ju Wu <jasonwucj@gmail.com>
11007 * config/nds32/nds32.c (nds32_compute_stack_frame,
11008 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11009 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11010 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11011 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11012 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11013 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11014 * config/nds32/nds32.md (prologue, epilogue): Use macro
11015 NDS32_V3PUSH_AVAILABLE_P to do checking.
11017 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11020 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11021 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11022 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11023 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11024 addition of most attributes on !orig_type_die or the attribute not
11025 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11027 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11028 Chung-Ju Wu <jasonwucj@gmail.com>
11030 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11031 __NDS32_VH__ macro.
11032 * config/nds32/nds32.opt (mvh): New option.
11034 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11035 Chung-Ju Wu <jasonwucj@gmail.com>
11037 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11039 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11040 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11043 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11044 Chung-Ju Wu <jasonwucj@gmail.com>
11046 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11048 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11049 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11051 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11052 Kito Cheng <kito.cheng@gmail.com>
11053 Chung-Ju Wu <jasonwucj@gmail.com>
11055 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11056 UNSPEC_FFMISM and UNSPEC_FLMISM.
11057 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11058 for ffb, ffmism and flmism.
11059 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11060 (unspec_ffmism): Ditto.
11061 (unspec_flmism): Ditto.
11062 (nds32_expand_builtin_impl): Check if string extension is available.
11063 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11064 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11066 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11069 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11072 * lra-assigns.c (assign_by_spills): Return a flag of reload
11073 assignment failure. Do not process the reload assignment
11074 failures. Do not spill other reload pseudos if they has the same
11076 (lra_assign): Add a return arg. Set up from the result of
11077 assign_by_spills call.
11078 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11079 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11080 usage_insns if it is not NULL.
11081 (spill_hard_reg_in_range): New function.
11082 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11083 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11084 function prototypes.
11085 (lra_assign): Change prototype.
11086 * lra.c (lra): Add code to deal with fails by splitting hard reg
11089 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11092 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11094 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11097 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11098 (dw_val_node): Add val_symbolic_view.
11099 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11100 (symview_upper_bound): New.
11101 (new_line_info_table): Initialize symviews_since_reset.
11102 (dwarf2out_source_line): Count symviews_since_reset and set
11103 symview_upper_bound.
11104 (dw_val_equal_p): Handle symview.
11105 (add_AT_symview): New.
11106 (print_dw_val): Handle symview.
11107 (attr_checksum, attr_checksum_ordered): Likewise.
11108 (same_dw_val_p, size_of_die): Likewise.
11109 (value_format, output_die): Likewise.
11110 (add_high_low_attributes): Use add_AT_symview for entry_view.
11111 (dwarf2out_finish): Reset symview_upper_bound, clear
11114 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11117 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11118 Add strict argument and use it.
11119 (rs6000_split_multireg_move): Update for new strict argument.
11120 (mem_operand_gpr): Disallow all non-offsettable addresses.
11121 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11123 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11126 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11127 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11128 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11131 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11132 decl, use remap_type if we want to use the type.
11134 2018-03-09 Martin Sebor <msebor@redhat.com>
11136 PR tree-optimization/84526
11137 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11139 (builtin_access::generic_overlap): Be prepared to handle non-array
11142 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11144 PR rtl-optimization/84682
11145 * lra-constraints.c (process_address_1): Check is_address flag
11146 for address constraints.
11147 (process_alt_operands): Likewise.
11148 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11149 preprocess_constraints.
11150 * recog.h (preprocess_constraints): Add oploc parameter.
11152 * recog.c (preprocess_constraints): Test address_operand for
11153 CT_ADDRESS constraints.
11155 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11158 * lra-assigns.c (assign_by_spills): Return a flag of reload
11159 assignment failure. Do not process the reload assignment
11160 failures. Do not spill other reload pseudos if they has the same
11162 (lra_assign): Add a return arg. Set up from the result of
11163 assign_by_spills call.
11164 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11165 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11166 usage_insns if it is not NULL.
11167 (spill_hard_reg_in_range): New function.
11168 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11169 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11170 function prototypes.
11171 (lra_assign): Change prototype.
11172 * lra.c (lra): Add code to deal with fails by splitting hard reg
11175 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11178 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11179 Accept complain bool parameter. Only emit errors if it is true.
11180 (arm_parse_cpu_option_name): Likewise.
11181 (arm_target_thumb_only): Adjust callers of the above.
11182 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11183 prototype to take a default true bool parameter.
11184 (arm_parse_arch_option_name): Likewise.
11186 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11187 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11191 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11192 * configure: Regenerate.
11193 * configure.ac ("linker --version-script option"): New.
11194 ("linker soname option"): New.
11196 2018-03-09 Richard Biener <rguenther@suse.de>
11198 PR tree-optimization/84775
11199 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11200 immediate uses of predicate stmts and mark them modified.
11203 PR tree-optimization/84178
11204 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11206 (version_loop_for_if_conversion): Delay update_ssa call.
11207 (tree_if_conversion): Delay update_ssa until after predicate
11210 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11213 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11214 when the function accesses prior frames.
11216 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11219 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11220 gen_llsym, otherwise call maybe_gen_llsym.
11222 PR inline-asm/84742
11223 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11224 has ',' character inside of it.
11226 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11229 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11230 as clobbering CC_REGNUM.
11232 2018-03-08 Richard Biener <rguenther@suse.de>
11234 PR middle-end/84552
11235 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11236 (follow_copies_to_constant): Do not follow SSA names registered
11239 2018-03-08 Richard Biener <rguenther@suse.de>
11241 PR tree-optimization/84178
11242 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11244 (version_loop_for_if_conversion): Delay update_ssa call.
11245 (tree_if_conversion): Delay update_ssa until after predicate
11248 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11250 PR tree-optimization/84178
11251 * tree-if-conv.c (release_bb_predicate): Remove the
11252 the assertion that the stmts have NULL use_ops.
11253 Discard the statements, asserting that they haven't
11254 yet been added to a BB.
11256 2018-03-08 Richard Biener <rguenther@suse.de>
11258 PR tree-optimization/84746
11259 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11260 (phi_translate): Pass in destination ANTIC_OUT set.
11261 (phi_translate_1): Likewise. For a simplified result lookup
11262 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11263 (phi_translate_set): Adjust.
11264 (do_pre_regular_insertion): Likewise.
11265 (do_pre_partial_partial_insertion): Likewise.
11267 2018-03-08 Martin Liska <mliska@suse.cz>
11269 PR gcov-profile/84735
11270 * doc/gcov.texi: Document usage of profile files.
11271 * gcov-io.h: Document changes in the format.
11273 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11277 * dwarf2out.c (struct dw_line_info_table): Update comments for
11279 (FORCE_RESET_NEXT_VIEW): New.
11280 (FORCE_RESETTING_VIEW_P): New.
11281 (RESETTING_VIEW_P): Check for -1 too.
11282 (ZERO_VIEW_P): Likewise.
11283 (new_line_info_table): Force-reset next view.
11284 (dwarf2out_begin_function): Likewise.
11285 (dwarf2out_source_line): Simplify zero_view_p initialization.
11286 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11287 view directly. Omit view when omitting .loc at line 0.
11289 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11291 PR tree-optimization/84740
11292 * tree-switch-conversion.c (process_switch): Call build_constructors
11293 only if info.phi_count is non-zero.
11295 PR tree-optimization/84739
11296 * tree-tailcall.c (find_tail_calls): Check call arguments against
11297 DECL_ARGUMENTS (current_function_decl) rather than
11298 DECL_ARGUMENTS (func) when checking for tail recursion.
11300 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11302 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11303 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11304 Volker Reichelt's entry and add entries for people that perform
11305 GCC fuzzy testing and report numerous bugs.
11307 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11310 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11311 readonly data in sdata, if that is disabled.
11312 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11313 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11314 -mreadonly-in-sdata option.
11316 2018-03-07 Martin Sebor <msebor@redhat.com>
11318 PR tree-optimization/84468
11319 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11320 basic block when looking for nul assignment.
11322 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11325 * except.h (output_function_exception_table): Adjust prototype.
11326 * except.c (output_function_exception_table): Remove FNNAME parameter
11327 and add SECTION parameter. Ouput one part of the table at a time.
11328 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11329 the first part of the exception table and emit unwind directives.
11330 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11331 (i386_pe_seh_cold_init): Likewise.
11332 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11333 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11334 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11335 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11336 * config/i386/winnt.c: Include except.h.
11337 (struct seh_frame_state): Add reg_offset, after_prologue and
11338 in_cold_section fields.
11339 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11340 (i386_pe_seh_cold_init): New function.
11341 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11342 to seh->in_cold_section.
11343 (seh_emit_push): Record the offset of the push.
11344 (seh_emit_save): Record the offet of the save.
11345 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11346 Test seh->after_prologue to disregard the epilogue.
11347 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11348 (i386_pe_end_cold_function): New function.
11350 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11353 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11354 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11357 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11359 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11360 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11362 PR middle-end/84723
11363 * multiple_target.c: Include tree-inline.h and intl.h.
11364 (expand_target_clones): Diagnose and fail if node->definition and
11365 !tree_versionable_function_p (node->decl).
11367 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11369 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11371 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11372 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11373 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11375 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11378 * combine.c (try_combine): Use reg_or_subregno instead of handling
11379 just paradoxical SUBREGs and REGs.
11381 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11383 * config/arc/arc.c (arc_finalize_pic): Remove function.
11384 (arc_must_save_register): We use single base PIC register, remove
11385 checks to save/restore the PIC register.
11386 (arc_expand_prologue): Likewise.
11387 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11389 (arc_verify_short): Likewise.
11390 (arc_attr_type): Likewise.
11391 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11392 (walk_stores): Likewise.
11393 (arc_address_cost): Make it static.
11394 (arc_verify_short): Likewise.
11395 (branch_dest): Likewise.
11396 (arc_attr_type): Likewise.
11397 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11398 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11399 (arc_final_prescan_insn): Remove inserting the nops due to
11400 hardware hazards. It is done in reorg step.
11401 (insn_length_variant_t): Remove.
11402 (insn_length_parameters_t): Likewise.
11403 (arc_insn_length_parameters): Likewise.
11404 (arc_get_insn_variants): Likewise.
11405 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11407 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11409 PR inline-asm/84683
11410 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11413 PR tree-optimization/84687
11414 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11416 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11418 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11420 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11421 Rename to ppc_speculation_barrier.
11422 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11423 __builtin_ppc_speculation_barrier.
11425 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11428 * combine.c (combine_simplify_rtx): Don't try to simplify if
11429 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11432 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11434 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11435 to 32 bytes when compiling for POWER9.
11437 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11440 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11441 regparm >= 3 with no arg reg available also for calls with
11442 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11445 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11447 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11449 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11452 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11454 2018-03-05 Richard Biener <rguenther@suse.de>
11456 PR tree-optimization/84486
11457 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11458 When inserting a __builtin_assume_aligned call set the LHS
11459 SSA name alignment info accordingly.
11461 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11463 PR tree-optimization/84114
11464 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11465 Avoid reassociation of FLOAT_MODE addition.
11467 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11469 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11470 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11471 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11472 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11473 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11474 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11475 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11477 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11478 __builtin_ia32_wbinvd): New builtins.
11479 (SPECIAL_ARGS2): New.
11480 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11481 (SPECIAL_ARGS2): New.
11482 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11483 (ix86_valid_target_attribute_inner_p): Ditto.
11484 (ix86_init_mmx_sse_builtins): Add special_args2.
11485 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11486 TARGET_WBNOINVD_P): New.
11487 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11488 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11489 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11490 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11491 * config/i386/pconfigintrin.h: New file.
11492 * config/i386/wbnoinvdintrin.h: Ditto.
11493 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11495 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11497 2018-03-05 Richard Biener <rguenther@suse.de>
11499 PR tree-optimization/84670
11500 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11502 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11503 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11504 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11505 doesn't grow if all successors (recursively) were visited at least
11508 2018-03-05 Richard Biener <rguenther@suse.de>
11510 PR tree-optimization/84650
11511 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11512 if executed in the loop pipeline.
11514 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11516 * doc/configfiles.texi (Configuration Files): Move info about
11517 conditionalizing $target-protos.h to...
11518 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11519 differs from $target-protos.h.
11521 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11522 Chung-Ju Wu <jasonwucj@gmail.com>
11524 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11525 * config/nds32/nds32-multiple.md (setmemsi): Define.
11526 * config/nds32/nds32-memory-manipulation.c
11527 (nds32_gen_dup_4_byte_to_word_value): New.
11528 (emit_setmem_word_loop): New.
11529 (emit_setmem_byte_loop): New.
11530 (nds32_expand_setmem_loop): New.
11531 (nds32_expand_setmem_loop_v3m): New.
11532 (nds32_expand_setmem_unroll): New.
11533 (nds32_expand_setmem): New.
11535 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11536 Chung-Ju Wu <jasonwucj@gmail.com>
11538 * config/nds32/nds32-memory-manipulation.c
11539 (nds32_emit_load_store): New.
11540 (nds32_emit_post_inc_load_store): New.
11541 (nds32_emit_mem_move): New.
11542 (nds32_emit_mem_move_block): New.
11543 (nds32_expand_movmemsi_loop_unknown_size): New.
11544 (nds32_expand_movmemsi_loop_known_size): New.
11545 (nds32_expand_movmemsi_loop): New.
11546 (nds32_expand_movmemsi_unroll): New.
11547 (nds32_expand_movmemqi): Rename ...
11548 (nds32_expand_movmemsi): ... to this.
11549 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11550 (movmemsi): ... to this.
11551 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11552 (nds32_expand_movmemsi): ... to this.
11554 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11555 Monk Chiang <sh.chiang04@gmail.com>
11556 Chung-Ju Wu <jasonwucj@gmail.com>
11558 * config/nds32/nds32-protos.h
11559 (nds32_expand_load_multiple): New arguments.
11560 (nds32_expand_store_multiple): Ditto.
11561 (nds32_valid_multiple_load_store): Rename ...
11562 (nds32_valid_multiple_load_store_p): ... to this.
11563 * config/nds32/nds32-memory-manipulation.c
11564 (nds32_expand_load_multiple): Refine implementation.
11565 (nds32_expand_store_multiple): Ditto.
11566 * config/nds32/nds32-multiple.md
11567 (load_multiple): Update nds32_expand_load_multiple interface.
11568 (store_multiple): Update nds32_expand_store_multiple interface.
11569 * config/nds32/nds32-predicates.c
11570 (nds32_valid_multiple_load_store): Rename ...
11571 (nds32_valid_multiple_load_store_p): ... to this and refine
11573 * config/nds32/predicates.md
11574 (nds32_load_multiple_and_update_address_operation): New predicate.
11575 (nds32_store_multiple_and_update_address_operation): New predicate.
11577 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11578 Chung-Ju Wu <jasonwucj@gmail.com>
11580 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11581 (combo): New attribute.
11582 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11584 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11586 * config/nds32/nds32.opt: Change -mcmodel= default value.
11588 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11589 Monk Chiang <sh.chiang04@gmail.com>
11590 Chung-Ju Wu <jasonwucj@gmail.com>
11592 * config/nds32/constants.md (unspec_element): New enum.
11593 * config/nds32/constraints.md (Umw): New constraint.
11594 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11595 * config/nds32/nds32-intrinsic.md: Likewise.
11596 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11597 (nds32_valid_smw_lwm_base_p): New.
11598 (nds32_output_smw_single_word): New.
11599 (nds32_output_lmw_single_word): New.
11600 (nds32_expand_unaligned_load): New.
11601 (nds32_expand_unaligned_store): New.
11602 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11603 (nds32_output_smw_single_word): Declare.
11604 (nds32_output_lmw_single_word): Declare.
11605 (nds32_expand_unaligned_load): Declare.
11606 (nds32_expand_unaligned_store): Declare.
11607 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11608 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11609 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11610 NDS32_BUILTIN_UASTORE_DW.
11611 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11614 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11615 Kito Cheng <kito.cheng@gmail.com>
11616 Chung-Ju Wu <jasonwucj@gmail.com>
11618 * config/nds32/nds32-intrinsic.c
11619 (nds32_expand_builtin_null_ftype_reg): Delete.
11620 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11621 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11622 (nds32_read_argument): New.
11623 (nds32_legitimize_target): Ditto.
11624 (nds32_legitimize_argument): Ditto.
11625 (nds32_check_constant_argument): Ditto.
11626 (nds32_expand_unop_builtin): Ditto.
11627 (nds32_expand_unopimm_builtin): Ditto.
11628 (nds32_expand_binop_builtin): Ditto.
11629 (nds32_builtin_decl_impl): Ditto.
11630 (builtin_description): Ditto.
11631 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11632 (nds32_init_builtins_impl): Ditto.
11633 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11634 (nds32_builtin_decl): New.
11635 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11636 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11638 2018-03-02 Jeff Law <law@redhat.com>
11640 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11641 (redundant_insn, fill_simple_delay_slots): Likewise.
11642 (fill_slots_from_thread): Likewise.
11643 * resource.c (mark_referenced_resources): Likewise.
11644 (mark_set_resources, find_dead_or_set_registers): Likewise.
11646 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11648 * substring-locations.h (format_warning_va): Formatting fix for
11649 ATTRIBUTE_GCC_DIAG.
11650 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11652 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11653 * substring-locations.c: Include intl.h.
11654 (format_warning_va): Turned into small wrapper around
11655 format_warning_n_va, renamed to ...
11656 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11657 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11659 (format_warning_at_substring_n): New function.
11660 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11661 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11662 format_warning_at_substring with just a shorter name instead of
11663 const function pointer.
11664 (fmtwarn_n): New function.
11665 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11666 appropriate, get rid of all the fmtstr temporaries, move conditionals
11667 with G_() wrapped string literals directly into fmtwarn arguments,
11668 cast dir.len to (int), formatting fixes.
11670 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11672 * doc/invoke.texi: Remove "Cilk Plus" references.
11674 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11675 Richard Biener <rguenther@suse.de>
11678 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11679 for error or warning attributes if CALL_FROM_THUNK_P is set.
11682 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11685 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11686 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11689 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11690 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11692 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11693 instead of -1U in last predictors element's probability member.
11695 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11698 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11699 arguments if they are comparable.
11701 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11703 PR tree-optimization/84634
11704 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11705 masks and masked_loop_p with a single loop_masks, making sure it's
11706 null for bb vectorization.
11708 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11710 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11711 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11712 loop->force_vectorize to check whether there is no alias.
11714 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11717 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11719 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11721 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11722 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11723 instead of a loop around prev_real_insn.
11724 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11727 PR inline-asm/84625
11728 * config/i386/i386.c (ix86_print_operand): Use conditional
11729 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11732 2018-03-02 Richard Biener <rguenther@suse.de>
11734 PR tree-optimization/84427
11735 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11736 (bitmap_set_subtract_values): Rewrite to handle multiple
11739 (prune_clobbered_mems): Likewise.
11740 (phi_translate): Take edge instead of pred/phiblock.
11741 (phi_translate_1): Likewise.
11742 (phi_translate_set): Likewise. Insert all translated
11743 exprs for a value into the set, keeping possibly multiple
11744 expressions per value.
11745 (compute_antic_aux): Adjust for phi_translate changes.
11746 When intersecting union the expressions and prune those
11747 not in the final value set, keeping possibly multiple
11748 expressions per value. Do not use value-insertion
11749 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11750 all expressions. Add verification that the value-sets
11751 only shrink during iteration.
11752 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11753 (do_pre_regular_insertion): Likewise.
11754 (do_pre_partial_partial_insertion): Likewise.
11756 2018-03-02 Richard Biener <rguenther@suse.de>
11759 * config/darwin.c (saved_debug_info_level): New static global.
11760 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11761 (darwin_asm_lto_end): Restore debug info generation settings.
11763 2018-03-01 Martin Liska <mliska@suse.cz>
11766 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11767 volatile arguments.
11769 2018-03-01 Richard Biener <rguenther@suse.de>
11772 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11773 type annotation with LTO when debug was disabled at compile-time.
11775 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11777 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11779 (mips_final_postscan_insn): Likewise.
11781 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11783 PR rtl-optimization/84528
11784 * alias.c (init_alias_target): Add commentary.
11785 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11786 a unique base value if the frame pointer is not eliminated
11787 to the stack pointer.
11789 2018-03-01 Tom de Vries <tom@codesourcery.com>
11791 PR rtl-optimization/83327
11792 * lra-int.h (hard_regs_spilled_into): Declare.
11793 * lra.c (hard_regs_spilled_into): Define.
11794 (init_reg_info): Init hard_regs_spilled_into.
11795 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11796 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11797 (process_bb_lives): Handle hard_regs_spilled_into.
11798 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11799 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11801 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11803 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11804 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11805 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11806 * config/rs6000/aix72.h: New file.
11808 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11810 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11811 instead of warning_at with conditional singular and plural messages
11815 * stor-layout.c (update_alignment_for_field): For
11816 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11817 && !DECL_PACKED (field), do the alignment update, just use
11818 only desired_align instead of MAX (type_align, desired_align)
11820 (place_field): Don't do known_align < desired_align handling
11821 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11822 is non-NULL, instead do it after rli->prev_field handling and
11823 only if not within a bitfield word. For DECL_PACKED (field)
11824 use type_align of BITS_PER_UNIT.
11826 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11828 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11829 superfluous parentheses and trailing spaces.
11831 2018-02-28 Richard Biener <rguenther@suse.de>
11833 PR tree-optimization/84584
11834 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11835 SCoPs with fake exit edge.
11837 2018-02-28 Martin Liska <mliska@suse.cz>
11840 * timevar.c (timer::print): Fix format to properly print 100%
11843 2018-02-28 Richard Biener <rguenther@suse.de>
11845 PR middle-end/84607
11846 * genmatch.c (capture_info::walk_match): Do not mark
11847 captured expressions without operands as expr_p given
11848 they act more like predicates and should be subject to
11849 "lost tail" side-effect preserving.
11851 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11853 PR rtl-optimization/81611
11854 * auto-inc-dec.c (attempt_change): Move dead note from
11855 mem_insn if it's the next use of regno
11856 (find_address): Take address use of reg holding
11857 non-incremented value. Add parm to limit search to the named
11859 (merge_in_block): Attempt to use a mem insn that is the next
11860 use of the original regno.
11862 2018-02-27 Martin Sebor <msebor@redhat.com>
11865 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11866 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11868 2018-02-27 Martin Sebor <msebor@redhat.com>
11870 PR translation/84207
11871 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11872 n argument to unsigned HOST_WIDE_INT.
11873 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11874 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11875 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11876 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11878 2018-02-27 Richard Biener <rguenther@suse.de>
11880 PR tree-optimization/84512
11881 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11882 Do not use the estimate returned from record_stmt_cost for
11883 the scalar iteration cost but sum properly using add_stmt_cost.
11885 2018-02-27 Richard Biener <rguenther@suse.de>
11887 PR tree-optimization/84466
11888 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11889 Adjust last change to less strictly validate use operands.
11891 2018-02-27 Martin Liska <mliska@suse.cz>
11893 PR gcov-profile/84548
11894 * gcov.c (process_file): Allow partial overlap and consider it
11895 also as group functions.
11896 (output_lines): Properly calculate range of lines for a group.
11898 2018-02-27 Martin Liska <mliska@suse.cz>
11900 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11901 'ggc' suffixes. Change first column width.
11902 (timer::print): Fix formatting of the column.
11904 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11906 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11907 preserve inline entry blocks for the sake of debug inline
11908 entry point markers alone.
11909 (remove_unused_locals): Suggest in comments a better place to
11910 force the preservation of inline entry blocks that are
11911 otherwise unused, but do not preserve them.
11913 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11915 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11917 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11920 * config/i386/constraints.md (Bs): Replace
11921 ix86_indirect_branch_register with
11922 TARGET_INDIRECT_BRANCH_REGISTER.
11924 * config/i386/i386.md (indirect_jump): Likewise.
11925 (tablejump): Likewise.
11926 (*sibcall_memory): Likewise.
11927 (*sibcall_value_memory): Likewise.
11928 Peepholes of indirect call and jump via memory: Likewise.
11929 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11930 (*sibcall_value_GOT_32): Likewise.
11931 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11932 (GOT_memory_operand): Likewise.
11933 (call_insn_operand): Likewise.
11934 (sibcall_insn_operand): Likewise.
11935 (GOT32_symbol_operand): Likewise.
11936 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11938 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11940 PR rtl-optimization/83496
11941 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11942 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11943 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11944 redundant insn, if any.
11945 (relax_delay_slots): Likewise.
11946 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11948 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11950 PR tree-optimization/83965
11951 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11952 that grouped statements are part of a reduction chain. Return
11953 true if the statement is not marked as a reduction itself but
11954 is part of a group.
11955 (vect_recog_dot_prod_pattern): Don't check whether the statement
11956 is part of a group here.
11957 (vect_recog_sad_pattern): Likewise.
11958 (vect_recog_widen_sum_pattern): Likewise.
11960 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11963 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11965 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11968 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11970 (ix86_output_indirect_function_return): New prototype.
11971 (ix86_split_simple_return_pop_internal): Likewise.
11972 * config/i386/i386.c (indirect_return_via_cx): New.
11973 (indirect_return_via_cx_bnd): Likewise.
11974 (indirect_thunk_name): Handle return va CX_REG.
11975 (output_indirect_thunk_function): Create alias for
11976 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11977 (ix86_output_indirect_jmp): Remove the bool argument.
11978 (ix86_output_indirect_function_return): New function.
11979 (ix86_split_simple_return_pop_internal): Likewise.
11980 * config/i386/i386.md (*indirect_jump): Don't pass false
11981 to ix86_output_indirect_jmp.
11982 (*tablejump_1): Likewise.
11983 (simple_return_pop_internal): Change it to define_insn_and_split.
11984 Call ix86_split_simple_return_pop_internal to split it for
11985 -mfunction-return=.
11986 (simple_return_indirect_internal): Call
11987 ix86_output_indirect_function_return instead of
11988 ix86_output_indirect_jmp.
11990 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11993 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11994 memset and value initialization afterwards.
11996 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11998 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12000 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12003 * common/config/aarch64/aarch64-common.c
12004 (aarch_option_optimization_table[]): Switch
12005 off fomit-frame-pointer
12007 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12008 Chung-Ju Wu <jasonwucj@gmail.com>
12010 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12012 (store_multiple): Ditto.
12014 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12016 * config.gcc: Add --with-cpu support for nds32 target.
12017 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12018 * config/nds32/nds32.opt: Add -mcpu= option.
12020 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12022 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12023 isel=yes): Warn for these deprecated options.
12025 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12027 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12028 ISA_2_5_MASKS_EMBEDDED.
12030 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12032 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12033 p->max as pointers rather than using iterative_hash_expr.
12035 2018-02-23 Carl Love <cel@us.ibm.com>
12037 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12038 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12040 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12041 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12042 P8V_BUILTIN_VEC_VUNSIGNED2.
12044 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12047 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12048 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12050 * lra-constraints.c (curr_insn_transform): Set up
12051 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12053 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12054 * lra-eliminations.c (spill_pseudos): Ditto.
12055 (process_insn_for_elimination): Ditto.
12056 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12057 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12058 LRA_NON_CLOBBERED_ALT.
12060 2018-02-22 Martin Sebor <msebor@redhat.com>
12062 PR tree-optimization/84480
12063 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12064 to maybe_diag_stxncpy_trunc. Call it.
12065 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12066 from gimple_fold_builtin_strcpy. Print inlining stack.
12067 (handle_builtin_stxncpy): Print inlining stack.
12068 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12070 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12073 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12074 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12075 and -fcheck-pointer-bounds are used together.
12076 (indirect_thunk_prefix): New enum.
12077 (indirect_thunk_need_prefix): New function.
12078 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12079 "_nt" instead of "_bnd" for NOTRACK prefix.
12080 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12081 (output_indirect_thunk_function): Likewise.
12083 (ix86_code_end): Update output_indirect_thunk_function calls.
12084 (ix86_output_indirect_branch_via_reg): Replace
12085 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12086 (ix86_output_indirect_branch_via_push): Likewise.
12087 (ix86_output_function_return): Likewise.
12088 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12089 incompatible with -fcf-protection=branch and
12090 -fcheck-pointer-bounds.
12092 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12095 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12096 Change gcc_assert call to output_operand_lossage.
12098 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12100 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12102 2018-02-22 DJ Delorie <dj@redhat.com>
12103 Sebastian Perta <sebastian.perta@renesas.com>
12104 Oleg Endo <olegendo@gcc.gnu.org>
12106 * config/rx/rx.c (rx_rtx_costs): New function.
12107 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12109 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12111 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12113 2018-02-22 Martin Liska <mliska@suse.cz>
12116 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12117 Add "native" as a possible value.
12119 2018-02-22 Martin Liska <mliska@suse.cz>
12122 * config/i386/i386.c (ix86_option_override_internal):
12123 Add "native" as a possible value for -march and -mtune.
12125 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12128 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12129 to all type variants.
12131 PR tree-optimization/84503
12132 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12133 width as info->bitpos + info->bitsize - start.
12134 (merged_store_group::merge_overlapping): Simplify width computation.
12135 (check_no_overlap): New function.
12136 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12137 start + width and last_order of the group, fail if check_no_overlap
12139 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12140 to group if check_no_overlap fails.
12142 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12144 * config/rs6000/altivec.md: Delete contraint arguments to
12145 define_expand, define_split, and define_peephole2, and in
12146 define_insn_and_split if always unused.
12147 * config/rs6000/darwin.md: Ditto.
12148 * config/rs6000/dfp.md: Ditto.
12149 * config/rs6000/rs6000.md: Ditto.
12150 * config/rs6000/sync.md: Ditto.
12151 * config/rs6000/vector.md: Ditto.
12152 * config/rs6000/vsx.md: Ditto.
12154 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12156 * config/rs6000/altivec.md: Write output control strings as braced
12157 blocks instead of double-quoted strings.
12158 * config/rs6000/darwin.md: Ditto.
12159 * config/rs6000/rs6000.md: Ditto.
12160 * config/rs6000/vector.md: Ditto.
12161 * config/rs6000/vsx.md: Ditto.
12163 2018-02-21 Jason Merrill <jason@redhat.com>
12165 PR c++/84314 - ICE with templates and fastcall attribute.
12166 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12168 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12170 * ipa-cp.c (determine_versionability): Fix comment typos.
12172 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12175 * ipa-cp.c (determine_versionability): Do not version functions caling
12178 2018-02-21 Martin Liska <mliska@suse.cz>
12181 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12182 Add "native" as a possible value.
12183 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12184 the macro when native cpu detection is available.
12186 2018-02-21 Martin Liska <mliska@suse.cz>
12189 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12190 Add "native" as a possible value.
12191 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12192 when native cpu detection is available.
12194 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12195 Martin Sebor <msebor@redhat.com>
12197 PR tree-optimization/84478
12198 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12200 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12201 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12202 support which is conservatively correct, for 2 only stay conservative
12203 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12204 argument to the 2 argument get_range_strlen, adjust 6 arg
12205 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12207 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12208 (gimple_fold_builtin_strlen): Pass true as last argument to
12211 2018-02-20 Martin Sebor <msebor@redhat.com>
12213 PR middle-end/84095
12214 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12215 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12216 (builtin_memref::builtin_memref): Factor out parts into
12217 set_base_and_offset and call it.
12219 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12221 PR middle-end/84406
12222 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12223 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12224 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12225 search at the associated MODE_INT.
12227 2018-02-20 Jeff Law <law@redhat.com>
12229 PR middle-end/82123
12230 PR tree-optimization/81592
12231 PR middle-end/79257
12232 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12233 for range data rather than using global data.
12234 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12235 range data rather than using global data.
12236 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12237 pass it to children as needed.
12238 (struct directive::fmtresult): Similarly.
12239 (struct directive::set_width): Similarly.
12240 (struct directive::set_precision): Similarly.
12241 (format_integer, format_directive, parse_directive): Similarly.
12242 (format_none): Accept unnamed vr_values parameter.
12243 (format_percent, format_floating, format_character): Similarly.
12244 (format_string, format_plain): Similarly.
12245 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12246 the EVRP range analyzer for range data rather than using global data.
12247 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12248 gimple-ssa-evrp-analyze.h
12249 (class sprintf_dom_walker): Add after_dom_children member function.
12250 Add evrp_range_analyzer member.
12251 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12252 range analyzer as needed.
12253 (sprintf_dom_walker::after_dom_children): New member function.
12254 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12256 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12257 (evrp_range_analyzer::pop_to_marker): Likewise.
12259 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12261 PR tree-optimization/84419
12262 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12263 with the required type if its current type is compatible but
12266 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12268 PR middle-end/82004
12269 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12270 after vectorization.
12272 2018-02-20 Martin Liska <mliska@suse.cz>
12275 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12276 possible values if we don't have a hint.
12278 2018-02-20 Martin Liska <mliska@suse.cz>
12282 * final.c (shorten_branches): Build align_tab array with one
12284 * opts.c (finish_options): Add alignment option limit check.
12285 (MAX_CODE_ALIGN): Likewise.
12286 (MAX_CODE_ALIGN_VALUE): Likewise.
12287 * doc/invoke.texi: Document maximum allowed option value for
12288 all -falign-* options.
12290 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12293 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12294 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12295 * var-tracking.c (emit_note_insn_var_location): Remove all references
12296 to NOTE_INSN_CALL_ARG_LOCATION.
12297 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12298 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12299 Use copy_rtx_if_shared.
12300 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12301 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12302 (dwarf2out_var_location): Remove handling of
12303 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12305 * final.c (final_scan_insn): Remove all references to
12306 NOTE_INSN_CALL_ARG_LOCATION.
12307 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12308 before dumping final insns.
12309 * except.c (emit_note_eh_region_end): Remove all references to
12310 NOTE_INSN_CALL_ARG_LOCATION.
12311 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12312 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12313 * config/arc/arc.c (hwloop_optimize): Likewise.
12314 * config/arm/arm.c (create_fix_barrier): Likewise.
12315 * config/s390/s390.c (s390_chunkify_start): Likewise.
12316 * config/sh/sh.c (find_barrier): Likewise.
12317 * config/i386/i386.c (rest_of_insert_endbranch,
12318 ix86_seh_fixup_eh_fallthru): Likewise.
12319 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12320 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12321 * config/frv/frv.c (frv_function_prologue): Likewise.
12322 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12324 (note_outside_basic_block_p): Remove all references to
12325 NOTE_INSN_CALL_ARG_LOCATION.
12326 * gengtype.c (adjust_field_rtx_def): Likewise.
12327 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12329 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12330 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12333 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12336 PR tree-optimization/84452
12337 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12338 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12341 2018-02-19 Martin Liska <mliska@suse.cz>
12344 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12346 2018-02-19 Martin Liska <mliska@suse.cz>
12347 Richard Sandiford <richard.sandiford@linaro.org>
12349 PR tree-optimization/82491
12350 * gimple-fold.c (get_base_constructor): Make earlier bail out
12353 2018-02-19 Carl Love <cel@us.ibm.com>
12355 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12356 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12358 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12359 P8V_BUILTIN_VEC_NEG.
12361 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12363 * config/rl78/rl78.md (movdf): New define expand.
12365 2018-02-19 Martin Liska <mliska@suse.cz>
12368 * doc/invoke.texi: Fix typo.
12369 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12371 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12373 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12374 handle rs6000_single_float and rs6000_double_float specially for
12377 2018-02-16 Jeff Law <law@redhat.com>
12379 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12380 (pop_regs): Use it.
12382 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12385 * ipa-inline.c (inline_small_functions): Fix a typo.
12387 2018-02-16 Nathan Sidwell <nathan@acm.org>
12389 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12391 2018-02-16 Carl Love <cel@us.ibm.com>
12393 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12394 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12395 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12396 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12397 expansion to P8V_BUILTIN_VEC_FLOAT2.
12399 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12401 PR rtl-optimization/70023
12402 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12403 src_regno into account.
12405 2018-02-16 Carl Love <cel@us.ibm.com>
12407 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12408 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12409 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12410 * config/rs6000/rs6000.c: Remove case statements for
12411 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12412 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12413 and P9V_BUILTIN_VEC_VINSERT4B.
12414 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12415 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12416 * config/rs6000/vsx.md:
12417 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12420 2018-02-16 Carl Love <cel@us.ibm.com>
12422 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12424 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12426 * config/rs6000/rs6000-c.c: Add the definitions for
12427 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12428 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12429 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12430 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12431 definition for insert4b and define insn *insert3b_internal.
12432 * doc/extend.texi: Add documentation for vec_extract4b.
12434 2018-02-16 Nathan Sidwell <nathan@acm.org>
12436 * doc/extend.texi (Backwards Compatibility): Mention friend
12437 injection. Note for-scope is deprecated.
12438 * doc/invoke.texi (-ffriend-injection): Deprecate.
12440 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12442 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12443 that moved to I2, also allow destinations that are a paradoxical
12444 subreg (instead of a normal reg).
12446 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12449 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12452 2018-02-16 Richard Biener <rguenther@suse.de>
12454 PR tree-optimization/84037
12455 PR tree-optimization/84016
12457 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12458 Adjust vec_construct for the fact we need additional higher latency
12459 128bit inserts for AVX256 and AVX512 vector builds.
12460 (ix86_add_stmt_cost): Scale vector construction cost for
12463 2018-02-16 Richard Biener <rguenther@suse.de>
12465 PR tree-optimization/84417
12466 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12467 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12468 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12470 2018-02-16 Martin Liska <mliska@suse.cz>
12473 * internal-fn.def (ASAN_CHECK): Set proper flags.
12474 (ASAN_MARK): Likewise.
12476 2018-02-16 Julia Koval <julia.koval@intel.com>
12478 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12479 from PTA_CANNONLAKE.
12481 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12484 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12485 Use ++iter rather than iter++ for std::list iterators.
12486 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12487 defer deleting them until all nodes in the forest are processed. Do
12488 free even leaf nodes. Change to_process into auto_vec.
12491 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12492 * vec.h (vec_default_construct): Use memset instead of placement new
12493 if BROKEN_VALUE_INITIALIZATION is defined.
12494 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12495 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12498 PR rtl-optimization/83723
12499 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12500 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12501 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12503 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12505 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12507 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12509 PR rtl-optimization/81443
12510 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12511 from inner REGs to paradoxical SUBREGs.
12513 2018-02-16 Richard Biener <rguenther@suse.de>
12515 PR tree-optimization/84399
12516 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12517 For operands we can analyze at their definition make sure we can
12518 analyze them at each use as well.
12520 2018-02-16 Richard Biener <rguenther@suse.de>
12522 PR tree-optimization/84190
12523 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12524 volatile accesses if the decl isn't volatile.
12526 2018-02-15 Jason Merrill <jason@redhat.com>
12528 PR c++/84314 - ICE with templates and fastcall attribute.
12529 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12530 TYPE_CANONICAL on an existing type.
12532 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12534 PR tree-optimization/84383
12535 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12536 dstoff nor call operand_equal_p if dstbase is NULL.
12538 PR tree-optimization/84334
12539 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12540 also a CONSTANT_CLASS_P, punt.
12542 2018-02-14 Jim Wilson <jimw@sifive.com>
12544 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12545 first SMALL_OPERAND check. New local min_second_step. Move assert
12546 to where locals are set. Add TARGET_RVC support.
12547 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12549 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12551 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12553 2018-02-14 Martin Sebor <msebor@redhat.com>
12555 PR tree-optimization/83698
12556 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12557 arrays constrain the offset range to their bounds.
12558 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12559 (builtin_access::overlap): Avoid setting the size of overlap if it's
12561 (maybe_diag_overlap): Also consider arrays when deciding what values
12562 of offsets to include in diagnostics.
12564 2018-02-14 Martin Sebor <msebor@redhat.com>
12567 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12568 that correspond to the kind of a declaration.
12570 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12573 * config/pa/pa.md: Load address of PIC label using the linkage table
12574 if the label is nonlocal.
12576 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12578 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12579 warning message if user requests -maltivec=be.
12580 * doc/invoke.texi: Document deprecation of -maltivec=be.
12582 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12585 * config/rs6000/rs6000-c.c: Update definitions for
12586 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12587 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12589 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12592 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12593 add _get_ssp intrinsics. Remove argument from
12594 __builtin_ia32_rdssp[d|q].
12595 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12596 * config/i386/i386-builtin.def: Remove argument from
12597 __builtin_ia32_rdssp[d|q].
12598 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12599 ix86_expand_special_args_builtin for _rdssp[d|q].
12600 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12601 Clear register before usage.
12602 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12603 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12605 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12607 PR tree-optimization/84357
12608 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12609 operand 1 of an ARRAY_REF too.
12611 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12614 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12615 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12617 (set_of_reg): New struct.
12618 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12619 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12620 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12622 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12623 Split into bitclr, bitset, bitinvert patterns if appropriate.
12624 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12625 use rx_fuse_in_memory_bitop.
12626 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12627 to named insn, correct maximum insn length.
12629 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12632 * machmode.def: Define a complex mode for PARTIAL_INT.
12633 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12635 * doc/rtl.texi: Document CSPImode.
12636 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12638 (msp430_hard_regno_nregs_with_padding): Likewise.
12640 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12643 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12645 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12647 PR rtl-optimization/84169
12648 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12649 we generated a parallel as new i3 and we split that to new i2 and i3
12650 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12651 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12652 those to i2, not i1. Partially rewrite this scan code.
12654 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12657 * stor-layout.c (place_field): For variable length fields, adjust
12658 offset_align afterwards not just based on the field's alignment,
12659 but also on the size.
12661 PR middle-end/84309
12662 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12663 of exps and logs in the use_exp2 case.
12665 2018-02-13 Jeff Law <law@redhat.com>
12667 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12668 entry for "vector".
12670 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12673 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12677 * common.opt (gas-loc-support, gas-locview-support): New.
12678 (ginline-points, ginternal-reset-location-views): New.
12679 * doc/invoke.texi: Document them. Use @itemx where intended.
12680 (gvariable-location-views): Adjust.
12681 * target.def (reset_location_view): New.
12682 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12683 (TARGET_RESET_LOCATION_VIEW): New.
12684 * doc/tm.texi: Rebuilt.
12685 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12686 (dwarf2out_default_as_locview_support): New.
12687 (output_asm_line_debug_info): Use option variables.
12688 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12689 (output_loc_list): Likewise.
12690 (add_high_low_attributes): Check option variables.
12691 Don't output entry view attribute in strict mode.
12692 (gen_inlined_subroutine_die): Check option variables.
12693 (dwarf2out_inline_entry): Likewise.
12694 (init_sections_and_labels): Likewise.
12695 (dwarf2out_early_finish): Likewise.
12696 (maybe_reset_location_view): New, from...
12697 (dwarf2out_var_location): ... here. Call it.
12698 * debug.h (dwarf2out_default_as_loc_support): Declare.
12699 (dwarf2out_default_as_locview_support): Declare.
12700 * hooks.c (hook_int_rtx_insn_0): New.
12701 * hooks.h (hook_int_rtx_insn_0): Declare.
12702 * toplev.c (process_options): Take -gas-loc-support and
12703 -gas-locview-support from dwarf2out. Enable
12704 -gvariable-location-views by default only with locview
12705 assembler support. Enable -ginternal-reset-location-views by
12706 default only if the target defines the corresponding hook.
12707 Enable -ginline-points by default if location views are
12708 enabled; force it disabled if statement frontiers are
12710 * tree-inline.c (expand_call_inline): Check option variables.
12711 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12713 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12715 PR tree-optimization/84321
12716 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12717 handling. Also check whether the anti-range contains any values
12718 that satisfy the mask; switch to a VR_RANGE if not.
12720 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12723 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12725 2018-02-13 Martin Jambor <mjambor@suse.cz>
12728 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12729 of call statements, also set location of a load to a temporary.
12731 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12733 * config/rl78/rl78.c (add_vector_labels): New function.
12734 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12735 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12736 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12737 which checks that no arguments are passed.
12738 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12739 * doc/extend.texi: Documentation for the new attribute.
12741 2018-02-13 Andreas Schwab <schwab@suse.de>
12743 * config/riscv/linux.h (CPP_SPEC): Define.
12745 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12748 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12749 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12750 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12751 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12752 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12753 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12754 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12755 temporarily for AES and PCLMUL builtins.
12757 PR tree-optimization/84339
12758 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12759 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12762 PR middle-end/84309
12763 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12764 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12765 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12767 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12769 * omp-simd-clone.h: New file.
12770 * omp-simd-clone.c: Include omp-simd-clone.h.
12771 (expand_simd_clones): No longer static.
12772 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12773 cgraph.h and omp-simd-clone.h.
12774 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12775 (vect_recog_widen_shift_pattern): Formatting fix.
12776 (vect_pattern_recog_1): Don't check optab for calls.
12779 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12780 operands[2] into a REG before using gen_lowpart on it.
12782 2018-02-12 Jeff Law <law@redhat.com>
12785 * config/sh/sh.c (find_barrier): Consider a sibling call
12788 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12789 successfully back substituting a reg.
12791 2018-02-12 Richard Biener <rguenther@suse.de>
12793 PR tree-optimization/84037
12794 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12795 parameter, move visited init to caller.
12796 (vect_slp_analyze_operations): Separate cost from validity
12797 check, initialize visited once for all instances.
12798 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12800 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12802 (vect_model_promotion_demotion_cost): Likewise.
12803 (vectorizable_bswap): Guard cost modeling with !slp_node
12804 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12806 (vectorizable_call): Likewise.
12807 (vectorizable_conversion): Likewise.
12808 (vectorizable_assignment): Likewise.
12809 (vectorizable_shift): Likewise.
12810 (vectorizable_operation): Likewise.
12811 (vectorizable_store): Likewise.
12812 (vectorizable_load): Likewise.
12813 (vectorizable_condition): Likewise.
12814 (vectorizable_comparison): Likewise.
12816 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12819 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12820 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12821 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12823 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12825 PR middle-end/83665
12826 * params.def (inline-min-speedup): Increase from 8 to 15.
12827 (max-inline-insns-auto): Decrease from 40 to 30.
12828 * ipa-split.c (consider_split): Add some buffer for function to
12829 be considered inlining candidate.
12830 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12833 2018-02-12 Richard Biener <rguenther@suse.de>
12835 PR tree-optimization/84037
12836 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12837 matched stmts if we cannot swap the non-matched ones.
12839 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12841 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12842 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12843 _mm_maskz_scalef_round_ss): New intrinsics.
12844 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12845 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12846 __builtin_ia32_scalefss_round): Remove.
12847 (__builtin_ia32_scalefsd_mask_round,
12848 __builtin_ia32_scalefss_mask_round): New intrinsics.
12849 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12850 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12851 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12852 "<round_constraint>")): Changed to ...
12853 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12854 "<round_scalar_constraint>")): ... this.
12855 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12856 %0, %1, %2<round_op3>}"): Changed to ...
12857 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12858 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12859 %2<round_scalar_mask_op3>}"): ... this.
12860 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12862 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12864 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12865 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12866 (_mm_maskz_sqrt_round_ss): New intrinsics.
12867 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12868 (__builtin_ia32_sqrtsd_mask_round)
12869 (__builtin_ia32_sqrtss_mask_round): New builtins.
12870 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12871 (__builtin_ia32_sqrtss_round): Remove.
12872 (__builtin_ia32_sqrtsd_mask_round)
12873 (__builtin_ia32_sqrtss_mask_round): New builtins.
12874 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12875 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12876 ((match_operand:VF_128 1 "vector_operand"
12877 "xBm,<round_constraint>")): Changed to ...
12878 ((match_operand:VF_128 1 "vector_operand"
12879 "xBm,<round_scalar_constraint>")): ... this.
12880 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12881 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12882 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12883 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12884 %<iptr>1<round_scalar_mask_op3>}): ... this.
12885 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12886 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12888 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12891 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12892 Cast vec_cmpeq result to correct type.
12893 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12894 Cast vec_cmpgt result to correct type.
12896 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12898 * final.c (final_scan_insn_1): Renamed from...
12899 (final_scan_insn): ... this. New wrapper, to recover
12900 seen from the outermost call in recursive ones.
12901 * config/sparc/sparc.c (output_return): Drop seen from call.
12902 (output_sibcall): Likewise.
12903 * config/visium/visium.c (output_branch): Likewise.
12905 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12907 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12910 2018-02-10 Alan Modra <amodra@gmail.com>
12913 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12914 Specify LR as an input.
12916 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12919 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12920 remove_member_access_dummy_vars): New functions.
12921 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12922 lower_omp_1, execute_lower_omp): Use them.
12924 PR rtl-optimization/84308
12925 * shrink-wrap.c (spread_components): Release todo vector.
12927 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12929 PR rtl-optimization/57193
12930 * ira-color.c (struct allocno_color_data): Add member
12931 conflict_allocno_hard_prefs.
12932 (update_conflict_allocno_hard_prefs): New.
12933 (bucket_allocno_compare_func): Add a preference based on
12934 conflict_allocno_hard_prefs.
12935 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12936 (color_allocnos): Remove a dead code. Initiate
12937 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12939 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12942 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12943 constraint from =wa to wa. Avoid a subreg on the output operand,
12944 instead use a pseudo and subreg it in a move.
12945 (p9_xxbrd_<mode>): Changed to ...
12946 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12947 (p9_xxbrd_v2df): New expander.
12948 (p9_xxbrw_<mode>): Changed to ...
12949 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12950 (p9_xxbrw_v4sf): New expander.
12952 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12954 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12956 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12959 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12960 multiply in 32-bit mode.
12961 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12962 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12965 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12967 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12968 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12969 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12970 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12972 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12975 * dwarf2out.c (is_trivial_indirect_ref): New function.
12976 (dwarf2out_late_global_decl): Do not generate a location
12977 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12978 and that are not defined in the current unit.
12980 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12982 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12983 instead of a libcall for UNORDERED.
12985 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12988 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12989 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12991 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12994 * config/s390/s390.c (s390_set_current_function): Invoke
12995 s390_indirect_branch_settings also if fndecl didn't change.
12997 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12999 * config/rs6000/rs6000.md (blockage): Set length to zero.
13001 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13003 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13005 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13008 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13009 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13013 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13014 PARALLEL incoming that failed vt_get_decl_and_offset check.
13016 PR middle-end/84237
13017 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13018 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13020 (get_variable_section): For decls in named .bss* sections pass true as
13021 second argument to bss_initializer_p.
13023 2018-02-09 Marek Polacek <polacek@redhat.com>
13024 Jakub Jelinek <jakub@redhat.com>
13027 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13028 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13029 Sync some changes from cxx_fold_indirect_ref.
13031 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13033 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13035 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13036 (BLOCK_INLINE_ENTRY_LABEL): New.
13037 (dwarf2out_var_location): Disregard inline entry markers.
13038 (inline_entry_data): New struct.
13039 (inline_entry_data_hasher): New hashtable type.
13040 (inline_entry_data_hasher::hash): New.
13041 (inline_entry_data_hasher::equal): New.
13042 (inline_entry_data_table): New variable.
13043 (add_high_low_attributes): Add DW_AT_entry_pc and
13044 DW_AT_GNU_entry_view attributes if a pending entry is found
13045 in inline_entry_data_table. Add old entry_pc attribute only
13046 if debug nonbinding markers are disabled.
13047 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13048 markers are enabled.
13049 (block_within_block_p, dwarf2out_inline_entry): New.
13050 (dwarf2out_finish): Check that no entries remained in
13051 inline_entry_data_table.
13052 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13053 (final_scan_insn, notice_source_line): Likewise.
13054 (rest_of_clean_state): Skip inline entry markers.
13055 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13057 * gimple.c (gimple_build_debug_inline_entry): New.
13058 * gimple.h (enum gimple_debug_subcode): Add
13059 GIMPLE_DEBUG_INLINE_ENTRY.
13060 (gimple_build_debug_inline_entry): Declare.
13061 (gimple_debug_inline_entry_p): New.
13062 (gimple_debug_nonbind_marker_p): Adjust.
13063 * insn-notes.def (INLINE_ENTRY): New.
13064 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13065 inline entry marker notes.
13066 (print_insn): Likewise.
13067 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13068 (INSN_DEBUG_MARKER_KIND): Likewise.
13069 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13070 * tree-inline.c (expand_call_inline): Build and insert
13071 debug_inline_entry stmt.
13072 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13073 inline entry blocks early, if nonbind markers are enabled.
13074 (dump_scope_block): Dump fragment info.
13075 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13076 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13077 (gimple_build_debug_inline_entry): New.
13078 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13079 Enable/disable inline entry points too.
13080 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13081 (DEBUG_INSN): Describe inline entry markers.
13083 * common.opt (gvariable-location-views): New.
13084 (gvariable-location-views=incompat5): New.
13085 * config.in: Rebuilt.
13086 * configure: Rebuilt.
13087 * configure.ac: Test assembler for view support.
13088 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13089 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13090 * dwarf2out.c (var_loc_view): New typedef.
13091 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13092 (dwarf2out_locviews_in_attribute): New.
13093 (dwarf2out_locviews_in_loclist): New.
13094 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13095 (enum dw_line_info_opcode): Add LI_adv_address.
13096 (struct dw_line_info_table): Add view.
13097 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13098 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13099 (zero_view_p): New variable.
13100 (ZERO_VIEW_P): New macro.
13101 (output_asm_line_debug_info): New.
13102 (struct var_loc_node): Add view.
13103 (add_AT_view_list, AT_loc_list): New.
13104 (add_var_loc_to_decl): Add view param. Test it against last.
13105 (new_loc_list): Add view params. Record them.
13106 (AT_loc_list_ptr): Handle loc and view lists.
13107 (view_list_to_loc_list_val_node): New.
13108 (print_dw_val): Handle dw_val_class_view_list.
13109 (size_of_die): Likewise.
13110 (value_format): Likewise.
13111 (loc_list_has_views): New.
13112 (gen_llsym): Set vl_symbol too.
13113 (maybe_gen_llsym, skip_loc_list_entry): New.
13114 (dwarf2out_maybe_output_loclist_view_pair): New.
13115 (output_loc_list): Output view list or entries too.
13116 (output_view_list_offset): New.
13117 (output_die): Handle dw_val_class_view_list.
13118 (output_dwarf_version): New.
13119 (output_compilation_unit_header): Use it.
13120 (output_skeleton_debug_sections): Likewise.
13121 (output_rnglists, output_line_info): Likewise.
13122 (output_pubnames, output_aranges): Update version comments.
13123 (output_one_line_info_table): Output view numbers in asm comments.
13124 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13125 Call maybe_gen_llsym.
13126 (loc_list_from_tree_1): Adjust.
13127 (add_AT_location_description): Create view list attribute if
13128 needed, check it's absent otherwise.
13129 (convert_cfa_to_fb_loc_list): Adjust.
13130 (maybe_emit_file): Call output_asm_line_debug_info for test.
13131 (dwarf2out_var_location): Reset views as needed. Precompute
13132 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13133 attribute. Set view.
13134 (new_line_info_table): Reset next view.
13135 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13136 (dwarf2out_source_line): Likewise. Output view resets and labels to
13137 the assembler, or select appropriate line info opcodes.
13138 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13139 (optimize_string_length): Catch it. Adjust.
13140 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13141 dw_val_class_view_list, and remove it if no longer needed.
13142 (hash_loc_list): Hash view numbers.
13143 (loc_list_hasher::equal): Compare them.
13144 (optimize_location_lists): Check whether a view list symbol is
13145 needed, and whether the locview attribute is present, and
13146 whether they match. Remove the locview attribute if no longer
13148 (index_location_lists): Call skip_loc_list_entry for test.
13149 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13150 Use output_dwarf_version.
13151 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13152 (struct dw_val_node): Add val_view_list.
13153 * final.c (SEEN_NEXT_VIEW): New.
13154 (set_next_view_needed): New.
13155 (clear_next_view_needed): New.
13156 (maybe_output_next_view): New.
13157 (final_start_function): Rename to...
13158 (final_start_function_1): ... this. Take pointer to FIRST,
13159 add SEEN parameter. Emit param bindings in the initial view.
13160 (final_start_function): Reintroduce SEEN-less interface.
13161 (final): Rename to...
13162 (final_1): ... this. Take SEEN parameter. Output final pending
13163 next view at the end.
13164 (final): Reintroduce seen-less interface.
13165 (final_scan_insn): Output pending next view before switching
13166 sections or ending a block. Mark the next view as needed when
13167 outputting variable locations. Notify debug backend of section
13168 changes, and of location view changes.
13169 (rest_of_handle_final): Adjust.
13170 * toplev.c (process_options): Autodetect value for debug variable
13171 location views option. Warn on incompat5 without -gdwarf-5.
13172 * doc/invoke.texi (gvariable-location-views): New.
13173 (gvariable-location-views=incompat5): New.
13174 (gno-variable-location-views): New.
13176 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13178 PR tree-optimization/84136
13179 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13180 that the result of find_edge is non-NULL.
13182 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13185 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13186 storing integer register in SImode. Fix cost of 256 and 512
13187 byte aligned SSE register store.
13189 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13191 * config/i386/i386.c (ix86_multiplication_cost): Fix
13192 multiplication cost for TARGET_AVX512DQ.
13194 2018-02-08 Marek Polacek <polacek@redhat.com>
13196 PR tree-optimization/84238
13197 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13200 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13202 PR tree-optimization/84265
13203 * tree-vect-stmts.c (vectorizable_store): Don't treat
13204 VMAT_CONTIGUOUS accesses as grouped.
13205 (vectorizable_load): Likewise.
13207 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13209 PR tree-optimization/81635
13210 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13211 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13212 (test_round_for_mask): New functions.
13213 (wide_int_cc_tests): Call test_round_for_mask.
13214 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13215 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13216 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13217 range returned by get_range_info.
13219 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13222 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13223 * symtab.c: Include builtins.h
13224 (symtab_node::output_to_lto_symbol_table_p): Move here
13225 from lto-streamer-out.c:output_symbol_p.
13226 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13227 (output_symbol_p): Move all logic to symtab.c
13228 (produce_symtab): Update.
13230 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13232 * config/s390/s390-opts.h (enum indirect_branch): Define.
13233 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13234 (s390_indirect_branch_via_thunk)
13235 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13236 (enum s390_indirect_branch_type): Define.
13237 * config/s390/s390.c (struct s390_frame_layout, struct
13238 machine_function): Remove.
13239 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13240 (indirect_branch_table_label_no, indirect_branch_table_name):
13242 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13243 (enum s390_indirect_branch_option): Define.
13244 (s390_return_addr_from_memory): New function.
13245 (s390_handle_string_attribute): New function.
13246 (s390_attribute_table): Add new attribute handler.
13247 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13248 (s390_indirect_branch_via_thunk): New function.
13249 (s390_indirect_branch_via_inline_thunk): New function.
13250 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13251 sibling call optimization for non z10 compiles.
13252 (s390_emit_call): Force indirect branch target to be a single
13253 register. Add r1 clobber for non-z10 compiles.
13254 (s390_emit_epilogue): Emit return jump via return_use expander.
13255 (s390_reorg): Handle JUMP_INSNs as execute targets.
13256 (s390_option_override_internal): Perform validity checks for the
13257 new command line options.
13258 (s390_indirect_branch_attrvalue): New function.
13259 (s390_indirect_branch_settings): New function.
13260 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13261 (s390_output_indirect_thunk_function): New function.
13262 (s390_code_end): Implement target hook.
13263 (s390_case_values_threshold): Implement target hook.
13264 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13266 * config/s390/s390.h (struct s390_frame_layout)
13267 (struct machine_function): Move here from s390.c.
13268 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13269 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13270 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13271 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13272 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13273 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13274 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13275 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13276 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13277 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13278 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13279 (mnemonic attribute): Add values which aren't recognized
13281 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13282 pattern for branch conversion. Fix mnemonic attribute.
13283 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13284 indirect branch via thunk if requested.
13285 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13286 ("*indirect_jump"): Disable for branch conversion using out of
13288 ("indirect_jump_via_thunk<mode>_z10")
13289 ("indirect_jump_via_thunk<mode>")
13290 ("indirect_jump_via_inlinethunk<mode>_z10")
13291 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13292 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13293 ("casesi_jump_via_inlinethunk<mode>_z10")
13294 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13295 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13296 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13297 ("*indirect2_jump"): Disable for branch conversion.
13298 ("casesi_jump"): Turn into expander and expand patterns for branch
13300 ("return_use"): New expander.
13301 ("*return"): Emit return via thunk and rename it to ...
13302 ("*return<mode>"): ... this one.
13303 * config/s390/s390.opt: Add new options and and enum for the
13306 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13308 * lra-constraints.c (match_reload): Unconditionally use
13309 gen_lowpart_SUBREG, rather than selecting between that
13310 and equivalent gen_rtx_SUBREG code.
13312 2018-02-08 Richard Biener <rguenther@suse.de>
13314 PR tree-optimization/84233
13315 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13316 changed flag instead of boguously re-using phi_inserted.
13318 2018-02-08 Martin Jambor <mjambor@suse.cz>
13320 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13321 static local variables.
13323 2018-02-08 Richard Biener <rguenther@suse.de>
13325 PR tree-optimization/84278
13326 * tree-vect-stmts.c (vectorizable_store): When looking for
13327 smaller vector types to perform grouped strided loads/stores
13328 make sure the mode is supported by the target.
13329 (vectorizable_load): Likewise.
13331 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13333 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13334 Increase LDP/STP opportunities by adding adjacent callee-saves.
13336 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13338 PR rtl-optimization/84068
13339 PR rtl-optimization/83459
13340 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13342 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13344 PR tree-optimization/84224
13345 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13346 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13347 non-zero arguments.
13349 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13352 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13353 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13354 count, remove one USE.
13356 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13358 * doc/install.texi (Configuration): Document the
13359 --with-long-double-format={ibm,ieee} PowerPC configuration
13363 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13364 Convert from define_expand to be define_insn_and_split. Rework
13365 float/double/_Float128 conversions to QI/HI/SImode to work with
13366 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13367 conversions to QI/HImode types did a store and then a load to
13368 truncate the value. For conversions to VSX registers, don't split
13369 the insn, instead emit the code directly. Use the code iterator
13370 any_fix to combine signed and unsigned conversions.
13371 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13372 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13373 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13374 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13375 (fix_<mode>di2_hw): Likewise.
13376 (fixuns_<mode>di2_hw): Likewise.
13377 (fix_<mode>si2_hw): Likewise.
13378 (fixuns_<mode>si2_hw): Likewise.
13379 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13380 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13381 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13382 fix<uns>_trunc<SFDF:mode>si2_p8.
13383 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13384 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13385 (fix<uns>_<mode>_mem): Likewise.
13386 (fctiw<u>z_<mode>_mem): Likewise.
13387 (fix<uns>_<mode>_mem): Likewise.
13388 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13389 the register allocator from doing a direct move to the GPRs to do
13390 a store, and instead use the ISA 3.0 store byte/half-word from
13391 vector register instruction. For IEEE 128-bit floating point,
13392 also optimize stores of 32-bit ints.
13393 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13395 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13397 * genextract.c (push_pathstr_operand): New function to support
13399 (walk_rtx): Call push_pathstr_operand.
13400 (print_path): Support [a-zA-Z].
13402 2018-02-07 Richard Biener <rguenther@suse.de>
13404 PR tree-optimization/84037
13405 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13406 (cse_and_gimplify_to_preheader): Declare.
13407 (vect_get_place_in_interleaving_chain): Likewise.
13408 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13410 (_loop_vec_info::~_loop_vec_info): Delete it.
13411 (cse_and_gimplify_to_preheader): New function.
13412 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13413 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13414 (vectorizable_load): Likewise. For grouped stores always base
13415 the IV on the first element.
13416 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13417 condition before gimplifying.
13419 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13421 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13422 *DIV_EXPR and *MOD_EXPR.
13424 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13427 * config/i386/i386.c (ix86_option_override_internal): Mask out
13428 the CF_SET bit when checking -fcf-protection.
13430 2018-02-07 Tom de Vries <tom@codesourcery.com>
13433 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13436 2018-02-07 Richard Biener <rguenther@suse.de>
13438 PR tree-optimization/84204
13439 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13442 PR tree-optimization/84205
13443 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13444 special-case isl_ast_op_zdiv_r.
13446 PR tree-optimization/84223
13447 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13448 Only add conditions from within the region.
13449 (gather_bbs::after_dom_children): Adjust.
13451 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13454 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13455 * config/avr/avr.md: Only post-reload split REG-REG moves if
13456 either register is GENERAL_REG_P.
13458 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13460 PR tree-optimization/84235
13461 * tree-ssa-scopedtables.c
13462 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13463 if the subtraction is performed in floating point type where NaNs are
13464 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13465 build 1. Formatting fix.
13467 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13470 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13471 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13472 and skip it regardless of bb boundaries. Use CALL_P macro,
13473 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13474 unnecessarily, formatting fix.
13476 2018-02-06 Michael Collison <michael.collison@arm.com>
13478 * config/arm/thumb2.md:
13479 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13480 (*thumb_mov_notscc): Ditto.
13482 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13485 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13486 unsigned_fix, not "s".
13488 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13490 * configure.ac (gcc_fn_eh_frame_ro): New function.
13491 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13492 correct .eh_frame permissions.
13493 * configure: Regenerate.
13495 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13497 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13498 irrelevant options.
13500 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13502 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13503 Display warning message for -mno-speculate-indirect-jumps.
13505 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13507 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13509 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13511 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13513 PR tree-optimization/84225
13514 * tree-eh.c (find_trapping_overflow): Only call
13515 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13517 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13520 * config/i386/i386.c: Reimplement the check of possible options
13521 -mibt/-mshstk conbination. Change error messages.
13522 * doc/invoke.texi: Fix a typo: remove extra '='.
13524 2018-02-06 Marek Polacek <polacek@redhat.com>
13526 PR tree-optimization/84228
13527 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13529 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13532 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13533 emitted arch directives.
13534 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13535 __ARM_FEATURE_COPROC before changing architectures.
13537 2018-02-06 Richard Biener <rguenther@suse.de>
13539 * config/i386/i386.c (print_reg): Fix typo.
13540 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13542 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13544 * configure: Regenerate.
13546 2018-02-05 Martin Sebor <msebor@redhat.com>
13548 PR tree-optimization/83369
13549 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13552 2018-02-05 Martin Liska <mliska@suse.cz>
13554 * doc/invoke.texi: Cherry-pick upstream r323995.
13556 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13558 * ira.c (ira_init_register_move_cost): Adjust comment.
13560 2018-02-05 Martin Liska <mliska@suse.cz>
13562 PR gcov-profile/84137
13563 * doc/gcov.texi: Fix typo in documentation.
13565 2018-02-05 Martin Liska <mliska@suse.cz>
13567 PR gcov-profile/83879
13568 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13569 using dlopen functionality.
13571 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13573 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13574 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13575 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13576 _mm_maskz_range_round_ss): New intrinsics.
13577 (__builtin_ia32_rangesd128_round)
13578 (__builtin_ia32_rangess128_round): Remove.
13579 (__builtin_ia32_rangesd128_mask_round,
13580 __builtin_ia32_rangess128_mask_round): New builtins.
13581 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13582 __builtin_ia32_rangess128_round): Remove.
13583 (__builtin_ia32_rangesd128_mask_round,
13584 __builtin_ia32_rangess128_mask_round): New builtins.
13585 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13586 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13587 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13588 "<round_saeonly_constraint>")): Changed to ...
13589 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13590 "<round_saeonly_scalar_constraint>")): ... this.
13591 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13592 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13593 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13594 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13595 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13597 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13599 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13601 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13602 Remove all values except native, 8540 and 8548.
13604 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13606 * config/i386/i386.c (ix86_output_function_return): Pass
13607 INVALID_REGNUM, instead of -1, as invalid register number to
13608 indirect_thunk_name and output_indirect_thunk.
13610 2018-02-02 Julia Koval <julia.koval@intel.com>
13612 * config.gcc: Add -march=icelake.
13613 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13614 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13615 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13616 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13617 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13618 (processor_target_table): Add icelake.
13619 (ix86_option_override_internal): Handle new PTAs.
13620 (get_builtin_code_for_version): Handle icelake.
13621 (M_INTEL_COREI7_ICELAKE): New.
13622 (fold_builtin_cpu): Handle icelake.
13623 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13624 * doc/invoke.texi: Add -march=icelake.
13626 2018-02-02 Julia Koval <julia.koval@intel.com>
13628 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13629 to wide_int_bitmask.
13630 * wide-int-bitmask.h: New.
13632 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13635 * config/i386/i386.md: Replace Pmode with word_mode in
13636 builtin_setjmp_setup and builtin_longjmp to support x32.
13638 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13642 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13644 (rs6000_supported_cpu_names): New static variable.
13645 (linux_cpu_translation_table): Likewise.
13646 (elf_platform) <cpu>: Define new static variable and use it.
13647 Translate kernel AT_PLATFORM name to canonical name if needed.
13648 Error if platform name is unknown.
13650 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13653 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13655 2018-02-01 Jeff Law <law@redhat.com>
13658 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13659 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13660 the scratch if RELEASE_VIA_POP is false.
13661 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13662 If we have to save a temporary register, decrement SIZE appropriately.
13663 Pass new arguments to release_scratch_register_on_entry.
13664 (ix86_adjust_stack_and_probe): Likewise.
13665 (ix86_emit_probe_stack_range): Pass new arguments to
13666 release_scratch_register_on_entry.
13668 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13670 PR rtl-optimization/84157
13671 * combine.c (change_zero_ext): Use REG_P predicate in
13672 front of HARD_REGISTER_P predicate.
13674 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13676 * config/avr/avr.c (avr_option_override): Move disabling of
13677 -fdelete-null-pointer-checks to...
13678 * common/config/avr/avr-common.c (avr_option_optimization_table):
13681 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13683 PR tree-optimization/81635
13684 * tree-data-ref.c (split_constant_offset_1): For types that
13685 wrap on overflow, try to use range info to prove that wrapping
13688 2018-02-01 Renlin Li <renlin.li@arm.com>
13691 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13692 TAILCALL_ADDR_REGS.
13693 (aarch64_register_move_cost): Likewise.
13694 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13695 TAILCALL_ADDR_REGS.
13696 (REG_CLASS_NAMES): Likewise.
13697 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13698 TAILCALL_ADDR_REGS. Remove IP registers.
13699 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13701 2018-02-01 Richard Biener <rguenther@suse.de>
13703 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13704 for specifying RPO order and allow NULL for that.
13705 * domwalk.c (dom_walker::dom_walker): Likewise.
13706 (dom_walker::walk): Handle NULL RPO order.
13707 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13709 (rewrite_update_dom_walker): Likewise.
13710 (mark_def_dom_walker): Likewise.
13712 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13714 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13715 (aarch64_maybe_expand_sve_subreg_move): Declare.
13716 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13717 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13719 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13720 that are semantically a reverse operation.
13721 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13722 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13723 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13725 (aarch64_can_change_mode_class): For big-endian, forbid changes
13726 between two SVE modes if they have different element sizes.
13728 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13730 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13731 the TImode handling for big-endian targets.
13733 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13735 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13736 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13738 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13739 Remove BSWAP handing for big-endian targets and use the form of
13740 LD1RQ appropariate for the mode.
13742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13744 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13745 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13746 duplicated element.
13748 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13751 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13752 check for operands that need to go through aarch64_sve_reload_be.
13754 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13756 PR tree-optimization/81661
13757 PR tree-optimization/84117
13758 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13759 * tree-eh.c: Include gimplify.h.
13760 (find_trapping_overflow, replace_trapping_overflow,
13761 rewrite_to_non_trapping_overflow): New functions.
13762 * tree-vect-loop.c: Include tree-eh.h.
13763 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13764 * tree-data-ref.c: Include tree-eh.h.
13765 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13767 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13769 PR rtl-optimization/84123
13770 * combine.c (change_zero_ext): Check if hard register satisfies
13771 can_change_dest_mode before calling gen_lowpart_SUBREG.
13773 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13776 * ira.c (ira_init_register_move_cost): Remove assert.
13778 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13780 PR rtl-optimization/84071
13781 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13782 * doc/tm.texi: Regenerate.
13784 2018-01-31 Richard Biener <rguenther@suse.de>
13786 PR tree-optimization/84132
13787 * tree-data-ref.c (analyze_miv_subscript): Properly
13788 check whether evolution_function_is_affine_multivariate_p
13789 before calling gcd_of_steps_may_divide_p.
13791 2018-01-31 Julia Koval <julia.koval@intel.com>
13794 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13795 * config/i386/i386.md (rdpid_rex64) New.
13796 (rdpid): Make 32bit only.
13798 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13801 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13802 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13804 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13807 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13809 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13811 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13813 PR rtl-optimization/84071
13814 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13815 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13817 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13819 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13820 (arc_attribute_table): Add 'aux' attribute.
13821 (arc_in_small_data_p): Consider aux like variables.
13822 (arc_is_aux_reg_p): New function.
13823 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13824 (arc_get_aux_arg): New function.
13825 (prepare_move_operands): Handle aux-register access.
13826 (arc_handle_aux_attribute): New function.
13827 * doc/extend.texi (ARC Variable attributes): Add subsection.
13829 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13831 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13832 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13833 (arc_attribute_table): Add 'uncached' attribute.
13834 (arc_print_operand): Print '.di' flag for uncached memory
13836 (arc_in_small_data_p): Do not consider for small data the uncached
13838 (arc_is_uncached_mem_p): New function.
13839 * config/arc/predicates.md (compact_store_memory_operand): Check
13840 for uncached memory accesses.
13841 (nonvol_nonimm_operand): Likewise.
13842 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13844 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13847 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13848 falign-loops=): Add Optimization flag.
13850 2018-01-30 Jeff Law <law@redhat.com>
13853 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13854 INT_REGISTERS_SAVED. Check it prior to calling
13855 get_scratch_register_on_entry.
13856 (ix86_adjust_stack_and_probe): Similarly.
13857 (ix86_emit_probe_stack_range): Similarly.
13858 (ix86_expand_prologue): Corresponding changes.
13860 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13863 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13864 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13866 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13869 * lra-constraints.c (curr_insn_transform): Process AND in the
13872 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13874 PR rtl-optimization/83986
13875 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13876 dependence against last_pending_memory_flush in addition to
13877 pending_jump_insns.
13879 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13881 PR tree-optimization/81611
13882 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13885 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13888 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13891 2018-01-30 Richard Biener <rguenther@suse.de>
13892 Jakub Jelinek <jakub@redhat.com>
13894 PR tree-optimization/84111
13895 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13896 inner loops added during recursion, as they don't have up-to-date
13899 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13902 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13903 (can_inline_edge_by_limits_p): ... here.
13904 (can_early_inline_edge_p, check_callers,
13905 update_caller_keys, update_callee_keys, recursive_inlining,
13906 add_new_edges_to_heap, speculation_useful_p,
13907 inline_small_functions,
13908 inline_small_functions, flatten_function,
13909 inline_to_all_callers_1): Update.
13911 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13913 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13916 2018-01-30 Richard Biener <rguenther@suse.de>
13918 PR tree-optimization/83008
13919 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13920 invariant and constant vector uses in stmts when they need
13921 more than one stmt.
13923 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13926 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13927 * configure: Regenerate.
13929 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13931 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13933 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13934 Use gen_rtx_REG rather than gen_lowpart.
13936 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13938 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13939 rather than 0 when creating partial subregs.
13941 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13943 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13946 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13949 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13950 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13951 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13952 flags. This restores the settings used before the 2017-07-24.
13953 Turning off pre increment/decrement/modify allows IVOPTS to
13954 optimize DF/SF loops where the index is an int.
13956 2018-01-29 Richard Biener <rguenther@suse.de>
13957 Kelvin Nilsen <kelvin@gcc.gnu.org>
13960 * tree-vect-stmts.c (vectorizable_call): Don't call
13961 targetm.vectorize_builtin_md_vectorized_function if callee is
13964 2018-01-22 Carl Love <cel@us.ibm.com>
13966 * doc/extend.tex: Fix typo in second arg in
13967 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13969 2018-01-29 Richard Biener <rguenther@suse.de>
13971 PR tree-optimization/84086
13972 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13973 (flush_ssaname_freelist): When SSA names were released reset
13974 the SCEV hash table.
13976 2018-01-29 Richard Biener <rguenther@suse.de>
13978 PR tree-optimization/84057
13979 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13980 removed paths when removing edges.
13982 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13984 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13985 -mfunction-return=@var{choice}.
13987 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13989 PR diagnostic/84034
13990 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13991 Handle CR like TAB.
13992 (layout::print_source_line): Likewise.
13993 (test_get_line_width_without_trailing_whitespace): Add test cases.
13995 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13997 PR middle-end/84040
13998 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14001 2018-01-26 Jim Wilson <jimw@sifive.com>
14003 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14005 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14008 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14010 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14011 and CMP + SUB-immediate -> SUBS.
14013 2018-01-26 Martin Sebor <msebor@redhat.com>
14015 PR tree-optimization/83896
14016 * tree-ssa-strlen.c (get_string_len): Rename...
14017 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14018 Avoid assuming length is constant.
14019 (handle_char_store): Use HOST_WIDE_INT for string length.
14021 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14024 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14025 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14027 2018-01-26 Richard Biener <rguenther@suse.de>
14029 PR rtl-optimization/84003
14030 * dse.c (record_store): Only record redundant stores when
14031 the earlier store aliases at least all accesses the later one does.
14033 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14035 PR rtl-optimization/83985
14036 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14037 REG_CFA_RESTORE insns.
14038 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14039 REG_CFA_RESTORE insns here.
14042 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14043 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14045 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14047 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14048 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14049 (arc_init): Likewise.
14050 (arc_override_options): Likewise.
14051 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14053 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14055 * config/arc/arc.h (TARGET_DBNZ): Define.
14056 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14057 properly set the tune attribute.
14058 (dbnz): Use TARGET_DBNZ guard.
14059 * config/arc/arc.opt (mtune): Add core3 option.
14061 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14063 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14064 recognize new pic like addresses.
14065 (arc_delegitimize_address): Clean up.
14067 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14069 * config/arc/arc-arches.def: Option mrf16 valid for all
14071 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14072 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14073 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14074 * config/arc/arc-tables.opt: Regenerate.
14075 * config/arc/arc.c (arc_conditional_register_usage): Handle
14076 reduced register file case.
14077 (arc_file_start): Set must have build attributes.
14078 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14079 mrf16 option value.
14080 * config/arc/arc.opt (mrf16): Add new option.
14081 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14082 * config/arc/genmultilib.awk: Handle new mrf16 option.
14083 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14084 * config/arc/t-multilib: Regenerate.
14085 * doc/invoke.texi (ARC Options): Document mrf16 option.
14087 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14089 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14090 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14091 (arc_attribute_table): Add 'secure_call' attribute.
14092 (arc_print_operand): Print secure call operand.
14093 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14095 (arc_is_secure_call_p): New function. * config/arc/arc.md
14096 (call_i): Add support for sjli instruction.
14097 (call_value_i): Likewise.
14098 * config/arc/constraints.md (Csc): New constraint.
14100 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14101 John Eric Martin <John.Martin@emmicro-us.com>
14103 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14104 * config/arc/arc.c (_arc_jli_section): New struct.
14105 (arc_jli_section): New type.
14106 (rc_jli_sections): New static variable.
14107 (arc_handle_jli_attribute): New function.
14108 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14109 (arc_file_end): New function.
14110 (TARGET_ASM_FILE_END): Define.
14111 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14112 (arc_add_jli_section): New function.
14113 (jli_call_scan): Likewise.
14114 (arc_reorg): Call jli_call_scan.
14115 (arc_output_addsi): Remove 'S' from printing asm operand.
14116 (arc_is_jli_call_p): New function.
14117 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14119 (movhi_insn): Likewise.
14120 (movsi_insn): Likewise.
14121 (movsi_set_cc_insn): Likewise.
14122 (loadqi_update): Likewise.
14123 (load_zeroextendqisi_update): Likewise.
14124 (load_signextendqisi_update): Likewise.
14125 (loadhi_update): Likewise.
14126 (load_zeroextendhisi_update): Likewise.
14127 (load_signextendhisi_update): Likewise.
14128 (loadsi_update): Likewise.
14129 (loadsf_update): Likewise.
14130 (movsicc_insn): Likewise.
14131 (bset_insn): Likewise.
14132 (bxor_insn): Likewise.
14133 (bclr_insn): Likewise.
14134 (bmsk_insn): Likewise.
14135 (bicsi3_insn): Likewise.
14136 (cmpsi_cc_c_insn): Likewise.
14137 (movsi_ne): Likewise.
14138 (movsi_cond_exec): Likewise.
14139 (clrsbsi2): Likewise.
14140 (norm_f): Likewise.
14150 (call_i): Remove 'S' asm letter, add jli instruction.
14151 (call_value_i): Likewise.
14152 * config/arc/arc.op (mjli-always): New option.
14153 * config/arc/constraints.md (Cji): New constraint.
14154 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14156 (subsf3_fpx): Likewise.
14157 (mulsf3_fpx): Likewise.
14158 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14160 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14161 function attrbutes.
14162 * doc/invoke.texi (ARC): Document mjli-always option.
14164 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14166 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14167 avoid addition with 0 and use incw and decw where possible.
14169 2018-01-26 Richard Biener <rguenther@suse.de>
14171 PR tree-optimization/81082
14172 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14173 association if it requires casting to unsigned.
14174 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14175 from fold_plusminus_mult_expr to catch important cases late when
14176 range info is available.
14178 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14180 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14181 * configure.ac (hidden_linkonce): New test.
14182 * configure: Regenerate.
14183 * config.in: Regenerate.
14185 2018-01-26 Julia Koval <julia.koval@intel.com>
14187 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14188 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14189 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14190 _mm_mask_bitshuffle_epi64_mask): Fix type.
14191 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14192 USI_FTYPE_V4DI_V4DI_USI): Remove.
14193 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14194 __builtin_ia32_vpshufbitqmb256_mask,
14195 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14196 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14197 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14199 2018-01-26 Alan Modra <amodra@gmail.com>
14202 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14203 UNSPEC_VBPERMQ. Sort other unspecs.
14205 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14207 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14209 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14211 PR middle-end/83055
14212 * predict.c (drop_profile): Do not push/pop cfun; update also
14214 (handle_missing_profiles): Fix logic looking for zero profiles.
14216 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14218 PR middle-end/83977
14219 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14220 on functions with #pragma omp declare simd or functions with simd
14222 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14223 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14224 Remove trailing \n from warning_at calls.
14226 2018-01-25 Tom de Vries <tom@codesourcery.com>
14229 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14230 for neutered workers.
14232 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14235 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14236 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14238 2018-01-24 Jeff Law <law@redhat.com>
14241 * i386.c (get_probe_interval): Move to earlier point.
14242 (ix86_compute_frame_layout): If -fstack-clash-protection and
14243 the frame is larger than the probe interval, then use pushes
14244 to save registers rather than reg->mem moves.
14245 (ix86_expand_prologue): Remove conditional for int_registers_saved
14248 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14251 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14252 min/max for never referenced object.
14254 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14256 PR middle-end/83977
14257 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14259 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14260 attributes from DECL_ATTRIBUTES (decl) without affecting
14261 DECL_ATTRIBUTES (current_function_decl).
14262 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14263 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14265 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14267 PR tree-optimization/83979
14268 * fold-const.c (fold_comparison): Use constant_boolean_node
14269 instead of boolean_{true,false}_node.
14271 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14273 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14276 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14278 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14279 Simplify the clause that sets the length attribute.
14280 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14281 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14282 clause that sets the length attribute.
14283 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14285 2018-01-24 Tom de Vries <tom@codesourcery.com>
14288 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14289 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14290 Add strict parameter.
14291 (prevent_branch_around_nothing): Insert dummy insn between branch to
14292 label and label with no ptx insn inbetween.
14293 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14295 2018-01-24 Tom de Vries <tom@codesourcery.com>
14298 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14299 for neutered threads in warp.
14300 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14302 2018-01-24 Richard Biener <rguenther@suse.de>
14304 PR tree-optimization/83176
14305 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14308 2018-01-24 Richard Biener <rguenther@suse.de>
14310 PR tree-optimization/82819
14311 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14312 code generating pluses that are no-ops in the target precision.
14314 2018-01-24 Richard Biener <rguenther@suse.de>
14316 PR middle-end/84000
14317 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14319 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14321 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14322 to merge probabilities.
14323 * predict.c (probably_never_executed): Also mark as cold functions
14324 with global 0 profile and guessed local profile.
14325 * profile-count.c (profile_probability::combine_with_count): New
14327 * profile-count.h (profile_probability::operator*,
14328 profile_probability::operator*=, profile_probability::operator/,
14329 profile_probability::operator/=): Reduce precision to adjusted
14330 and set value to guessed on contradictory divisions.
14331 (profile_probability::combine_with_freq): Remove.
14332 (profile_probability::combine_wiht_count): Declare.
14333 (profile_count::force_nonzero):: Set to adjusted.
14334 (profile_count::probability_in):: Set quality to adjusted.
14335 * tree-ssa-tail-merge.c (replace_block_by): Use
14336 combine_with_count.
14338 2018-01-23 Andrew Waterman <andrew@sifive.com>
14339 Jim Wilson <jimw@sifive.com>
14341 * config/riscv/riscv.c (riscv_stack_boundary): New.
14342 (riscv_option_override): Set riscv_stack_boundary. Handle
14343 riscv_preferred_stack_boundary_arg.
14344 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14345 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14346 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14347 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14348 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14349 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14351 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14354 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14355 of struct ix86_frame.
14356 (ix86_expand_epilogue): Likewise. Add a local variable for
14357 the reg_save_offset field in struct ix86_frame.
14359 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14361 PR tree-optimization/82604
14362 * tree-loop-distribution.c (enum partition_kind): New enum item
14363 PKIND_PARTIAL_MEMSET.
14364 (partition_builtin_p): Support above new enum item.
14365 (generate_code_for_partition): Ditto.
14366 (compute_access_range): Differentiate cases that equality can be
14367 proven at all loops, the innermost loops or no loops.
14368 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14369 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14370 (finalize_partitions, distribute_loop): Don't fuse partition of
14371 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14372 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14373 parloop is enabled.
14375 2018-01-23 Martin Liska <mliska@suse.cz>
14377 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14378 order to ignore the predictor.
14379 (PRED_POLYMORPHIC_CALL): Likewise.
14380 (PRED_RECURSIVE_CALL): Likewise.
14382 2018-01-23 Martin Liska <mliska@suse.cz>
14384 * tree-profile.c (tree_profiling): Print function header to
14385 aware reader which function we are working on.
14386 * value-prof.c (gimple_find_values_to_profile): Do not print
14387 not interesting value histograms.
14389 2018-01-23 Martin Liska <mliska@suse.cz>
14391 * profile-count.h (enum profile_quality): Add
14392 profile_uninitialized as the first value. Do not number values
14393 as they are zero based.
14394 (profile_count::verify): Update sanity check.
14395 (profile_probability::verify): Likewise.
14397 2018-01-23 Nathan Sidwell <nathan@acm.org>
14399 * doc/invoke.texi (ffor-scope): Deprecate.
14401 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14403 PR tree-optimization/83510
14404 * domwalk.c (set_all_edges_as_executable): New function.
14405 (dom_walker::dom_walker): Convert bool param
14406 "skip_unreachable_blocks" to enum reachability. Move setup of
14407 edge flags to set_all_edges_as_executable and only do it when
14408 reachability is REACHABLE_BLOCKS.
14409 * domwalk.h (enum dom_walker::reachability): New enum.
14410 (dom_walker::dom_walker): Convert bool param
14411 "skip_unreachable_blocks" to enum reachability.
14412 (set_all_edges_as_executable): New decl.
14413 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14414 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14416 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14417 but converting true to REACHABLE_BLOCKS.
14418 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14420 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14421 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14422 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14424 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14425 if check_all_array_refs will be called.
14427 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14429 * tree.c (selftest::test_location_wrappers): Add more test
14432 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14434 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14435 (selftest::test_bit_in_range): Likewise.
14437 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14440 * doc/sourcebuild.texi (vect_float): Say that the selector
14441 only describes the situation when -funsafe-math-optimizations is on.
14442 (vect_float_strict): Document.
14444 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14446 PR tree-optimization/83965
14447 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14448 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14449 instead of checking only for a reduction.
14450 (vect_recog_widen_sum_pattern): Likewise.
14452 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14454 * predict.c (probably_never_executed): Only use precise profile info.
14455 (compute_function_frequency): Skip after inlining hack since we now
14456 have quality checking.
14458 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14460 * profile-count.h (profile_probability::very_unlikely,
14461 profile_probability::unlikely, profile_probability::even): Set
14462 precision to guessed.
14464 2018-01-23 Richard Biener <rguenther@suse.de>
14466 PR tree-optimization/83963
14467 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14468 Properly terminate dominator walk when crossing the exit edge not
14469 when visiting its source block.
14471 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14474 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14475 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14477 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14479 PR tree-optimization/83957
14480 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14481 semicolon after for body surrounded by braces.
14483 PR tree-optimization/83081
14484 * profile-count.h (profile_probability::split): New method.
14485 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14486 Use profile_probability::split.
14487 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14488 when splitting a single conditional jump into 2.
14490 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14492 PR tree-optimization/69452
14493 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14496 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14498 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14499 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14500 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14502 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14504 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14506 * config/rl78/rl78.md (movdi): New define_expand.
14507 * config/rl78/rl78.c (rl78_split_movdi): New function.
14509 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14512 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14514 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14515 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14516 128-bit to produce an UNSPEC move to get the double word with the
14517 signbit and then a shift directly to do signbit.
14518 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14519 implementation with a new version that just does either a direct
14520 move or a regular move. Move memory interface to separate insns.
14521 Move insns so they are next to the expander.
14522 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14523 with signbit move. Split big and little endian case.
14524 (signbit<mode>2_dm_mem_le): Likewise.
14525 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14526 (signbit<mode>2_dm2): Likewise.
14528 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14530 * config/rl78/rl78.md (anddi3): New define_expand.
14532 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14534 * config/rl78/rl78.md (umindi3): New define_expand.
14536 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14538 * config/rl78/rl78.md (smindi3): New define_expand.
14540 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14542 * config/rl78/rl78.md (smaxdi3): New define_expand.
14544 2018-01-22 Carl Love <cel@us.ibm.com>
14546 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14547 LVX_V1TI): Add macro expansion.
14548 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14549 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14550 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14551 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14552 Change check to determine if the instruction is a byte reversing
14553 entry. Fix typo in comment.
14554 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14555 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14556 Add def_builtin calls for new builtins.
14557 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14558 Add define_insn expansion.
14560 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14562 * config/rl78/rl78.md (umaxdi3): New define_expand.
14564 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14566 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14567 for non-QImode registers.
14569 2018-01-22 Richard Biener <rguenther@suse.de>
14571 PR tree-optimization/83963
14572 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14573 including the loop exit block.
14574 (scop_detection::merge_sese): Likewise.
14575 (scop_detection::add_scop): Do it here instead.
14577 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14579 * doc/sourcebuild.texi (arm_softfloat): Document.
14581 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14584 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14585 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14586 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14588 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14589 David Edelsohn <dje.gcc@gmail.com>
14592 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14593 Change "crset eq" to "crset 2".
14594 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14595 (*call_indirect_aix<mode>_nospec): Likewise.
14596 (*call_value_indirect_aix<mode>_nospec): Likewise.
14597 (*call_indirect_elfv2<mode>_nospec): Likewise.
14598 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14599 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14600 change assembly output from . to $.
14601 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14602 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14603 (*tablejump<mode>_internal1_nospec): Likewise.
14605 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14608 * config/sh/sh_optimize_sett_clrt.cc:
14609 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14611 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14613 PR tree-optimization/83940
14614 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14615 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14616 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14617 use it to pass back the definition type.
14618 (vect_check_store_rhs): Likewise rhs_dt_out.
14619 (vect_build_gather_load_calls): Add a mask_dt argument and use
14620 it instead of a call to vect_is_simple_use.
14621 (vectorizable_store): Update calls to vect_check_load_store_mask
14622 and vect_check_store_rhs. Use the dt returned by the latter instead
14623 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14624 instead of calls to vect_is_simple_use. Pass the scalar rather
14625 than the vector operand to vect_is_simple_use when handling
14626 second and subsequent copies of an rhs value.
14627 (vectorizable_load): Update calls to vect_check_load_store_mask
14628 and vect_build_gather_load_calls. Use the cached mask_dt and
14629 gs_info.offset_dt instead of calls to vect_is_simple_use.
14631 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14633 PR middle-end/83945
14634 * tree-emutls.c: Include gimplify.h.
14635 (lower_emutls_2): New function.
14636 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14637 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14638 it before further processing.
14641 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14642 UINTVAL (trueop1) instead of INTVAL (op1).
14644 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14648 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14649 INCOMING_FRAME_SP_OFFSET if not defined.
14650 (scan_trace): Add ENTRY argument. If true and
14651 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14652 emit a note to adjust the CFA offset.
14653 (create_cfi_notes): Adjust scan_trace callers.
14654 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14655 INCOMING_FRAME_SP_OFFSET in the CIE.
14656 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14657 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14659 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14660 * doc/tm.texi: Regenerated.
14662 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14664 PR rtl-optimization/83147
14665 * lra-constraints.c (remove_inheritance_pseudos): Use
14666 lra_substitute_pseudo_within_insn.
14668 2018-01-19 Tom de Vries <tom@codesourcery.com>
14669 Cesar Philippidis <cesar@codesourcery.com>
14672 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14674 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14677 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14678 spaces for function labels.
14680 2018-01-19 Martin Liska <mliska@suse.cz>
14682 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14683 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14684 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14685 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14686 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14687 (PRED_CONST_RETURN): Change from 69 to 65.
14688 (PRED_NULL_RETURN): Change from 91 to 71.
14689 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14690 (PRED_LOOP_GUARD): Change from 66 to 73.
14692 2018-01-19 Martin Liska <mliska@suse.cz>
14694 * predict.c (predict_insn_def): Add new assert.
14695 (struct branch_predictor): Change type to signed integer.
14696 (test_prediction_value_range): Amend test to cover
14697 PROB_UNINITIALIZED.
14698 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14699 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14700 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14701 (PRED_LOOP_IV_COMPARE): Likewise.
14702 * predict.h (PROB_UNINITIALIZED): Define new constant.
14704 2018-01-19 Martin Liska <mliska@suse.cz>
14706 * predict.c (dump_prediction): Add new format for
14707 analyze_brprob.py script which is enabled with -details
14709 * profile-count.h (precise_p): New function.
14711 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14713 PR tree-optimization/83922
14714 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14715 there are no statements that need masking.
14716 (vect_active_double_reduction_p): New function.
14717 (vect_analyze_loop_operations): Use it when handling phis that
14718 are not in the loop header.
14720 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14722 PR tree-optimization/83914
14723 * tree-vect-loop.c (vectorizable_induction): Don't convert
14724 init_expr or apply the peeling adjustment for inductions
14725 that are nested within the vectorized loop.
14727 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14729 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14732 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14736 * function.h (gimplify_parameters): Add gimple_seq * argument.
14737 * function.c: Include gimple.h and options.h.
14738 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14739 for the added local temporaries if needed.
14740 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14741 if there are any parameter cleanups, wrap whole body into a
14742 try/finally with the cleanups.
14744 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14747 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14748 Use GET_MODE_CLASS for scalar floating point.
14750 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14754 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14755 Fix call of call_cgraph_insertion_hooks.
14757 2018-01-18 Martin Sebor <msebor@redhat.com>
14759 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14761 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14764 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14767 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14770 * common.opt: (-ffile-prefix-map): New option.
14771 * opts.c (common_handle_option): Defer it.
14772 * opts-global.c (handle_common_deferred_options): Handle it.
14773 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14774 * file-prefix-map.h: New file.
14775 (remap_debug_filename, add_debug_prefix_map): ...here.
14776 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14777 * final.c (debug_prefix_map, add_debug_prefix_map
14778 remap_debug_filename): Move to...
14779 * file-prefix-map.c: New file.
14780 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14781 generalize, get rid of alloca(), use strrchr() instead of strchr().
14782 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14783 Implement in terms of add_prefix_map().
14784 (remap_macro_filename, remap_debug_filename): Implement in term of
14786 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14787 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14788 * dbxout.c: Include file-prefix-map.h.
14789 * varasm.c: Likewise.
14790 * vmsdbgout.c: Likewise.
14791 * xcoffout.c: Likewise.
14792 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14793 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14794 * doc/invoke.texi (-ffile-prefix-map): Document.
14795 (-fdebug-prefix-map): Update description.
14797 2018-01-18 Martin Liska <mliska@suse.cz>
14799 * config/i386/i386.c (indirect_thunk_name): Document that also
14801 (output_indirect_thunk): Document why both instructions
14802 (pause and lfence) are generated.
14804 2018-01-18 Richard Biener <rguenther@suse.de>
14806 PR tree-optimization/83887
14807 * graphite-scop-detection.c
14808 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14809 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14810 (scop_detection::merge_sese): Re-implement with a flood-fill
14811 algorithm that properly finds a SESE region if it exists.
14813 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14816 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14817 pointer_diff optimizations use view_convert instead of convert.
14819 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14821 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14822 Generate different code for -mno-speculate-indirect-jumps.
14823 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14824 (*call_indirect_aix<mode>): Disable for
14825 -mno-speculate-indirect-jumps.
14826 (*call_indirect_aix<mode>_nospec): New define_insn.
14827 (*call_value_indirect_aix<mode>): Disable for
14828 -mno-speculate-indirect-jumps.
14829 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14830 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14831 -mno-speculate-indirect-jumps.
14832 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14834 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14836 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14837 long double type, set the flags for noting the default long double
14838 type, even if we don't pass or return a long double type.
14840 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14843 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14846 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14848 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14849 support for merge[hl].
14850 (fold_mergehl_helper): New helper function.
14851 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14852 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14853 (altivec_vmrglw_direct): Add xxmrglw insn.
14855 2018-01-17 Andrew Waterman <andrew@sifive.com>
14857 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14858 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14860 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14863 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14864 call the lto_location_cache before reading the
14865 DECL_SOURCE_LOCATION of the types.
14867 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14868 Richard Sandiford <richard.sandiford@linaro.org>
14870 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14871 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14872 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14873 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14874 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14876 * config/aarch64/constraints.md (aarch64_movti_operand):
14878 * config/aarch64/predicates.md (Uti): Add new constraint.
14880 2018-01-17 Carl Love <cel@us.ibm.com>
14882 * config/rs6000/vsx.md (define_expand xl_len_r,
14883 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14884 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14886 (define_expand, define_insn): Move the shift left from the
14887 define_insn to the define_expand for lxvl and stxvl instructions.
14888 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14889 and XL_LEN_R definitions to PURE.
14891 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14893 * config/i386/i386.c (indirect_thunk_name): Declare regno
14894 as unsigned int. Compare regno with INVALID_REGNUM.
14895 (output_indirect_thunk): Ditto.
14896 (output_indirect_thunk_function): Ditto.
14897 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14898 in the call to output_indirect_thunk_function.
14900 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14902 PR middle-end/83884
14903 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14904 rather than the size of inner_type to determine the stack slot size
14905 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14907 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14910 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14913 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14915 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14916 endian Linux systems to optionally enable multilibs for selecting
14917 the long double type if the user configured an explicit type.
14918 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14919 have no long double multilibs if not defined.
14920 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14921 warn if the user used -mabi={ieee,ibm}longdouble and we built
14922 multilibs for long double.
14923 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14924 appropriate multilib option.
14925 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14927 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14928 for building long double multilibs.
14929 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14931 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14933 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14936 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14938 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14941 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14944 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14947 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14949 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14950 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14952 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14954 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14955 different rtl trees depending on TARGET_64BIT.
14956 (rs6000_gen_lvx): Likewise.
14958 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14960 * config/visium/visium.md (nop): Tweak comment.
14961 (hazard_nop): Likewise.
14963 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14965 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14966 -mspeculate-indirect-jumps.
14967 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14968 for -mno-speculate-indirect-jumps.
14969 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14970 (*call_value_indirect_elfv2<mode>): Disable for
14971 -mno-speculate-indirect-jumps.
14972 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14973 (indirect_jump): Emit different RTL for
14974 -mno-speculate-indirect-jumps.
14975 (*indirect_jump<mode>): Disable for
14976 -mno-speculate-indirect-jumps.
14977 (*indirect_jump<mode>_nospec): New define_insn.
14978 (tablejump): Emit different RTL for
14979 -mno-speculate-indirect-jumps.
14980 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14981 (tablejumpsi_nospec): New define_expand.
14982 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14983 (tablejumpdi_nospec): New define_expand.
14984 (*tablejump<mode>_internal1): Disable for
14985 -mno-speculate-indirect-jumps.
14986 (*tablejump<mode>_internal1_nospec): New define_insn.
14987 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14990 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14992 * caller-save.c (insert_save): Drop unnecessary parameter. All
14995 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14996 Richard Biener <rguenth@suse.de>
14999 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15000 return early, inline manually is_gimple_sizepos. Make sure if we
15001 call gimplify_expr we don't end up with a gimple constant.
15002 * tree.c (variably_modified_type_p): Don't return true for
15003 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15004 * gimplify.h (is_gimple_sizepos): Remove.
15006 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15008 PR tree-optimization/83857
15009 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15010 vectorizable_live_operation for pure SLP statements.
15011 (vectorizable_live_operation): Handle PHIs.
15013 2018-01-16 Richard Biener <rguenther@suse.de>
15015 PR tree-optimization/83867
15016 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15017 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15019 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15022 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15023 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15024 If off is not INTEGER_CST, issue a may not be aligned warning
15025 rather than isn't aligned. Use isn%'t rather than isn't.
15026 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15028 <case MULT_EXPR>: Improve the case when bottom and one of the
15029 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15030 operand, in that case check if the other operand is multiple of
15031 bottom divided by the INTEGER_CST operand.
15033 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15036 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15037 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15038 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15039 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15040 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15041 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15042 (pa_function_arg_size): New function.
15044 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15046 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15047 in a separate statement.
15049 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15051 PR tree-optimization/83847
15052 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15053 group gathers and scatters.
15055 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15057 PR rtl-optimization/86620
15058 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15060 PR rtl-optimization/83213
15061 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15062 to last if both are JUMP_INSNs.
15064 PR tree-optimization/83843
15065 * gimple-ssa-store-merging.c
15066 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15067 store_immediate_info for bswap/nop orig_stores.
15069 2018-01-15 Andrew Waterman <andrew@sifive.com>
15071 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15073 <UDIV>: Increase cost if !TARGET_DIV.
15075 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15077 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15078 (define_attr "cr_logical_3op"): New.
15079 (cceq_ior_compare): Adjust.
15080 (cceq_ior_compare_complement): Adjust.
15081 (*cceq_rev_compare): Adjust.
15082 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15083 (is_cracked_insn): Adjust.
15084 (insn_must_be_first_in_group): Adjust.
15085 * config/rs6000/40x.md: Adjust.
15086 * config/rs6000/440.md: Adjust.
15087 * config/rs6000/476.md: Adjust.
15088 * config/rs6000/601.md: Adjust.
15089 * config/rs6000/603.md: Adjust.
15090 * config/rs6000/6xx.md: Adjust.
15091 * config/rs6000/7450.md: Adjust.
15092 * config/rs6000/7xx.md: Adjust.
15093 * config/rs6000/8540.md: Adjust.
15094 * config/rs6000/cell.md: Adjust.
15095 * config/rs6000/e300c2c3.md: Adjust.
15096 * config/rs6000/e500mc.md: Adjust.
15097 * config/rs6000/e500mc64.md: Adjust.
15098 * config/rs6000/e5500.md: Adjust.
15099 * config/rs6000/e6500.md: Adjust.
15100 * config/rs6000/mpc.md: Adjust.
15101 * config/rs6000/power4.md: Adjust.
15102 * config/rs6000/power5.md: Adjust.
15103 * config/rs6000/power6.md: Adjust.
15104 * config/rs6000/power7.md: Adjust.
15105 * config/rs6000/power8.md: Adjust.
15106 * config/rs6000/power9.md: Adjust.
15107 * config/rs6000/rs64.md: Adjust.
15108 * config/rs6000/titan.md: Adjust.
15110 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15112 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15113 ix86_indirect_branch_register logic.
15115 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15117 * config/i386/constraints.md (Bs): Update
15118 ix86_indirect_branch_register check. Don't check
15119 ix86_indirect_branch_register with GOT_memory_operand.
15121 * config/i386/predicates.md (GOT_memory_operand): Don't check
15122 ix86_indirect_branch_register here.
15123 (GOT32_symbol_operand): Likewise.
15125 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15127 * config/i386/predicates.md (constant_call_address_operand):
15128 Rewrite ix86_indirect_branch_register logic.
15129 (sibcall_insn_operand): Likewise.
15131 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15133 * config/i386/constraints.md (Bs): Replace
15134 ix86_indirect_branch_thunk_register with
15135 ix86_indirect_branch_register.
15137 * config/i386/i386.md (indirect_jump): Likewise.
15138 (tablejump): Likewise.
15139 (*sibcall_memory): Likewise.
15140 (*sibcall_value_memory): Likewise.
15141 Peepholes of indirect call and jump via memory: Likewise.
15142 * config/i386/i386.opt: Likewise.
15143 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15144 (GOT_memory_operand): Likewise.
15145 (call_insn_operand): Likewise.
15146 (sibcall_insn_operand): Likewise.
15147 (GOT32_symbol_operand): Likewise.
15149 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15151 PR middle-end/83837
15152 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15153 type rather than type addr's type points to.
15154 (expand_omp_atomic_mutex): Likewise.
15155 (expand_omp_atomic): Likewise.
15157 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15160 * config/i386/i386.c (output_indirect_thunk_function): Use
15161 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15162 for __x86_return_thunk.
15164 2018-01-15 Richard Biener <rguenther@suse.de>
15166 PR middle-end/83850
15167 * expmed.c (extract_bit_field_1): Fix typo.
15169 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15172 * config/arm/iterators.md (VF): New mode iterator.
15173 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15174 Remove integer-related logic from pattern.
15175 (neon_vabd<mode>_3): Likewise.
15177 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15179 PR middle-end/82694
15180 * common.opt (fstrict-overflow): No longer an alias.
15181 (fwrapv-pointer): New option.
15182 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15183 also for pointer types based on flag_wrapv_pointer.
15184 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15185 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15186 opts->x_flag_wrapv got set.
15187 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15188 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15189 POINTER_TYPE_OVERFLOW_UNDEFINED.
15190 * match.pd: Likewise in address comparison pattern.
15191 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15193 2018-01-15 Richard Biener <rguenther@suse.de>
15196 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15197 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15198 Reset type names to their identifier if their TYPE_DECL doesn't
15199 have linkage (and thus is used for ODR and devirt).
15200 (save_debug_info_for_decl): Remove.
15201 (save_debug_info_for_type): Likewise.
15202 (add_tree_to_fld_list): Adjust.
15203 * tree-pretty-print.c (dump_generic_node): Make dumping of
15204 type names more robust.
15206 2018-01-15 Richard Biener <rguenther@suse.de>
15208 * BASE-VER: Bump to 8.0.1.
15210 2018-01-14 Martin Sebor <msebor@redhat.com>
15213 * builtins.c (check_access): Avoid warning when the no-warning bit
15216 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15218 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15219 * ira-color (allocno_hard_regs_compare): Likewise.
15221 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15224 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15225 Use .pushsection/.popsection.
15227 2018-01-14 Martin Sebor <msebor@redhat.com>
15230 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15232 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15234 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15235 entry from extra_headers.
15236 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15237 extra_headers, make the list bitwise identical to the i?86-*-* one.
15239 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15241 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15242 -mcmodel=large with -mindirect-branch=thunk,
15243 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15244 -mfunction-return=thunk-extern.
15245 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15246 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15247 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15249 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15251 * config/i386/i386.c (print_reg): Print the name of the full
15252 integer register without '%'.
15253 (ix86_print_operand): Handle 'V'.
15254 * doc/extend.texi: Document 'V' modifier.
15256 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15258 * config/i386/constraints.md (Bs): Disallow memory operand for
15259 -mindirect-branch-register.
15261 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15262 (GOT_memory_operand): Likewise.
15263 (call_insn_operand): Likewise.
15264 (sibcall_insn_operand): Likewise.
15265 (GOT32_symbol_operand): Likewise.
15266 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15267 for -mindirect-branch-register.
15268 (tablejump): Likewise.
15269 (*sibcall_memory): Likewise.
15270 (*sibcall_value_memory): Likewise.
15271 Disallow peepholes of indirect call and jump via memory for
15272 -mindirect-branch-register.
15273 (*call_pop): Replace m with Bw.
15274 (*call_value_pop): Likewise.
15275 (*sibcall_pop_memory): Replace m with Bs.
15276 * config/i386/i386.opt (mindirect-branch-register): New option.
15277 * doc/invoke.texi: Document -mindirect-branch-register option.
15279 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15281 * config/i386/i386-protos.h (ix86_output_function_return): New.
15282 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15283 set function_return_type.
15284 (indirect_thunk_name): Add ret_p to indicate thunk for function
15286 (output_indirect_thunk_function): Pass false to
15287 indirect_thunk_name.
15288 (ix86_output_indirect_branch_via_reg): Likewise.
15289 (ix86_output_indirect_branch_via_push): Likewise.
15290 (output_indirect_thunk_function): Create alias for function
15291 return thunk if regno < 0.
15292 (ix86_output_function_return): New function.
15293 (ix86_handle_fndecl_attribute): Handle function_return.
15294 (ix86_attribute_table): Add function_return.
15295 * config/i386/i386.h (machine_function): Add
15296 function_return_type.
15297 * config/i386/i386.md (simple_return_internal): Use
15298 ix86_output_function_return.
15299 (simple_return_internal_long): Likewise.
15300 * config/i386/i386.opt (mfunction-return=): New option.
15301 (indirect_branch): Mention -mfunction-return=.
15302 * doc/extend.texi: Document function_return function attribute.
15303 * doc/invoke.texi: Document -mfunction-return= option.
15305 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15307 * config/i386/i386-opts.h (indirect_branch): New.
15308 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15309 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15310 with local indirect jump when converting indirect call and jump.
15311 (ix86_set_indirect_branch_type): New.
15312 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15313 (indirectlabelno): New.
15314 (indirect_thunk_needed): Likewise.
15315 (indirect_thunk_bnd_needed): Likewise.
15316 (indirect_thunks_used): Likewise.
15317 (indirect_thunks_bnd_used): Likewise.
15318 (INDIRECT_LABEL): Likewise.
15319 (indirect_thunk_name): Likewise.
15320 (output_indirect_thunk): Likewise.
15321 (output_indirect_thunk_function): Likewise.
15322 (ix86_output_indirect_branch_via_reg): Likewise.
15323 (ix86_output_indirect_branch_via_push): Likewise.
15324 (ix86_output_indirect_branch): Likewise.
15325 (ix86_output_indirect_jmp): Likewise.
15326 (ix86_code_end): Call output_indirect_thunk_function if needed.
15327 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15329 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15330 (ix86_attribute_table): Add indirect_branch.
15331 * config/i386/i386.h (machine_function): Add indirect_branch_type
15332 and has_local_indirect_jump.
15333 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15335 (tablejump): Likewise.
15336 (*indirect_jump): Use ix86_output_indirect_jmp.
15337 (*tablejump_1): Likewise.
15338 (simple_return_indirect_internal): Likewise.
15339 * config/i386/i386.opt (mindirect-branch=): New option.
15340 (indirect_branch): New.
15343 (thunk-inline): Likewise.
15344 (thunk-extern): Likewise.
15345 * doc/extend.texi: Document indirect_branch function attribute.
15346 * doc/invoke.texi: Document -mindirect-branch= option.
15348 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15351 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15353 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15355 * ipa-inline.c (want_inline_small_function_p): Return false if
15356 inlining has already failed with CIF_FINAL_ERROR.
15357 (update_caller_keys): Call want_inline_small_function_p before
15359 (update_callee_keys): Likewise.
15361 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15363 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15365 (rs6000_quadword_masked_address_p): Likewise.
15366 (quad_aligned_load_p): Likewise.
15367 (quad_aligned_store_p): Likewise.
15368 (const_load_sequence_p): Add comment to describe the outer-most loop.
15369 (mimic_memory_attributes_and_flags): New function.
15370 (rs6000_gen_stvx): Likewise.
15371 (replace_swapped_aligned_store): Likewise.
15372 (rs6000_gen_lvx): Likewise.
15373 (replace_swapped_aligned_load): Likewise.
15374 (replace_swapped_load_constant): Capitalize argument name in
15375 comment describing this function.
15376 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15377 and stores that access quad-word aligned addresses and replace
15378 with stvx or lvx instructions when appropriate.
15379 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15380 New function prototype.
15381 (rs6000_quadword_masked_address_p): Likewise.
15382 (rs6000_gen_lvx): Likewise.
15383 (rs6000_gen_stvx): Likewise.
15384 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15385 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15386 when memory address is aligned.
15387 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15388 this split to select lvx instruction when memory address is aligned.
15389 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15390 instruction when memory address is aligned.
15391 (*vsx_le_perm_load_v16qi): Likewise.
15392 (four unnamed splitters): Modify to select the stvx instruction
15393 when memory is aligned.
15395 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15397 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15398 which appears in the queue multiple times.
15400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15401 Alan Hayward <alan.hayward@arm.com>
15402 David Sherwood <david.sherwood@arm.com>
15404 * tree-vectorizer.h (vec_lower_bound): New structure.
15405 (_loop_vec_info): Add check_nonzero and lower_bounds.
15406 (LOOP_VINFO_CHECK_NONZERO): New macro.
15407 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15408 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15409 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15410 fields. Make seg_len the distance travelled, not including the
15412 (dr_direction_indicator): Declare.
15413 (dr_zero_step_indicator): Likewise.
15414 (dr_known_forward_stride_p): Likewise.
15415 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15417 (runtime_alias_check_p): Allow runtime alias checks with
15419 (operator ==): Compare access_size and align.
15420 (prune_runtime_alias_test_list): Rework for new distinction between
15421 the access_size and seg_len.
15422 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15424 (get_segment_min_max): New function.
15425 (create_intersect_range_checks): Use it.
15426 (dr_step_indicator): New function.
15427 (dr_direction_indicator): Likewise.
15428 (dr_zero_step_indicator): Likewise.
15429 (dr_known_forward_stride_p): Likewise.
15430 * tree-loop-distribution.c (data_ref_segment_size): Return
15431 DR_STEP * (niters - 1).
15432 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15434 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15435 (vect_preserves_scalar_order_p): New function, split out from...
15436 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15437 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15438 (vect_vfa_access_size): New function.
15439 (vect_vfa_align): Likewise.
15440 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15441 (dump_lower_bound): New function.
15442 (vect_check_lower_bound): Likewise.
15443 (vect_small_gap_p): Likewise.
15444 (vectorizable_with_step_bound_p): Likewise.
15445 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15446 depencies if the vectorization factor is 1. Convert the checks
15447 for nonzero steps into checks on the bounds of DR_STEP. Try using
15448 a bunds check for variable steps if the minimum required step is
15449 relatively small. Update calls to the dr_with_seg_len
15450 constructor and to vect_compile_time_alias.
15451 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15453 (vect_loop_versioning): Call it.
15454 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15456 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15459 Alan Hayward <alan.hayward@arm.com>
15460 David Sherwood <david.sherwood@arm.com>
15462 * doc/sourcebuild.texi (vect_scatter_store): Document.
15463 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15465 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15467 * genopinit.c (main): Add supports_vec_scatter_store and
15468 supports_vec_scatter_store_cached to target_optabs.
15469 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15470 IFN_MASK_SCATTER_STORE.
15471 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15473 * internal-fn.h (internal_store_fn_p): Declare.
15474 (internal_fn_stored_value_index): Likewise.
15475 * internal-fn.c (scatter_store_direct): New macro.
15476 (expand_scatter_store_optab_fn): New function.
15477 (direct_scatter_store_optab_supported_p): New macro.
15478 (internal_store_fn_p): New function.
15479 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15480 IFN_MASK_SCATTER_STORE.
15481 (internal_fn_mask_index): Likewise.
15482 (internal_fn_stored_value_index): New function.
15483 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15484 for scatter stores.
15485 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15486 * optabs-query.c (supports_vec_scatter_store_p): New function.
15487 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15488 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15489 true for scatter stores.
15490 (vect_gather_scatter_fn_p): Handle scatter stores too.
15491 (vect_check_gather_scatter): Consider using scatter stores if
15492 supports_vec_scatter_store_p.
15493 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15494 scatter stores too.
15495 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15496 internal_fn_stored_value_index.
15497 (check_load_store_masking): Handle scatter stores too.
15498 (vect_get_store_rhs): Make public.
15499 (vectorizable_call): Use internal_store_fn_p.
15500 (vectorizable_store): Handle scatter store internal functions.
15501 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15502 when deciding whether the end of the group has been reached.
15503 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15504 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15505 (mask_scatter_store<mode>): New insns.
15507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15508 Alan Hayward <alan.hayward@arm.com>
15509 David Sherwood <david.sherwood@arm.com>
15511 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15512 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15513 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15515 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15516 Use vect_truncate_gather_scatter_offset if we can't treat the
15517 operation as a normal gather load or scatter store.
15518 (get_group_load_store_type): Take the gather_scatter_info
15519 as argument. Try using a gather load or scatter store for
15520 single-element groups.
15521 (get_load_store_type): Update calls to get_group_load_store_type
15522 and vect_use_strided_gather_scatters_p.
15524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15525 Alan Hayward <alan.hayward@arm.com>
15526 David Sherwood <david.sherwood@arm.com>
15528 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15529 optional tree argument.
15530 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15532 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15533 but continue to use the current value as a fallback.
15534 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15535 to compare the updates.
15536 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15537 (get_load_store_type): Use it when handling a strided access.
15538 (vect_get_strided_load_store_ops): New function.
15539 (vect_get_data_ptr_increment): Likewise.
15540 (vectorizable_load): Handle strided gather loads. Always pass
15541 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15544 Alan Hayward <alan.hayward@arm.com>
15545 David Sherwood <david.sherwood@arm.com>
15547 * doc/md.texi (gather_load@var{m}): Document.
15548 (mask_gather_load@var{m}): Likewise.
15549 * genopinit.c (main): Add supports_vec_gather_load and
15550 supports_vec_gather_load_cached to target_optabs.
15551 * optabs-tree.c (init_tree_optimization_optabs): Use
15552 ggc_cleared_alloc to allocate target_optabs.
15553 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15554 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15556 * internal-fn.h (internal_load_fn_p): Declare.
15557 (internal_gather_scatter_fn_p): Likewise.
15558 (internal_fn_mask_index): Likewise.
15559 (internal_gather_scatter_fn_supported_p): Likewise.
15560 * internal-fn.c (gather_load_direct): New macro.
15561 (expand_gather_load_optab_fn): New function.
15562 (direct_gather_load_optab_supported_p): New macro.
15563 (direct_internal_fn_optab): New function.
15564 (internal_load_fn_p): Likewise.
15565 (internal_gather_scatter_fn_p): Likewise.
15566 (internal_fn_mask_index): Likewise.
15567 (internal_gather_scatter_fn_supported_p): Likewise.
15568 * optabs-query.c (supports_at_least_one_mode_p): New function.
15569 (supports_vec_gather_load_p): Likewise.
15570 * optabs-query.h (supports_vec_gather_load_p): Declare.
15571 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15572 and memory_type field.
15573 (NUM_PATTERNS): Bump to 15.
15574 * tree-vect-data-refs.c: Include internal-fn.h.
15575 (vect_gather_scatter_fn_p): New function.
15576 (vect_describe_gather_scatter_call): Likewise.
15577 (vect_check_gather_scatter): Try using internal functions for
15578 gather loads. Recognize existing calls to a gather load function.
15579 (vect_analyze_data_refs): Consider using gather loads if
15580 supports_vec_gather_load_p.
15581 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15582 (vect_get_gather_scatter_offset_type): Likewise.
15583 (vect_convert_mask_for_vectype): Likewise.
15584 (vect_add_conversion_to_patterm): Likewise.
15585 (vect_try_gather_scatter_pattern): Likewise.
15586 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15587 (vect_vect_recog_func_ptrs): Add it.
15588 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15589 internal_fn_mask_index and internal_gather_scatter_fn_p.
15590 (check_load_store_masking): Take the gather_scatter_info as an
15591 argument and handle gather loads.
15592 (vect_get_gather_scatter_ops): New function.
15593 (vectorizable_call): Check internal_load_fn_p.
15594 (vectorizable_load): Likewise. Handle gather load internal
15596 (vectorizable_store): Update call to check_load_store_masking.
15597 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15598 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15599 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15600 (aarch64_gather_scale_operand_d): New predicates.
15601 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15602 (mask_gather_load<mode>): New insns.
15604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15605 Alan Hayward <alan.hayward@arm.com>
15606 David Sherwood <david.sherwood@arm.com>
15608 * optabs.def (fold_left_plus_optab): New optab.
15609 * doc/md.texi (fold_left_plus_@var{m}): Document.
15610 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15611 * internal-fn.c (fold_left_direct): Define.
15612 (expand_fold_left_optab_fn): Likewise.
15613 (direct_fold_left_optab_supported_p): Likewise.
15614 * fold-const-call.c (fold_const_fold_left): New function.
15615 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15616 * tree-parloops.c (valid_reduction_p): New function.
15617 (gather_scalar_reductions): Use it.
15618 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15619 (vect_finish_replace_stmt): Declare.
15620 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15621 (needs_fold_left_reduction_p): New function, split out from...
15622 (vect_is_simple_reduction): ...here. Accept reductions that
15623 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15624 (vect_force_simple_reduction): Also store the reduction type in
15625 the assignment's STMT_VINFO_REDUC_TYPE.
15626 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15627 (merge_with_identity): New function.
15628 (vect_expand_fold_left): Likewise.
15629 (vectorize_fold_left_reduction): Likewise.
15630 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15631 scalar phi in place for it. Check for target support and reject
15632 cases that would reassociate the operation. Defer the transform
15633 phase to vectorize_fold_left_reduction.
15634 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15635 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15636 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15640 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15641 call to ifc_temp_var.
15643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15644 Alan Hayward <alan.hayward@arm.com>
15645 David Sherwood <david.sherwood@arm.com>
15647 * target.def (legitimize_address_displacement): Take the original
15648 offset as a poly_int.
15649 * targhooks.h (default_legitimize_address_displacement): Update
15651 * targhooks.c (default_legitimize_address_displacement): Likewise.
15652 * doc/tm.texi: Regenerate.
15653 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15654 as an argument, moving assert of ad->disp == ad->disp_term to...
15655 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15656 Try calling targetm.legitimize_address_displacement before expanding
15657 the address rather than afterwards, and adjust for the new interface.
15658 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15659 Match the new hook interface. Handle SVE addresses.
15660 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15661 new hook interface.
15663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15665 * Makefile.in (OBJS): Add early-remat.o.
15666 * target.def (select_early_remat_modes): New hook.
15667 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15668 * doc/tm.texi: Regenerate.
15669 * targhooks.h (default_select_early_remat_modes): Declare.
15670 * targhooks.c (default_select_early_remat_modes): New function.
15671 * timevar.def (TV_EARLY_REMAT): New timevar.
15672 * passes.def (pass_early_remat): New pass.
15673 * tree-pass.h (make_pass_early_remat): Declare.
15674 * early-remat.c: New file.
15675 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15677 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15680 Alan Hayward <alan.hayward@arm.com>
15681 David Sherwood <david.sherwood@arm.com>
15683 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15684 vfm1 with a bound_epilog parameter.
15685 (vect_do_peeling): Update calls accordingly, and move the prologue
15686 call earlier in the function. Treat the base bound_epilog as 0 for
15687 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15688 this base when peeling for gaps.
15689 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15690 with fully-masked loops.
15691 (vect_estimate_min_profitable_iters): Handle the single peeled
15692 iteration in that case.
15694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15695 Alan Hayward <alan.hayward@arm.com>
15696 David Sherwood <david.sherwood@arm.com>
15698 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15699 single-element interleaving even if the size is not a power of 2.
15700 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15701 accesses for single-element interleaving if the group size is
15704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15705 Alan Hayward <alan.hayward@arm.com>
15706 David Sherwood <david.sherwood@arm.com>
15708 * doc/md.texi (fold_extract_last_@var{m}): Document.
15709 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15710 * optabs.def (fold_extract_last_optab): New optab.
15711 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15712 * internal-fn.c (fold_extract_direct): New macro.
15713 (expand_fold_extract_optab_fn): Likewise.
15714 (direct_fold_extract_optab_supported_p): Likewise.
15715 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15716 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15717 EXTRACT_LAST_REDUCTION.
15718 (get_initial_def_for_reduction): Do not create an initial vector
15719 for EXTRACT_LAST_REDUCTION reductions.
15720 (vectorizable_reduction): Leave the scalar phi in place for
15721 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15722 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15723 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15724 transform phase to vectorizable_condition.
15725 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15727 (vect_finish_stmt_generation): ...here.
15728 (vect_finish_replace_stmt): New function.
15729 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15730 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15732 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15735 Alan Hayward <alan.hayward@arm.com>
15736 David Sherwood <david.sherwood@arm.com>
15738 * doc/md.texi (extract_last_@var{m}): Document.
15739 * optabs.def (extract_last_optab): New optab.
15740 * internal-fn.def (EXTRACT_LAST): New internal function.
15741 * internal-fn.c (cond_unary_direct): New macro.
15742 (expand_cond_unary_optab_fn): Likewise.
15743 (direct_cond_unary_optab_supported_p): Likewise.
15744 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15745 loops using EXTRACT_LAST.
15746 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15747 (extract_last_<mode>): ...this optab.
15748 (vec_extract<mode><Vel>): Update accordingly.
15750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15751 Alan Hayward <alan.hayward@arm.com>
15752 David Sherwood <david.sherwood@arm.com>
15754 * target.def (empty_mask_is_expensive): New hook.
15755 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15756 * doc/tm.texi: Regenerate.
15757 * targhooks.h (default_empty_mask_is_expensive): Declare.
15758 * targhooks.c (default_empty_mask_is_expensive): New function.
15759 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15760 if the target says that empty masks are expensive.
15761 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15763 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15766 Alan Hayward <alan.hayward@arm.com>
15767 David Sherwood <david.sherwood@arm.com>
15769 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15770 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15771 (vect_use_loop_mask_for_alignment_p): New function.
15772 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15773 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15774 niters_skip argument. Make sure that the first niters_skip elements
15775 of the first iteration are inactive.
15776 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15777 Update call to vect_set_loop_masks_directly.
15778 (get_misalign_in_elems): New function, split out from...
15779 (vect_gen_prolog_loop_niters): ...here.
15780 (vect_update_init_of_dr): Take a code argument that specifies whether
15781 the adjustment should be added or subtracted.
15782 (vect_update_init_of_drs): Likewise.
15783 (vect_prepare_for_masked_peels): New function.
15784 (vect_do_peeling): Skip prologue peeling if we're using a mask
15785 instead. Update call to vect_update_inits_of_drs.
15786 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15788 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15789 alignment. Do not include the number of peeled iterations in
15790 the minimum threshold in that case.
15791 (vectorizable_induction): Adjust the start value down by
15792 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15793 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15794 Take the number of skipped iterations into account when calculating
15796 * tree-vect-stmts.c (vect_gen_while_not): New function.
15798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15799 Alan Hayward <alan.hayward@arm.com>
15800 David Sherwood <david.sherwood@arm.com>
15802 * doc/sourcebuild.texi (vect_fully_masked): Document.
15803 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15804 default value to 0.
15805 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15807 (vect_analyze_loop_2): ...here. Don't check the vectorization
15808 factor against the number of loop iterations if the loop is
15811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15812 Alan Hayward <alan.hayward@arm.com>
15813 David Sherwood <david.sherwood@arm.com>
15815 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15816 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15817 (dump_groups): Update accordingly.
15818 (iv_use::mem_type): New member variable.
15819 (address_p): New function.
15820 (record_use): Add a mem_type argument and initialize the new
15822 (record_group_use): Add a mem_type argument. Use address_p.
15823 Remove obsolete null checks of base_object. Update call to record_use.
15824 (find_interesting_uses_op): Update call to record_group_use.
15825 (find_interesting_uses_cond): Likewise.
15826 (find_interesting_uses_address): Likewise.
15827 (get_mem_type_for_internal_fn): New function.
15828 (find_address_like_use): Likewise.
15829 (find_interesting_uses_stmt): Try find_address_like_use before
15830 calling find_interesting_uses_op.
15831 (addr_offset_valid_p): Use the iv mem_type field as the type
15832 of the addressed memory.
15833 (add_autoinc_candidates): Likewise.
15834 (get_address_cost): Likewise.
15835 (split_small_address_groups_p): Use address_p.
15836 (split_address_groups): Likewise.
15837 (add_iv_candidate_for_use): Likewise.
15838 (autoinc_possible_for_pair): Likewise.
15839 (rewrite_groups): Likewise.
15840 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15841 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15842 (get_alias_ptr_type_for_ptr_address): New function.
15843 (rewrite_use_address): Rewrite address uses in calls that were
15844 identified by find_address_like_use.
15846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15847 Alan Hayward <alan.hayward@arm.com>
15848 David Sherwood <david.sherwood@arm.com>
15850 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15852 * gimple-expr.h (is_gimple_addressable: Likewise.
15853 * gimple-expr.c (is_gimple_address): Likewise.
15854 * internal-fn.c (expand_call_mem_ref): New function.
15855 (expand_mask_load_optab_fn): Use it.
15856 (expand_mask_store_optab_fn): Likewise.
15858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15859 Alan Hayward <alan.hayward@arm.com>
15860 David Sherwood <david.sherwood@arm.com>
15862 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15863 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15864 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15865 (cond_umax@var{mode}): Document.
15866 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15867 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15868 (cond_umin_optab, cond_umax_optab): New optabs.
15869 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15870 (COND_IOR, COND_XOR): New internal functions.
15871 * internal-fn.h (get_conditional_internal_fn): Declare.
15872 * internal-fn.c (cond_binary_direct): New macro.
15873 (expand_cond_binary_optab_fn): Likewise.
15874 (direct_cond_binary_optab_supported_p): Likewise.
15875 (get_conditional_internal_fn): New function.
15876 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15877 Cope with reduction statements that are vectorized as calls rather
15879 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15880 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15881 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15882 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15883 (UNSPEC_COND_EOR): New unspecs.
15884 (optab): Add mappings for them.
15885 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15886 (sve_int_op, sve_fp_op): New int attributes.
15888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15889 Alan Hayward <alan.hayward@arm.com>
15890 David Sherwood <david.sherwood@arm.com>
15892 * optabs.def (while_ult_optab): New optab.
15893 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15894 * internal-fn.def (WHILE_ULT): New internal function.
15895 * internal-fn.h (direct_internal_fn_supported_p): New override
15896 that takes two types as argument.
15897 * internal-fn.c (while_direct): New macro.
15898 (expand_while_optab_fn): New function.
15899 (convert_optab_supported_p): Likewise.
15900 (direct_while_optab_supported_p): New macro.
15901 * wide-int.h (wi::udiv_ceil): New function.
15902 * tree-vectorizer.h (rgroup_masks): New structure.
15903 (vec_loop_masks): New typedef.
15904 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15905 and fully_masked_p.
15906 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15907 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15908 (vect_max_vf): New function.
15909 (slpeel_make_loop_iterate_ntimes): Delete.
15910 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15911 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15912 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15913 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15914 internal-fn.h, stor-layout.h and optabs-query.h.
15915 (vect_set_loop_mask): New function.
15916 (add_preheader_seq): Likewise.
15917 (add_header_seq): Likewise.
15918 (interleave_supported_p): Likewise.
15919 (vect_maybe_permute_loop_masks): Likewise.
15920 (vect_set_loop_masks_directly): Likewise.
15921 (vect_set_loop_condition_masked): Likewise.
15922 (vect_set_loop_condition_unmasked): New function, split out from
15923 slpeel_make_loop_iterate_ntimes.
15924 (slpeel_make_loop_iterate_ntimes): Rename to..
15925 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15926 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15927 (vect_do_peeling): Update call accordingly.
15928 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15930 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15931 mask_compare_type, can_fully_mask_p and fully_masked_p.
15932 (release_vec_loop_masks): New function.
15933 (_loop_vec_info): Use it to free the loop masks.
15934 (can_produce_all_loop_masks_p): New function.
15935 (vect_get_max_nscalars_per_iter): Likewise.
15936 (vect_verify_full_masking): Likewise.
15937 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15938 retries, and free the mask rgroups before retrying. Check loop-wide
15939 reasons for disallowing fully-masked loops. Make the final decision
15940 about whether use a fully-masked loop or not.
15941 (vect_estimate_min_profitable_iters): Do not assume that peeling
15942 for the number of iterations will be needed for fully-masked loops.
15943 (vectorizable_reduction): Disable fully-masked loops.
15944 (vectorizable_live_operation): Likewise.
15945 (vect_halve_mask_nunits): New function.
15946 (vect_double_mask_nunits): Likewise.
15947 (vect_record_loop_mask): Likewise.
15948 (vect_get_loop_mask): Likewise.
15949 (vect_transform_loop): Handle the case in which the final loop
15950 iteration might handle a partial vector. Call vect_set_loop_condition
15951 instead of slpeel_make_loop_iterate_ntimes.
15952 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15953 (check_load_store_masking): New function.
15954 (prepare_load_store_mask): Likewise.
15955 (vectorizable_store): Handle fully-masked loops.
15956 (vectorizable_load): Likewise.
15957 (supportable_widening_operation): Use vect_halve_mask_nunits for
15959 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15960 (vect_gen_while): New function.
15961 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15962 (aarch64_uqdec<mode>): New insn.
15964 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15965 Alan Hayward <alan.hayward@arm.com>
15966 David Sherwood <david.sherwood@arm.com>
15968 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15969 (reduc_xor_scal_optab): New optabs.
15970 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15971 (reduc_xor_scal_@var{m}): Document.
15972 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15973 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15974 internal functions.
15975 * fold-const-call.c (fold_const_call): Handle them.
15976 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15977 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15978 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15979 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15980 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15981 (UNSPEC_XORV): New unspecs.
15982 (optab): Add entries for them.
15983 (BITWISEV): New int iterator.
15984 (bit_reduc_op): New int attributes.
15986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15987 Alan Hayward <alan.hayward@arm.com>
15988 David Sherwood <david.sherwood@arm.com>
15990 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15991 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15992 * optabs.def (vec_shl_insert_optab): New optab.
15993 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15994 (duplicate_and_interleave): Likewise.
15995 * tree-vect-loop.c: Include internal-fn.h.
15996 (neutral_op_for_slp_reduction): New function, split out from
15997 get_initial_defs_for_reduction.
15998 (get_initial_def_for_reduction): Handle option 2 for variable-length
15999 vectors by loading the neutral value into a vector and then shifting
16000 the initial value into element 0.
16001 (get_initial_defs_for_reduction): Replace the code argument with
16002 the neutral value calculated by neutral_op_for_slp_reduction.
16003 Use gimple_build_vector for constant-length vectors.
16004 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16005 but the first group_size elements have a neutral value.
16006 Use duplicate_and_interleave otherwise.
16007 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16008 Update call to get_initial_defs_for_reduction. Handle SLP
16009 reductions for variable-length vectors by creating one vector
16010 result for each scalar result, with the elements associated
16011 with other scalar results stubbed out with the neutral value.
16012 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16013 Require IFN_VEC_SHL_INSERT for double reductions on
16014 variable-length vectors, or SLP reductions that have
16015 a neutral value. Require can_duplicate_and_interleave_p
16016 support for variable-length unchained SLP reductions if there
16017 is no neutral value, such as for MIN/MAX reductions. Also require
16018 the number of vector elements to be a multiple of the number of
16019 SLP statements when doing variable-length unchained SLP reductions.
16020 Update call to vect_create_epilog_for_reduction.
16021 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16022 and remove initial values.
16023 (duplicate_and_interleave): Make public.
16024 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16025 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16028 Alan Hayward <alan.hayward@arm.com>
16029 David Sherwood <david.sherwood@arm.com>
16031 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16032 (can_duplicate_and_interleave_p): New function.
16033 (vect_get_and_check_slp_defs): Take the vector of statements
16034 rather than just the current one. Remove excess parentheses.
16035 Restriction rejectinon of vect_constant_def and vect_external_def
16036 for variable-length vectors to boolean types, or types for which
16037 can_duplicate_and_interleave_p is false.
16038 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16039 (duplicate_and_interleave): New function.
16040 (vect_get_constant_vectors): Use gimple_build_vector for
16041 constant-length vectors and suitable variable-length constant
16042 vectors. Use duplicate_and_interleave for other variable-length
16043 vectors. Don't defer the update when inserting new statements.
16045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16046 Alan Hayward <alan.hayward@arm.com>
16047 David Sherwood <david.sherwood@arm.com>
16049 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16050 min_profitable_iters doesn't go negative.
16052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16053 Alan Hayward <alan.hayward@arm.com>
16054 David Sherwood <david.sherwood@arm.com>
16056 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16057 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16058 * optabs.def (vec_mask_load_lanes_optab): New optab.
16059 (vec_mask_store_lanes_optab): Likewise.
16060 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16061 (MASK_STORE_LANES): Likewise.
16062 * internal-fn.c (mask_load_lanes_direct): New macro.
16063 (mask_store_lanes_direct): Likewise.
16064 (expand_mask_load_optab_fn): Handle masked operations.
16065 (expand_mask_load_lanes_optab_fn): New macro.
16066 (expand_mask_store_optab_fn): Handle masked operations.
16067 (expand_mask_store_lanes_optab_fn): New macro.
16068 (direct_mask_load_lanes_optab_supported_p): Likewise.
16069 (direct_mask_store_lanes_optab_supported_p): Likewise.
16070 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16072 (vect_load_lanes_supported): Likewise.
16073 * tree-vect-data-refs.c (strip_conversion): New function.
16074 (can_group_stmts_p): Likewise.
16075 (vect_analyze_data_ref_accesses): Use it instead of checking
16076 for a pair of assignments.
16077 (vect_store_lanes_supported): Take a masked_p parameter.
16078 (vect_load_lanes_supported): Likewise.
16079 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16080 vect_store_lanes_supported and vect_load_lanes_supported.
16081 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16082 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16083 parameter. Don't allow gaps for masked accesses.
16084 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16085 and vect_load_lanes_supported.
16086 (get_load_store_type): Take a masked_p parameter and update
16087 call to get_group_load_store_type.
16088 (vectorizable_store): Update call to get_load_store_type.
16089 Handle IFN_MASK_STORE_LANES.
16090 (vectorizable_load): Update call to get_load_store_type.
16091 Handle IFN_MASK_LOAD_LANES.
16093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16094 Alan Hayward <alan.hayward@arm.com>
16095 David Sherwood <david.sherwood@arm.com>
16097 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16099 * config/aarch64/aarch64-protos.h
16100 (aarch64_sve_struct_memory_operand_p): Declare.
16101 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16102 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16103 (VPRED, vpred): Handle SVE structure modes.
16104 * config/aarch64/constraints.md (Utx): New constraint.
16105 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16106 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16107 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16108 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16109 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16110 structure modes. Split into pieces after RA.
16111 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16112 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16114 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16115 SVE structure modes.
16116 (aarch64_classify_address): Likewise.
16117 (sizetochar): Move earlier in file.
16118 (aarch64_print_operand): Handle SVE register lists.
16119 (aarch64_array_mode): New function.
16120 (aarch64_sve_struct_memory_operand_p): Likewise.
16121 (TARGET_ARRAY_MODE): Redefine.
16123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16124 Alan Hayward <alan.hayward@arm.com>
16125 David Sherwood <david.sherwood@arm.com>
16127 * target.def (array_mode): New target hook.
16128 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16129 * doc/tm.texi: Regenerate.
16130 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16131 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16132 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16133 targetm.array_mode.
16134 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16138 Alan Hayward <alan.hayward@arm.com>
16139 David Sherwood <david.sherwood@arm.com>
16141 * fold-const.c (fold_binary_loc): Check the argument types
16142 rather than the result type when testing for a vector operation.
16144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16146 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16147 * doc/tm.texi: Regenerate.
16149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16150 Alan Hayward <alan.hayward@arm.com>
16151 David Sherwood <david.sherwood@arm.com>
16153 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16154 (sve): Document new AArch64 extension.
16155 * doc/md.texi (w): Extend the description of the AArch64
16156 constraint to include SVE vectors.
16157 (Upl, Upa): Document new AArch64 predicate constraints.
16158 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16160 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16161 (msve-vector-bits=): New option.
16162 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16163 SVE when these are disabled.
16164 (sve): New extension.
16165 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16166 modes. Adjust their number of units based on aarch64_sve_vg.
16167 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16168 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16169 aarch64_addr_query_type.
16170 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16171 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16172 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16173 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16174 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16175 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16176 (aarch64_simd_imm_zero_p): Delete.
16177 (aarch64_check_zero_based_sve_index_immediate): Declare.
16178 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16179 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16180 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16181 (aarch64_sve_float_mul_immediate_p): Likewise.
16182 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16183 rather than an rtx.
16184 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16185 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16186 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16187 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16188 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16189 (aarch64_regmode_natural_size): Likewise.
16190 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16191 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16193 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16194 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16195 for VG and the SVE predicate registers.
16196 (V_ALIASES): Add a "z"-prefixed alias.
16197 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16198 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16199 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16200 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16201 (REG_CLASS_NAMES): Add entries for them.
16202 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16203 and the predicate registers.
16204 (aarch64_sve_vg): Declare.
16205 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16206 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16207 (REGMODE_NATURAL_SIZE): Define.
16208 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16210 * config/aarch64/aarch64.c: Include cfgrtl.h.
16211 (simd_immediate_info): Add a constructor for series vectors,
16212 and an associated step field.
16213 (aarch64_sve_vg): New variable.
16214 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16215 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16216 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16217 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16218 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16219 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16220 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16221 (aarch64_get_mask_mode): New functions.
16222 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16223 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16224 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16225 predicate modes and predicate registers. Explicitly restrict
16226 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16227 to store a vector mode if it is recognized by
16228 aarch64_classify_vector_mode.
16229 (aarch64_regmode_natural_size): New function.
16230 (aarch64_hard_regno_caller_save_mode): Return the original mode
16232 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16233 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16234 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16235 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16237 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16238 does not overlap dest if the function is frame-related. Handle
16240 (aarch64_split_add_offset): New function.
16241 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16242 them aarch64_add_offset.
16243 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16244 and update call to aarch64_sub_sp.
16245 (aarch64_add_cfa_expression): New function.
16246 (aarch64_expand_prologue): Pass extra temporary registers to the
16247 functions above. Handle the case in which we need to emit new
16248 DW_CFA_expressions for registers that were originally saved
16249 relative to the stack pointer, but now have to be expressed
16250 relative to the frame pointer.
16251 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16253 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16254 IP0 and IP1 values for SVE frames.
16255 (aarch64_expand_vec_series): New function.
16256 (aarch64_expand_sve_widened_duplicate): Likewise.
16257 (aarch64_expand_sve_const_vector): Likewise.
16258 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16259 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16260 into the register, rather than emitting a SET directly.
16261 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16262 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16263 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16264 (offset_9bit_signed_scaled_p): New functions.
16265 (aarch64_replicate_bitmask_imm): New function.
16266 (aarch64_bitmask_imm): Use it.
16267 (aarch64_cannot_force_const_mem): Reject expressions involving
16268 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16269 (aarch64_classify_index): Handle SVE indices, by requiring
16270 a plain register index with a scale that matches the element size.
16271 (aarch64_classify_address): Handle SVE addresses. Assert that
16272 the mode of the address is VOIDmode or an integer mode.
16273 Update call to aarch64_classify_symbol.
16274 (aarch64_classify_symbolic_expression): Update call to
16275 aarch64_classify_symbol.
16276 (aarch64_const_vec_all_in_range_p): New function.
16277 (aarch64_print_vector_float_operand): Likewise.
16278 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16279 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16280 and the FP immediates 1.0 and 0.5.
16281 (aarch64_print_address_internal): Handle SVE addresses.
16282 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16283 (aarch64_regno_regclass): Handle predicate registers.
16284 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16286 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16287 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16288 (aarch64_convert_sve_vector_bits): New function.
16289 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16290 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16291 rather than an rtx.
16292 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16293 Handle SVE vector and predicate modes. Accept VL-based constants
16294 that need only one temporary register, and VL offsets that require
16295 no temporary registers.
16296 (aarch64_conditional_register_usage): Mark the predicate registers
16297 as fixed if SVE isn't available.
16298 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16299 Return true for SVE vector and predicate modes.
16300 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16301 rather than an unsigned int. Handle SVE modes.
16302 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16304 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16306 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16307 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16308 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16309 (aarch64_sve_float_mul_immediate_p): New functions.
16310 (aarch64_sve_valid_immediate): New function.
16311 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16312 Explicitly reject structure modes. Check for INDEX constants.
16313 Handle PTRUE and PFALSE constants.
16314 (aarch64_check_zero_based_sve_index_immediate): New function.
16315 (aarch64_simd_imm_zero_p): Delete.
16316 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16317 vector modes. Accept constants in the range of CNT[BHWD].
16318 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16319 ask for an Advanced SIMD mode.
16320 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16321 (aarch64_simd_vector_alignment): Handle SVE predicates.
16322 (aarch64_vectorize_preferred_vector_alignment): New function.
16323 (aarch64_simd_vector_alignment_reachable): Use it instead of
16325 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16326 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16328 (MAX_VECT_LEN): Delete.
16329 (expand_vec_perm_d): Add a vec_flags field.
16330 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16331 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16332 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16334 (aarch64_evpc_rev): Rename to...
16335 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16336 (aarch64_evpc_rev_global): New function.
16337 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16338 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16340 (aarch64_evpc_sve_tbl): New function.
16341 (aarch64_expand_vec_perm_const_1): Update after rename of
16342 aarch64_evpc_rev. Handle SVE permutes too, trying
16343 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16344 than aarch64_evpc_tbl.
16345 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16346 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16347 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16348 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16349 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16350 (aarch64_expand_sve_vcond): New functions.
16351 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16352 of aarch64_vector_mode_p.
16353 (aarch64_dwarf_poly_indeterminate_value): New function.
16354 (aarch64_compute_pressure_classes): Likewise.
16355 (aarch64_can_change_mode_class): Likewise.
16356 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16357 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16358 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16359 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16360 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16361 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16362 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16363 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16365 (Dn, Dl, Dr): Accept const as well as const_vector.
16366 (Dz): Likewise. Compare against CONST0_RTX.
16367 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16368 of "vector" where appropriate.
16369 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16370 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16371 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16372 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16373 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16374 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16375 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16376 (v_int_equiv): Extend to SVE modes.
16377 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16379 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16380 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16381 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16382 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16383 (SVE_COND_FP_CMP): New int iterators.
16384 (perm_hilo): Handle the new unpack unspecs.
16385 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16387 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16388 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16389 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16390 (aarch64_equality_operator, aarch64_constant_vector_operand)
16391 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16392 (aarch64_sve_nonimmediate_operand): Likewise.
16393 (aarch64_sve_general_operand): Likewise.
16394 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16395 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16396 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16397 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16398 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16399 (aarch64_sve_float_arith_immediate): Likewise.
16400 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16401 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16402 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16403 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16404 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16405 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16406 (aarch64_sve_float_arith_operand): Likewise.
16407 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16408 (aarch64_sve_float_mul_operand): Likewise.
16409 (aarch64_sve_vec_perm_operand): Likewise.
16410 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16411 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16412 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16413 as well as const_vector.
16414 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16415 in file. Use CONST0_RTX and CONSTM1_RTX.
16416 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16417 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16418 Use aarch64_simd_imm_zero.
16419 * config/aarch64/aarch64-sve.md: New file.
16420 * config/aarch64/aarch64.md: Include it.
16421 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16422 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16423 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16424 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16425 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16426 (sve): New attribute.
16427 (enabled): Disable instructions with the sve attribute unless
16429 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16430 aarch64_expand_mov_immediate.
16431 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16432 CNT[BHSD] immediates.
16433 (movti): Split CONST_POLY_INT moves into two halves.
16434 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16435 Split additions that need a temporary here if the destination
16436 is the stack pointer.
16437 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16438 (*add<mode>3_poly_1): New instruction.
16439 (set_clobber_cc): New expander.
16441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16443 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16444 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16445 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16446 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16447 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16448 Change innermode from fixed_mode_size to machine_mode.
16449 (simplify_subreg): Update call accordingly. Handle a constant-sized
16450 subreg of a variable-length CONST_VECTOR.
16452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16453 Alan Hayward <alan.hayward@arm.com>
16454 David Sherwood <david.sherwood@arm.com>
16456 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16457 (add_offset_to_base): New function, split out from...
16458 (create_mem_ref): ...here. When handling a scale other than 1,
16459 check first whether the address is valid without the offset.
16460 Add it into the base if so, leaving the index and scale as-is.
16462 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16465 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16466 fold_for_warn before checking if arg2 is INTEGER_CST.
16468 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16470 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16471 (store_multiple_operation): Delete.
16472 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16473 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16474 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16475 guarded by TARGET_STRING.
16476 (rs6000_output_load_multiple): Delete.
16477 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16478 OPTION_MASK_STRING / TARGET_STRING handling.
16479 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16480 (const rs6000_opt_masks) <"string">: Change mask to 0.
16481 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16482 (MASK_STRING): Delete.
16483 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16485 (load_multiple): Delete.
16492 (store_multiple): Delete.
16499 (movmemsi_8reg): Delete.
16500 (corresponding unnamed define_insn): Delete.
16501 (movmemsi_6reg): Delete.
16502 (corresponding unnamed define_insn): Delete.
16503 (movmemsi_4reg): Delete.
16504 (corresponding unnamed define_insn): Delete.
16505 (movmemsi_2reg): Delete.
16506 (corresponding unnamed define_insn): Delete.
16507 (movmemsi_1reg): Delete.
16508 (corresponding unnamed define_insn): Delete.
16509 * config/rs6000/rs6000.opt (mno-string): New.
16510 (mstring): Replace by deprecation warning stub.
16511 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16513 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16515 * regrename.c (regrename_do_replace): If replacing the same
16516 reg multiple times, try to reuse last created gen_raw_REG.
16519 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16520 main to workaround a bug in GDB.
16522 2018-01-12 Tom de Vries <tom@codesourcery.com>
16525 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16527 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16529 PR rtl-optimization/80481
16530 * ira-color.c (get_cap_member): New function.
16531 (allocnos_conflict_by_live_ranges_p): Use it.
16532 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16533 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16535 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16538 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16539 (*saddl_se_1): Ditto.
16540 (*ssubsi_1): Ditto.
16541 (*ssubl_se_1): Ditto.
16543 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16545 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16546 rather than wi::to_widest for DR_INITs.
16547 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16548 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16549 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16551 (vect_analyze_group_access_1): Note that here.
16553 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16555 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16556 polynomial type sizes.
16558 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16560 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16561 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16562 (gimple_add_tmp_var): Likewise.
16564 2018-01-12 Martin Liska <mliska@suse.cz>
16566 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16567 (gimple_alloc_sizes): Likewise.
16568 (dump_gimple_statistics): Use PRIu64 in printf format.
16569 * gimple.h: Change uint64_t to int.
16571 2018-01-12 Martin Liska <mliska@suse.cz>
16573 * tree-core.h: Use uint64_t instead of int.
16574 * tree.c (tree_node_counts): Likewise.
16575 (tree_node_sizes): Likewise.
16576 (dump_tree_statistics): Use PRIu64 in printf format.
16578 2018-01-12 Martin Liska <mliska@suse.cz>
16580 * Makefile.in: As qsort_chk is implemented in vec.c, add
16581 vec.o to linkage of gencfn-macros.
16582 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16583 passing the info to record_node_allocation_statistics.
16584 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16586 * ggc-common.c (struct ggc_usage): Add operator== and use
16587 it in operator< and compare function.
16588 * mem-stats.h (struct mem_usage): Likewise.
16589 * vec.c (struct vec_usage): Remove operator< and compare
16590 function. Can be simply inherited.
16592 2018-01-12 Martin Jambor <mjambor@suse.cz>
16595 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16596 * tree-ssa-math-opts.c: Include domwalk.h.
16597 (convert_mult_to_fma_1): New function.
16598 (fma_transformation_info): New type.
16599 (fma_deferring_state): Likewise.
16600 (cancel_fma_deferring): New function.
16601 (result_of_phi): Likewise.
16602 (last_fma_candidate_feeds_initial_phi): Likewise.
16603 (convert_mult_to_fma): Added deferring logic, split actual
16604 transformation to convert_mult_to_fma_1.
16605 (math_opts_dom_walker): New type.
16606 (math_opts_dom_walker::after_dom_children): New method, body moved
16607 here from pass_optimize_widening_mul::execute, added deferring logic
16609 (pass_optimize_widening_mul::execute): Moved most of code to
16610 math_opts_dom_walker::after_dom_children.
16611 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16612 * config/i386/i386.c (ix86_option_override_internal): Added
16613 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16615 2018-01-12 Richard Biener <rguenther@suse.de>
16618 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16619 inline instance vars.
16621 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16624 * config/rx/rx.c (rx_is_restricted_memory_address):
16625 Handle SUBREG case.
16627 2018-01-12 Richard Biener <rguenther@suse.de>
16629 PR tree-optimization/80846
16630 * target.def (split_reduction): New target hook.
16631 * targhooks.c (default_split_reduction): New function.
16632 * targhooks.h (default_split_reduction): Declare.
16633 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16634 target requests first reduce vectors by combining low and high
16636 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16637 (get_vectype_for_scalar_type_and_size): Export.
16638 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16639 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16640 * doc/tm.texi: Regenerate.
16641 * config/i386/i386.c (ix86_split_reduction): Implement
16642 TARGET_VECTORIZE_SPLIT_REDUCTION.
16644 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16647 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16648 in PIC mode except for TARGET_VXWORKS_RTP.
16649 * config/sparc/sparc.c: Include cfgrtl.h.
16650 (TARGET_INIT_PIC_REG): Define.
16651 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16652 (sparc_pic_register_p): New predicate.
16653 (sparc_legitimate_address_p): Use it.
16654 (sparc_legitimize_pic_address): Likewise.
16655 (sparc_delegitimize_address): Likewise.
16656 (sparc_mode_dependent_address_p): Likewise.
16657 (gen_load_pcrel_sym): Remove 4th parameter.
16658 (load_got_register): Adjust call to above. Remove obsolete stuff.
16659 (sparc_expand_prologue): Do not call load_got_register here.
16660 (sparc_flat_expand_prologue): Likewise.
16661 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16662 (sparc_use_pseudo_pic_reg): New function.
16663 (sparc_init_pic_reg): Likewise.
16664 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16665 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16667 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16669 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16670 Add item for branch_cost.
16672 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16674 PR rtl-optimization/83565
16675 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16676 not extend the result to a larger mode for rotate operations.
16677 (num_sign_bit_copies1): Likewise.
16679 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16682 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16684 Use values-Xc.o for -pedantic.
16685 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16687 2018-01-12 Martin Liska <mliska@suse.cz>
16690 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16692 (possible_polymorphic_call_targets): Use it.
16693 (ipa_devirt): Likewise.
16695 2018-01-12 Martin Liska <mliska@suse.cz>
16697 * profile-count.h (enum profile_quality): Use 0 as invalid
16698 enum value of profile_quality.
16700 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16702 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16703 -mext-string options.
16705 2018-01-12 Richard Biener <rguenther@suse.de>
16707 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16708 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16709 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16711 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16713 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16715 * configure.ac (--with-long-double-format): Add support for the
16716 configuration option to change the default long double format on
16718 * config.gcc (powerpc*-linux*-*): Likewise.
16719 * configure: Regenerate.
16720 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16721 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16722 used without modification.
16724 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16726 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16727 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16728 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16729 MISC_BUILTIN_SPEC_BARRIER.
16730 (rs6000_init_builtins): Likewise.
16731 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16733 (speculation_barrier): New define_insn.
16734 * doc/extend.texi: Document __builtin_speculation_barrier.
16736 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16739 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16740 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16741 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16743 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16744 integral modes instead of "ss" and "sd".
16745 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16746 vectors with 32-bit and 64-bit elements.
16747 (vecdupssescalarmodesuffix): New mode attribute.
16748 (vec_dup<mode>): Use it.
16750 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16753 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16754 frame if argument is passed on stack.
16756 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16759 * ree.c (combine_reaching_defs): Optimize also
16760 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16761 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16763 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16765 PR middle-end/83189
16766 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16768 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16770 PR middle-end/83718
16771 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16772 after they are computed.
16774 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16776 PR tree-optimization/83695
16777 * gimple-loop-linterchange.cc
16778 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16779 reset cached scev information after interchange.
16780 (pass_linterchange::execute): Remove call to scev_reset_htab.
16782 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16784 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16785 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16786 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16787 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16788 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16789 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16790 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16791 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16792 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16793 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16794 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16795 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16796 (V_lane_reg): Likewise.
16797 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16799 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16800 (vfmal_lane_low<mode>_intrinsic,
16801 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16802 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16803 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16804 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16805 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16806 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16808 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16810 * config/arm/arm-cpus.in (fp16fml): New feature.
16811 (ALL_SIMD): Add fp16fml.
16812 (armv8.2-a): Add fp16fml as an option.
16813 (armv8.3-a): Likewise.
16814 (armv8.4-a): Add fp16fml as part of fp16.
16815 * config/arm/arm.h (TARGET_FP16FML): Define.
16816 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16818 * config/arm/arm-modes.def (V2HF): Define.
16819 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16820 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16821 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16822 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16823 vfmsl_low, vfmsl_high): New set of builtins.
16824 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16825 (vfml_op): New code attribute.
16826 (VFMLHALVES): New int iterator.
16827 (VFML, VFMLSEL): New mode attributes.
16828 (V_reg): Define mapping for V2HF.
16829 (V_hi, V_lo): New mode attributes.
16830 (VF_constraint): Likewise.
16831 (vfml_half, vfml_half_selector): New int attributes.
16832 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16834 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16835 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16837 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16838 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16839 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16840 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16842 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16843 Document new effective target and option set.
16845 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16847 * config/arm/arm-cpus.in (armv8_4): New feature.
16848 (ARMv8_4a): New fgroup.
16849 (armv8.4-a): New arch.
16850 * config/arm/arm-tables.opt: Regenerate.
16851 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16852 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16853 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16854 Add matching rules for -march=armv8.4-a and extensions.
16855 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16857 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16860 * config/rx/rx.md (BW): New mode attribute.
16861 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16863 2018-01-11 Richard Biener <rguenther@suse.de>
16865 PR tree-optimization/83435
16866 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16867 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16868 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16870 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16871 Alan Hayward <alan.hayward@arm.com>
16872 David Sherwood <david.sherwood@arm.com>
16874 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16876 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16877 (aarch64_print_address_internal): Use it to check for a zero offset.
16879 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16880 Alan Hayward <alan.hayward@arm.com>
16881 David Sherwood <david.sherwood@arm.com>
16883 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16884 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16885 Return a poly_int64 rather than a HOST_WIDE_INT.
16886 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16887 rather than a HOST_WIDE_INT.
16888 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16889 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16890 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16891 final_offset from HOST_WIDE_INT to poly_int64.
16892 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16893 to_constant when getting the number of units in an Advanced SIMD
16895 (aarch64_builtin_vectorized_function): Check for a constant number
16897 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16899 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16900 attribute instead of GET_MODE_NUNITS.
16901 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16902 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16903 GET_MODE_SIZE for fixed-size registers.
16904 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16905 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16906 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16907 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16908 (aarch64_print_operand, aarch64_print_address_internal)
16909 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16910 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16911 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16912 Handle polynomial GET_MODE_SIZE.
16913 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16914 wider than SImode without modification.
16915 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16916 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16917 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16918 passing and returning SVE modes.
16919 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16920 rather than GEN_INT.
16921 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16922 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16923 (aarch64_allocate_and_probe_stack_space): Likewise.
16924 (aarch64_layout_frame): Cope with polynomial offsets.
16925 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16926 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16927 polynomial offsets.
16928 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16929 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16930 poly_int64 rather than a HOST_WIDE_INT.
16931 (aarch64_get_separate_components, aarch64_process_components)
16932 (aarch64_expand_prologue, aarch64_expand_epilogue)
16933 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16934 (aarch64_anchor_offset): New function, split out from...
16935 (aarch64_legitimize_address): ...here.
16936 (aarch64_builtin_vectorization_cost): Handle polynomial
16937 TYPE_VECTOR_SUBPARTS.
16938 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16940 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16941 number of elements from the PARALLEL rather than the mode.
16942 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16943 rather than GET_MODE_BITSIZE.
16944 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16945 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16946 (aarch64_expand_vec_perm_const_1): Handle polynomial
16947 d->perm.length () and d->perm elements.
16948 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16949 Apply to_constant to d->perm elements.
16950 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16951 polynomial CONST_VECTOR_NUNITS.
16952 (aarch64_move_pointer): Take amount as a poly_int64 rather
16954 (aarch64_progress_pointer): Avoid temporary variable.
16955 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16956 the mode attribute instead of GET_MODE.
16958 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16959 Alan Hayward <alan.hayward@arm.com>
16960 David Sherwood <david.sherwood@arm.com>
16962 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16963 x exists before using it.
16964 (aarch64_add_constant_internal): Rename to...
16965 (aarch64_add_offset_1): ...this. Replace regnum with separate
16966 src and dest rtxes. Handle the case in which they're different,
16967 including when the offset is zero. Replace scratchreg with an rtx.
16968 Use 2 additions if there is no spare register into which we can
16969 move a 16-bit constant.
16970 (aarch64_add_constant): Delete.
16971 (aarch64_add_offset): Replace reg with separate src and dest
16972 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16973 Use aarch64_add_offset_1.
16974 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16975 an rtx rather than an int. Take the delta as a poly_int64
16976 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16977 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16978 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16979 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16980 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16981 and aarch64_add_sp.
16982 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16983 aarch64_add_constant.
16985 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16987 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16988 Use scalar_float_mode.
16990 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16992 * config/aarch64/aarch64-simd.md
16993 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16994 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16995 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16996 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16997 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16998 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16999 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17000 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17001 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17002 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17004 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17007 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17008 targ_options->x_arm_arch_string is non NULL.
17010 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17012 * config/aarch64/aarch64.h
17013 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17015 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17018 * expmed.c (emit_store_flag_force): Swap if const op0
17019 and change VOIDmode to mode of op0.
17021 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17023 PR rtl-optimization/83761
17024 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17025 than bytes to mode_for_size.
17027 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17029 PR middle-end/83189
17030 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17031 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17034 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17036 PR middle-end/83575
17037 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17038 when in layout mode.
17039 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17040 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17043 2018-01-10 Michael Collison <michael.collison@arm.com>
17045 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17046 * config/aarch64/aarch64-option-extension.def: Add
17047 AARCH64_OPT_EXTENSION of 'fp16fml'.
17048 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17049 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17050 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17051 * config/aarch64/constraints.md (Ui7): New constraint.
17052 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17053 (VFMLA_SEL_W): Ditto.
17056 (VFMLA16_LOW): New int iterator.
17057 (VFMLA16_HIGH): Ditto.
17058 (UNSPEC_FMLAL): New unspec.
17059 (UNSPEC_FMLSL): Ditto.
17060 (UNSPEC_FMLAL2): Ditto.
17061 (UNSPEC_FMLSL2): Ditto.
17062 (f16mac): New code attribute.
17063 * config/aarch64/aarch64-simd-builtins.def
17064 (aarch64_fmlal_lowv2sf): Ditto.
17065 (aarch64_fmlsl_lowv2sf): Ditto.
17066 (aarch64_fmlalq_lowv4sf): Ditto.
17067 (aarch64_fmlslq_lowv4sf): Ditto.
17068 (aarch64_fmlal_highv2sf): Ditto.
17069 (aarch64_fmlsl_highv2sf): Ditto.
17070 (aarch64_fmlalq_highv4sf): Ditto.
17071 (aarch64_fmlslq_highv4sf): Ditto.
17072 (aarch64_fmlal_lane_lowv2sf): Ditto.
17073 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17074 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17075 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17076 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17077 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17078 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17079 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17080 (aarch64_fmlal_lane_highv2sf): Ditto.
17081 (aarch64_fmlsl_lane_highv2sf): Ditto.
17082 (aarch64_fmlal_laneq_highv2sf): Ditto.
17083 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17084 (aarch64_fmlalq_lane_highv4sf): Ditto.
17085 (aarch64_fmlsl_lane_highv4sf): Ditto.
17086 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17087 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17088 * config/aarch64/aarch64-simd.md:
17089 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17090 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17091 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17092 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17093 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17094 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17095 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17096 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17097 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17098 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17099 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17100 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17101 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17102 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17103 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17104 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17105 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17106 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17107 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17108 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17109 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17110 (vfmlsl_low_u32): Ditto.
17111 (vfmlalq_low_u32): Ditto.
17112 (vfmlslq_low_u32): Ditto.
17113 (vfmlal_high_u32): Ditto.
17114 (vfmlsl_high_u32): Ditto.
17115 (vfmlalq_high_u32): Ditto.
17116 (vfmlslq_high_u32): Ditto.
17117 (vfmlal_lane_low_u32): Ditto.
17118 (vfmlsl_lane_low_u32): Ditto.
17119 (vfmlal_laneq_low_u32): Ditto.
17120 (vfmlsl_laneq_low_u32): Ditto.
17121 (vfmlalq_lane_low_u32): Ditto.
17122 (vfmlslq_lane_low_u32): Ditto.
17123 (vfmlalq_laneq_low_u32): Ditto.
17124 (vfmlslq_laneq_low_u32): Ditto.
17125 (vfmlal_lane_high_u32): Ditto.
17126 (vfmlsl_lane_high_u32): Ditto.
17127 (vfmlal_laneq_high_u32): Ditto.
17128 (vfmlsl_laneq_high_u32): Ditto.
17129 (vfmlalq_lane_high_u32): Ditto.
17130 (vfmlslq_lane_high_u32): Ditto.
17131 (vfmlalq_laneq_high_u32): Ditto.
17132 (vfmlslq_laneq_high_u32): Ditto.
17133 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17134 (AARCH64_FL_FOR_ARCH8_4): New.
17135 (AARCH64_ISA_F16FML): New ISA flag.
17136 (TARGET_F16FML): New feature flag for fp16fml.
17137 (doc/invoke.texi): Document new fp16fml option.
17139 2018-01-10 Michael Collison <michael.collison@arm.com>
17141 * config/aarch64/aarch64-builtins.c:
17142 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17143 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17144 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17145 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17146 (AARCH64_ISA_SHA3): New ISA flag.
17147 (TARGET_SHA3): New feature flag for sha3.
17148 * config/aarch64/iterators.md (sha512_op): New int attribute.
17149 (CRYPTO_SHA512): New int iterator.
17150 (UNSPEC_SHA512H): New unspec.
17151 (UNSPEC_SHA512H2): Ditto.
17152 (UNSPEC_SHA512SU0): Ditto.
17153 (UNSPEC_SHA512SU1): Ditto.
17154 * config/aarch64/aarch64-simd-builtins.def
17155 (aarch64_crypto_sha512hqv2di): New builtin.
17156 (aarch64_crypto_sha512h2qv2di): Ditto.
17157 (aarch64_crypto_sha512su0qv2di): Ditto.
17158 (aarch64_crypto_sha512su1qv2di): Ditto.
17159 (aarch64_eor3qv8hi): Ditto.
17160 (aarch64_rax1qv2di): Ditto.
17161 (aarch64_xarqv2di): Ditto.
17162 (aarch64_bcaxqv8hi): Ditto.
17163 * config/aarch64/aarch64-simd.md:
17164 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17165 (aarch64_crypto_sha512su0qv2di): Ditto.
17166 (aarch64_crypto_sha512su1qv2di): Ditto.
17167 (aarch64_eor3qv8hi): Ditto.
17168 (aarch64_rax1qv2di): Ditto.
17169 (aarch64_xarqv2di): Ditto.
17170 (aarch64_bcaxqv8hi): Ditto.
17171 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17172 (vsha512h2q_u64): Ditto.
17173 (vsha512su0q_u64): Ditto.
17174 (vsha512su1q_u64): Ditto.
17175 (veor3q_u16): Ditto.
17176 (vrax1q_u64): Ditto.
17177 (vxarq_u64): Ditto.
17178 (vbcaxq_u16): Ditto.
17179 * config/arm/types.md (crypto_sha512): New type attribute.
17180 (crypto_sha3): Ditto.
17181 (doc/invoke.texi): Document new sha3 option.
17183 2018-01-10 Michael Collison <michael.collison@arm.com>
17185 * config/aarch64/aarch64-builtins.c:
17186 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17187 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17188 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17189 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17190 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17191 (AARCH64_ISA_SM4): New ISA flag.
17192 (TARGET_SM4): New feature flag for sm4.
17193 * config/aarch64/aarch64-simd-builtins.def
17194 (aarch64_sm3ss1qv4si): Ditto.
17195 (aarch64_sm3tt1aq4si): Ditto.
17196 (aarch64_sm3tt1bq4si): Ditto.
17197 (aarch64_sm3tt2aq4si): Ditto.
17198 (aarch64_sm3tt2bq4si): Ditto.
17199 (aarch64_sm3partw1qv4si): Ditto.
17200 (aarch64_sm3partw2qv4si): Ditto.
17201 (aarch64_sm4eqv4si): Ditto.
17202 (aarch64_sm4ekeyqv4si): Ditto.
17203 * config/aarch64/aarch64-simd.md:
17204 (aarch64_sm3ss1qv4si): Ditto.
17205 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17206 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17207 (aarch64_sm4eqv4si): Ditto.
17208 (aarch64_sm4ekeyqv4si): Ditto.
17209 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17210 (sm3part_op): Ditto.
17211 (CRYPTO_SM3TT): Ditto.
17212 (CRYPTO_SM3PART): Ditto.
17213 (UNSPEC_SM3SS1): New unspec.
17214 (UNSPEC_SM3TT1A): Ditto.
17215 (UNSPEC_SM3TT1B): Ditto.
17216 (UNSPEC_SM3TT2A): Ditto.
17217 (UNSPEC_SM3TT2B): Ditto.
17218 (UNSPEC_SM3PARTW1): Ditto.
17219 (UNSPEC_SM3PARTW2): Ditto.
17220 (UNSPEC_SM4E): Ditto.
17221 (UNSPEC_SM4EKEY): Ditto.
17222 * config/aarch64/constraints.md (Ui2): New constraint.
17223 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17224 * config/arm/types.md (crypto_sm3): New type attribute.
17225 (crypto_sm4): Ditto.
17226 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17227 (vsm3tt1aq_u32): Ditto.
17228 (vsm3tt1bq_u32): Ditto.
17229 (vsm3tt2aq_u32): Ditto.
17230 (vsm3tt2bq_u32): Ditto.
17231 (vsm3partw1q_u32): Ditto.
17232 (vsm3partw2q_u32): Ditto.
17233 (vsm4eq_u32): Ditto.
17234 (vsm4ekeyq_u32): Ditto.
17235 (doc/invoke.texi): Document new sm4 option.
17237 2018-01-10 Michael Collison <michael.collison@arm.com>
17239 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17240 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17241 (AARCH64_FL_FOR_ARCH8_4): New.
17242 (AARCH64_FL_V8_4): New flag.
17243 (doc/invoke.texi): Document new armv8.4-a option.
17245 2018-01-10 Michael Collison <michael.collison@arm.com>
17247 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17248 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17249 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17250 * config/aarch64/aarch64-option-extension.def: Add
17251 AARCH64_OPT_EXTENSION of 'sha2'.
17252 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17253 (crypto): Disable sha2 and aes if crypto disabled.
17254 (crypto): Enable aes and sha2 if enabled.
17255 (simd): Disable sha2 and aes if simd disabled.
17256 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17258 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17259 (TARGET_SHA2): New feature flag for sha2.
17260 (TARGET_AES): New feature flag for aes.
17261 * config/aarch64/aarch64-simd.md:
17262 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17263 conditional on TARGET_AES.
17264 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17265 (aarch64_crypto_sha1hsi): Make pattern conditional
17267 (aarch64_crypto_sha1hv4si): Ditto.
17268 (aarch64_be_crypto_sha1hv4si): Ditto.
17269 (aarch64_crypto_sha1su1v4si): Ditto.
17270 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17271 (aarch64_crypto_sha1su0v4si): Ditto.
17272 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17273 (aarch64_crypto_sha256su0v4si): Ditto.
17274 (aarch64_crypto_sha256su1v4si): Ditto.
17275 (doc/invoke.texi): Document new aes and sha2 options.
17277 2018-01-10 Martin Sebor <msebor@redhat.com>
17279 PR tree-optimization/83781
17280 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17283 2018-01-11 Martin Sebor <msebor@gmail.com>
17284 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17286 PR tree-optimization/83501
17287 PR tree-optimization/81703
17289 * tree-ssa-strlen.c (get_string_cst): Rename...
17290 (get_string_len): ...to this. Handle global constants.
17291 (handle_char_store): Adjust.
17293 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17294 Jim Wilson <jimw@sifive.com>
17296 * config/riscv/riscv-protos.h (riscv_output_return): New.
17297 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17298 (riscv_attribute_table, riscv_output_return),
17299 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17300 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17301 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17303 (riscv_expand_prologue): Add early return for naked function.
17304 (riscv_expand_epilogue): Likewise.
17305 (riscv_function_ok_for_sibcall): Return false for naked function.
17306 (riscv_set_current_function): New.
17307 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17308 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17309 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17310 * doc/extend.texi (RISC-V Function Attributes): New.
17312 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17314 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17315 check for 128-bit long double before checking TCmode.
17316 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17317 128-bit long doubles before checking TFmode or TCmode.
17318 (FLOAT128_IBM_P): Likewise.
17320 2018-01-10 Martin Sebor <msebor@redhat.com>
17322 PR tree-optimization/83671
17323 * builtins.c (c_strlen): Unconditionally return zero for the empty
17325 Use -Warray-bounds for warnings.
17326 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17327 for non-constant array indices with COMPONENT_REF, arrays of
17328 arrays, and pointers to arrays.
17329 (gimple_fold_builtin_strlen): Determine and set length range for
17330 non-constant character arrays.
17332 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17334 PR middle-end/81897
17335 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17338 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17340 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17342 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17345 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17346 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17347 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17348 indexed_or_indirect_operand predicate.
17349 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17350 (*vsx_le_perm_load_v8hi): Likewise.
17351 (*vsx_le_perm_load_v16qi): Likewise.
17352 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17353 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17354 (*vsx_le_perm_store_v8hi): Likewise.
17355 (*vsx_le_perm_store_v16qi): Likewise.
17356 (eight unnamed splitters): Likewise.
17358 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17360 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17361 * config/rs6000/emmintrin.h: Likewise.
17362 * config/rs6000/mmintrin.h: Likewise.
17363 * config/rs6000/xmmintrin.h: Likewise.
17365 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17368 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17370 * tree.c (tree_nop_conversion): Return true for location wrapper
17372 (maybe_wrap_with_location): New function.
17373 (selftest::check_strip_nops): New function.
17374 (selftest::test_location_wrappers): New function.
17375 (selftest::tree_c_tests): Call it.
17376 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17377 (maybe_wrap_with_location): New decl.
17378 (EXPR_LOCATION_WRAPPER_P): New macro.
17379 (location_wrapper_p): New inline function.
17380 (tree_strip_any_location_wrapper): New inline function.
17382 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17385 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17386 stack_realign_offset for the largest alignment of stack slot
17388 (ix86_find_max_used_stack_alignment): New function.
17389 (ix86_finalize_stack_frame_flags): Use it. Set
17390 max_used_stack_alignment if we don't realign stack.
17391 * config/i386/i386.h (machine_function): Add
17392 max_used_stack_alignment.
17394 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17396 * config/arm/arm.opt (-mbranch-cost): New option.
17397 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17400 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17403 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17404 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17406 2018-01-10 Richard Biener <rguenther@suse.de>
17409 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17410 early out so it also covers the case where we have a non-NULL
17413 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17415 PR tree-optimization/83753
17416 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17417 for non-strided grouped accesses if the number of elements is 1.
17419 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17422 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17423 * i386.h (TARGET_USE_GATHER): Define.
17424 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17426 2018-01-10 Martin Liska <mliska@suse.cz>
17429 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17430 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17432 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17433 CLEANUP_NO_PARTITIONING is not set.
17435 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17437 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17438 for vectors, as a partial revert of r254296.
17439 * rtl.h (const_vec_p): Delete.
17440 (const_vec_duplicate_p): Don't test for vector CONSTs.
17441 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17442 * expmed.c (make_tree): Likewise.
17445 * common.md (E, F): Use CONSTANT_P instead of checking for
17447 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17448 checking for CONST_VECTOR.
17450 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17452 PR middle-end/83575
17453 * predict.c (force_edge_cold): Handle in more sane way edges
17454 with no prediction.
17456 2018-01-09 Carl Love <cel@us.ibm.com>
17458 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17460 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17461 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17462 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17463 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17464 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17465 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17466 * config/rs6000/rs6000-protos.h: Add extern defition for
17467 rs6000_generate_float2_double_code.
17468 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17470 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17471 (float2_v2df): Add define_expand.
17473 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17476 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17477 op_mode in the force_to_mode call.
17479 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17481 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17482 instead of checking each element individually.
17483 (aarch64_evpc_uzp): Likewise.
17484 (aarch64_evpc_zip): Likewise.
17485 (aarch64_evpc_ext): Likewise.
17486 (aarch64_evpc_rev): Likewise.
17487 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17488 instead of checking each element individually. Return true without
17490 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17491 whether all selected elements come from the same input, instead of
17492 checking each element individually. Remove calls to gen_rtx_REG,
17493 start_sequence and end_sequence and instead assert that no rtl is
17496 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17498 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17499 order of HIGH and CONST checks.
17501 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17503 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17504 if the destination isn't an SSA_NAME.
17506 2018-01-09 Richard Biener <rguenther@suse.de>
17508 PR tree-optimization/83668
17509 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17511 (canonicalize_loop_form): ... here, renamed from ...
17512 (canonicalize_loop_closed_ssa_form): ... this and amended to
17513 swap successor edges for loop exit blocks to make us use
17514 the RPO order we need for initial schedule generation.
17516 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17518 PR tree-optimization/64811
17519 * match.pd: When optimizing comparisons with Inf, avoid
17520 introducing or losing exceptions from comparisons with NaN.
17522 2018-01-09 Martin Liska <mliska@suse.cz>
17525 * asan.c (shadow_mem_size): Add gcc_assert.
17527 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17529 Don't save registers in main().
17532 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17533 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17534 * config/avr/avr.c (avr_set_current_function): Don't error if
17535 naked, OS_task or OS_main are specified at the same time.
17536 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17538 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17540 * common/config/avr/avr-common.c (avr_option_optimization_table):
17541 Switch on -mmain-is-OS_task for optimizing compilations.
17543 2018-01-09 Richard Biener <rguenther@suse.de>
17545 PR tree-optimization/83572
17546 * graphite.c: Include cfganal.h.
17547 (graphite_transform_loops): Connect infinite loops to exit
17548 and remove fake edges at the end.
17550 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17552 * ipa-inline.c (edge_badness): Revert accidental checkin.
17554 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17557 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17558 symbols; not inline clones.
17560 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17563 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17564 hard registers. Formatting fixes.
17566 PR preprocessor/83722
17567 * gcc.c (try_generate_repro): Pass
17568 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17569 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17572 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17573 Kito Cheng <kito.cheng@gmail.com>
17575 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17576 (riscv_leaf_function_p): Delete.
17577 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17579 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17581 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17583 (do_ifelse): New function.
17584 (do_isel): New function.
17585 (do_sub3): New function.
17586 (do_add3): New function.
17587 (do_load_mask_compare): New function.
17588 (do_overlap_load_compare): New function.
17589 (expand_compare_loop): New function.
17590 (expand_block_compare): Call expand_compare_loop() when appropriate.
17591 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17592 option description.
17593 (-mblock-compare-inline-loop-limit): New option.
17595 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17598 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17599 Reverse order of second and third operands in first alternative.
17600 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17601 of first and second elements in UNSPEC_VPERMR vector.
17602 (altivec_expand_vec_perm_le): Likewise.
17604 2018-01-08 Jeff Law <law@redhat.com>
17606 PR rtl-optimizatin/81308
17607 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17608 (process_switch): If group_case_labels makes a change, then set
17610 (pass_convert_switch::execute): If a switch is converted, then
17611 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17613 PR rtl-optimization/81308
17614 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17617 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17619 PR target/83663 - Revert r255946
17620 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17621 generation for cases where splatting a value is not useful.
17622 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17623 across a vec_duplicate and a paradoxical subreg forming a vector
17624 mode to a vec_concat.
17626 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17628 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17629 -march=armv8.3-a variants.
17630 * config/arm/t-multilib: Likewise.
17631 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17633 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17635 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17637 (cceq_ior_compare_complement): Give it a name so I can use it, and
17638 change boolean_or_operator predicate to boolean_operator so it can
17639 be used to generate a crand.
17640 (eqne): New code iterator.
17641 (bd/bd_neg): New code_attrs.
17642 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17643 a single define_insn.
17644 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17645 decrement (bdnzt/bdnzf/bdzt/bdzf).
17646 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17647 with the new names of the branch decrement patterns, and added the
17648 names of the branch decrement conditional patterns.
17650 2018-01-08 Richard Biener <rguenther@suse.de>
17652 PR tree-optimization/83563
17653 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17656 2018-01-08 Richard Biener <rguenther@suse.de>
17658 PR middle-end/83713
17659 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17661 2018-01-08 Richard Biener <rguenther@suse.de>
17663 PR tree-optimization/83685
17664 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17665 references to abnormals.
17667 2018-01-08 Richard Biener <rguenther@suse.de>
17670 * dwarf2out.c (output_indirect_strings): Handle empty
17671 skeleton_debug_str_hash.
17672 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17674 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17676 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17677 (emit_store_direct): Likewise.
17678 (arc_trampoline_adjust_address): Likewise.
17679 (arc_asm_trampoline_template): New function.
17680 (arc_initialize_trampoline): Use asm_trampoline_template.
17681 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17682 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17683 * config/arc/arc.md (flush_icache): Delete pattern.
17685 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17687 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17688 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17691 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17694 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17695 by not USED_FOR_TARGET.
17696 (make_pass_resolve_sw_modes): Likewise.
17698 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17700 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17703 2018-01-08 Richard Biener <rguenther@suse.de>
17705 PR middle-end/83580
17706 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17708 2018-01-08 Richard Biener <rguenther@suse.de>
17710 PR middle-end/83517
17711 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17713 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17715 PR middle-end/81897
17716 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17717 basic blocks with a small number of successors.
17718 (convert_control_dep_chain_into_preds): Improve handling of
17720 (dump_predicates): Split apart into...
17721 (dump_pred_chain): ...here...
17722 (dump_pred_info): ...and here.
17723 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17724 (can_chain_union_be_invalidated_p): Improve check for invalidation
17726 (uninit_uses_cannot_happen): Avoid unnecessary if
17727 convert_control_dep_chain_into_preds yielded nothing.
17729 2018-01-06 Martin Sebor <msebor@redhat.com>
17731 PR tree-optimization/83640
17732 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17733 subtracting negative offset from size.
17734 (builtin_access::overlap): Adjust offset bounds of the access to fall
17735 within the size of the object if possible.
17737 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17739 PR rtl-optimization/83699
17740 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17741 extract_bit_field_as_subreg to cases in which the extracted
17742 value is also a vector.
17744 * lra-constraints.c (process_alt_operands): Test for the equivalence
17745 substitutions when detecting a possible reload cycle.
17747 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17750 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17751 by default if flag_selective_schedling{,2}. Formatting fixes.
17753 PR rtl-optimization/83682
17754 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17755 if it has non-VECTOR_MODE element mode.
17756 (vec_duplicate_p): Likewise.
17758 PR middle-end/83694
17759 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17760 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17762 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17765 * config/i386/i386-builtin.def
17766 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17767 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17768 Require also OPTION_MASK_ISA_AVX512F in addition to
17769 OPTION_MASK_ISA_GFNI.
17770 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17771 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17772 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17773 to OPTION_MASK_ISA_GFNI.
17774 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17775 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17776 OPTION_MASK_ISA_AVX512BW.
17777 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17778 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17779 addition to OPTION_MASK_ISA_GFNI.
17780 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17781 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17782 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17783 to OPTION_MASK_ISA_GFNI.
17784 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17785 a requirement for all ISAs rather than any of them with a few
17787 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17789 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17790 bitmasks to be enabled with 3 exceptions, instead of requiring any
17791 enabled ISA with lots of exceptions.
17792 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17793 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17794 Change avx512bw in isa attribute to avx512f.
17795 * config/i386/sgxintrin.h: Add license boilerplate.
17796 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17797 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17798 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17799 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17801 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17802 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17803 temporarily sse2 rather than sse if not enabled already.
17806 * config/i386/sse.md (VI248_VLBW): Rename to ...
17807 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17808 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17809 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17810 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17811 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17812 mode iterator instead of VI248_VLBW.
17814 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17816 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17817 (record_modified): Skip clobbers; add debug output.
17818 (param_change_prob): Use sreal frequencies.
17820 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17822 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17823 punt for user-aligned variables.
17825 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17827 * tree-chrec.c (chrec_contains_symbols): Return true for
17830 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17833 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17834 of (x|y) == x for BICS pattern.
17836 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17838 PR tree-optimization/83605
17839 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17840 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17843 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17845 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17846 * config/epiphany/rtems.h: New file.
17848 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17849 Uros Bizjak <ubizjak@gmail.com>
17852 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17853 QIreg_operand instead of register_operand predicate.
17854 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17855 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17856 comments instead of -fmitigate[-_]rop.
17858 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17861 * cgraphunit.c (symbol_table::compile): Switch to text_section
17862 before calling assembly_start debug hook.
17863 * run-rtl-passes.c (run_rtl_passes): Likewise.
17866 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17868 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17869 range_int_cst_p rather than !symbolic_range_p before calling
17870 extract_range_from_multiplicative_op_1.
17872 2018-01-04 Jeff Law <law@redhat.com>
17874 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17875 redundant test in assertion.
17877 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17879 * doc/rtl.texi: Document machine_mode wrapper classes.
17881 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17883 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17884 using tree_to_uhwi.
17886 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17888 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17889 the VEC_PERM_EXPR fold to fail.
17891 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17894 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17895 to switched_sections.
17897 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17900 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17901 test for d.testing.
17903 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17906 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17907 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17909 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17912 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17913 is BLKmode and bitpos not zero or mode change is needed.
17915 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17918 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17921 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17924 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17925 instead of MULT rtx. Update all corresponding splitters.
17926 (*saddl_se): Ditto.
17927 (*ssub<modesuffix>): Ditto.
17928 (*ssubl_se): Ditto.
17929 (*cmp_sadd_di): Update split patterns.
17930 (*cmp_sadd_si): Ditto.
17931 (*cmp_sadd_sidi): Ditto.
17932 (*cmp_ssub_di): Ditto.
17933 (*cmp_ssub_si): Ditto.
17934 (*cmp_ssub_sidi): Ditto.
17935 * config/alpha/predicates.md (const23_operand): New predicate.
17936 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17937 Look for ASHIFT, not MULT inner operand.
17938 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17940 2018-01-04 Martin Liska <mliska@suse.cz>
17942 PR gcov-profile/83669
17943 * gcov.c (output_intermediate_file): Add version to intermediate
17945 * doc/gcov.texi: Document new field 'version' in intermediate
17946 file format. Fix location of '-k' option of gcov command.
17948 2018-01-04 Martin Liska <mliska@suse.cz>
17951 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17953 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17955 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17957 2018-01-03 Martin Sebor <msebor@redhat.com>
17959 PR tree-optimization/83655
17960 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17961 checking calls with invalid arguments.
17963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17965 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17966 (vectorizable_mask_load_store): Delete.
17967 (vectorizable_call): Return false for masked loads and stores.
17968 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17969 instead of gimple_assign_rhs1.
17970 (vectorizable_load): Handle IFN_MASK_LOAD.
17971 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17975 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17977 (vectorizable_mask_load_store): ...here.
17978 (vectorizable_load): ...and here.
17980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17982 * tree-vect-stmts.c (vect_build_all_ones_mask)
17983 (vect_build_zero_merge_argument): New functions, split out from...
17984 (vectorizable_load): ...here.
17986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17988 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17990 (vectorizable_mask_load_store): ...here.
17991 (vectorizable_store): ...and here.
17993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17995 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17997 (vectorizable_mask_load_store): ...here.
17999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18001 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18002 (vect_model_store_cost): Take a vec_load_store_type instead of a
18004 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18005 (vect_model_store_cost): Take a vec_load_store_type instead of a
18007 (vectorizable_mask_load_store): Update accordingly.
18008 (vectorizable_store): Likewise.
18009 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18013 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18014 IFN_MASK_LOAD calls here rather than...
18015 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18018 Alan Hayward <alan.hayward@arm.com>
18019 David Sherwood <david.sherwood@arm.com>
18021 * expmed.c (extract_bit_field_1): For vector extracts,
18022 fall back to extract_bit_field_as_subreg if vec_extract
18025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18026 Alan Hayward <alan.hayward@arm.com>
18027 David Sherwood <david.sherwood@arm.com>
18029 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18030 they are variable or constant sized.
18031 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18032 slots for constant-sized data.
18034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18035 Alan Hayward <alan.hayward@arm.com>
18036 David Sherwood <david.sherwood@arm.com>
18038 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18039 handling COND_EXPRs with boolean comparisons, try to find a better
18040 basis for the mask type than the boolean itself.
18042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18044 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18045 is calculated and how it can be overridden.
18046 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18047 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18049 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18053 Alan Hayward <alan.hayward@arm.com>
18054 David Sherwood <david.sherwood@arm.com>
18056 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18057 Remove the mode argument.
18058 (aarch64_simd_valid_immediate): Remove the mode and inverse
18060 * config/aarch64/iterators.md (bitsize): New iterator.
18061 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18062 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18063 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18064 aarch64_simd_valid_immediate.
18065 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18066 (aarch64_reg_or_bic_imm): Likewise.
18067 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18068 with an insn_type enum and msl with a modifier_type enum.
18069 Replace element_width with a scalar_mode. Change the shift
18070 to unsigned int. Add constructors for scalar_float_mode and
18071 scalar_int_mode elements.
18072 (aarch64_vect_float_const_representable_p): Delete.
18073 (aarch64_can_const_movi_rtx_p)
18074 (aarch64_simd_scalar_immediate_valid_for_move)
18075 (aarch64_simd_make_constant): Update call to
18076 aarch64_simd_valid_immediate.
18077 (aarch64_advsimd_valid_immediate_hs): New function.
18078 (aarch64_advsimd_valid_immediate): Likewise.
18079 (aarch64_simd_valid_immediate): Remove mode and inverse
18080 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18081 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18082 and aarch64_float_const_representable_p on the result.
18083 (aarch64_output_simd_mov_immediate): Remove mode argument.
18084 Update call to aarch64_simd_valid_immediate and use of
18085 simd_immediate_info.
18086 (aarch64_output_scalar_simd_mov_immediate): Update call
18089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18090 Alan Hayward <alan.hayward@arm.com>
18091 David Sherwood <david.sherwood@arm.com>
18093 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18094 (mode_nunits): Likewise CONST_MODE_NUNITS.
18095 * machmode.def (ADJUST_NUNITS): Document.
18096 * genmodes.c (mode_data::need_nunits_adj): New field.
18097 (blank_mode): Update accordingly.
18098 (adj_nunits): New variable.
18099 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18101 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18102 listed in adj_nunits.
18103 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18104 listed in adj_nunits. Don't emit case statements for such modes.
18105 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18106 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18107 nothing if adj_nunits is nonnull.
18108 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18109 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18110 (emit_mode_fbit): Update use of print_maybe_const_decl.
18111 (emit_move_size): Likewise. Treat the array as non-const
18113 (emit_mode_adjustments): Handle adj_nunits.
18115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18117 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18118 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18119 (VECTOR_MODES): Use it.
18120 (make_vector_modes): Take the prefix as an argument.
18122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18123 Alan Hayward <alan.hayward@arm.com>
18124 David Sherwood <david.sherwood@arm.com>
18126 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18127 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18128 for MODE_VECTOR_BOOL.
18129 * machmode.def (VECTOR_BOOL_MODE): Document.
18130 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18131 (make_vector_bool_mode): New function.
18132 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18134 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18135 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18137 * stor-layout.c (int_mode_for_mode): Likewise.
18138 * tree.c (build_vector_type_for_mode): Likewise.
18139 * varasm.c (output_constant_pool_2): Likewise.
18140 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18141 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18142 for MODE_VECTOR_BOOL.
18143 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18144 of mode class checks.
18145 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18146 instead of a list of mode class checks.
18147 (expand_vector_scalar_condition): Likewise.
18148 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18151 Alan Hayward <alan.hayward@arm.com>
18152 David Sherwood <david.sherwood@arm.com>
18154 * machmode.h (mode_size): Change from unsigned short to
18156 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18157 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18158 or if measurement_type is not polynomial.
18159 (fixed_size_mode::includes_p): Check for constant-sized modes.
18160 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18161 return a poly_uint16 rather than an unsigned short.
18162 (emit_mode_size): Change the type of mode_size from unsigned short
18163 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18164 (emit_mode_adjustments): Cope with polynomial vector sizes.
18165 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18167 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18169 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18170 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18171 * caller-save.c (setup_save_areas): Likewise.
18172 (replace_reg_with_saved_mem): Likewise.
18173 * calls.c (emit_library_call_value_1): Likewise.
18174 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18175 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18176 (gen_lowpart_for_combine): Likewise.
18177 * convert.c (convert_to_integer_1): Likewise.
18178 * cse.c (equiv_constant, cse_insn): Likewise.
18179 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18180 (cselib_subst_to_values): Likewise.
18181 * dce.c (word_dce_process_block): Likewise.
18182 * df-problems.c (df_word_lr_mark_ref): Likewise.
18183 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18184 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18185 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18186 (rtl_for_decl_location): Likewise.
18187 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18188 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18189 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18190 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18191 (expand_expr_real_1): Likewise.
18192 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18193 (pad_below): Likewise.
18194 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18195 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18196 * ira.c (get_subreg_tracking_sizes): Likewise.
18197 * ira-build.c (ira_create_allocno_objects): Likewise.
18198 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18199 (ira_sort_regnos_for_alter_reg): Likewise.
18200 * ira-costs.c (record_operand_costs): Likewise.
18201 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18202 (resolve_simple_move): Likewise.
18203 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18204 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18205 (lra_constraints): Likewise.
18206 (CONST_POOL_OK_P): Reject variable-sized modes.
18207 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18208 (add_pseudo_to_slot, lra_spill): Likewise.
18209 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18210 * optabs-query.c (get_best_extraction_insn): Likewise.
18211 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18212 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18213 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18214 * recog.c (offsettable_address_addr_space_p): Likewise.
18215 * regcprop.c (maybe_mode_change): Likewise.
18216 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18217 * regrename.c (build_def_use): Likewise.
18218 * regstat.c (dump_reg_info): Likewise.
18219 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18220 (find_reloads, find_reloads_subreg_address): Likewise.
18221 * reload1.c (eliminate_regs_1): Likewise.
18222 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18223 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18224 (simplify_binary_operation_1, simplify_subreg): Likewise.
18225 * targhooks.c (default_function_arg_padding): Likewise.
18226 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18227 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18228 (verify_gimple_assign_ternary): Likewise.
18229 * tree-inline.c (estimate_move_cost): Likewise.
18230 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18231 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18232 (get_address_cost_ainc): Likewise.
18233 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18234 (vect_supportable_dr_alignment): Likewise.
18235 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18236 (vectorizable_reduction): Likewise.
18237 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18238 (vectorizable_operation, vectorizable_load): Likewise.
18239 * tree.c (build_same_sized_truth_vector_type): Likewise.
18240 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18241 * var-tracking.c (emit_note_insn_var_location): Likewise.
18242 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18243 (ADDR_VEC_ALIGN): Likewise.
18245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18246 Alan Hayward <alan.hayward@arm.com>
18247 David Sherwood <david.sherwood@arm.com>
18249 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18251 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18252 or if measurement_type is polynomial.
18253 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18254 * combine.c (make_extraction): Likewise.
18255 * dse.c (find_shift_sequence): Likewise.
18256 * dwarf2out.c (mem_loc_descriptor): Likewise.
18257 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18258 (extract_bit_field, extract_low_bits): Likewise.
18259 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18260 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18261 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18262 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18263 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18264 * reload.c (find_reloads): Likewise.
18265 * reload1.c (alter_reg): Likewise.
18266 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18267 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18268 * tree-if-conv.c (predicate_mem_writes): Likewise.
18269 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18270 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18271 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18272 * valtrack.c (dead_debug_insert_temp): Likewise.
18273 * varasm.c (mergeable_constant_section): Likewise.
18274 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18277 Alan Hayward <alan.hayward@arm.com>
18278 David Sherwood <david.sherwood@arm.com>
18280 * expr.c (expand_assignment): Cope with polynomial mode sizes
18281 when assigning to a CONCAT.
18283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18284 Alan Hayward <alan.hayward@arm.com>
18285 David Sherwood <david.sherwood@arm.com>
18287 * machmode.h (mode_precision): Change from unsigned short to
18289 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18291 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18292 or if measurement_type is not polynomial.
18293 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18294 in which the mode is already known to be a scalar_int_mode.
18295 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18296 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18298 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18299 for GET_MODE_PRECISION.
18300 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18301 for GET_MODE_PRECISION.
18302 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18304 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18305 (expand_field_assignment, make_extraction): Likewise.
18306 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18307 (get_last_value): Likewise.
18308 * convert.c (convert_to_integer_1): Likewise.
18309 * cse.c (cse_insn): Likewise.
18310 * expr.c (expand_expr_real_1): Likewise.
18311 * lra-constraints.c (simplify_operand_subreg): Likewise.
18312 * optabs-query.c (can_atomic_load_p): Likewise.
18313 * optabs.c (expand_atomic_load): Likewise.
18314 (expand_atomic_store): Likewise.
18315 * ree.c (combine_reaching_defs): Likewise.
18316 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18317 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18318 * tree.h (type_has_mode_precision_p): Likewise.
18319 * ubsan.c (instrument_si_overflow): Likewise.
18321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18322 Alan Hayward <alan.hayward@arm.com>
18323 David Sherwood <david.sherwood@arm.com>
18325 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18326 polynomial numbers of units.
18327 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18328 (valid_vector_subparts_p): New function.
18329 (build_vector_type): Remove temporary shim and take the number
18330 of units as a poly_uint64 rather than an int.
18331 (build_opaque_vector_type): Take the number of units as a
18332 poly_uint64 rather than an int.
18333 * tree.c (build_vector_from_ctor): Handle polynomial
18334 TYPE_VECTOR_SUBPARTS.
18335 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18336 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18337 (build_vector_from_val): If the number of units is variable,
18338 use build_vec_duplicate_cst for constant operands and
18339 VEC_DUPLICATE_EXPR otherwise.
18340 (make_vector_type): Remove temporary is_constant ().
18341 (build_vector_type, build_opaque_vector_type): Take the number of
18342 units as a poly_uint64 rather than an int.
18343 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18345 * cfgexpand.c (expand_debug_expr): Likewise.
18346 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18347 (store_constructor, expand_expr_real_1): Likewise.
18348 (const_scalar_mask_from_tree): Likewise.
18349 * fold-const-call.c (fold_const_reduction): Likewise.
18350 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18351 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18352 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18353 (fold_relational_const): Likewise.
18354 (native_interpret_vector): Likewise. Change the size from an
18355 int to an unsigned int.
18356 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18357 TYPE_VECTOR_SUBPARTS.
18358 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18359 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18360 duplicating a non-constant operand into a variable-length vector.
18361 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18362 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18363 * ipa-icf.c (sem_variable::equals): Likewise.
18364 * match.pd: Likewise.
18365 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18366 * print-tree.c (print_node): Likewise.
18367 * stor-layout.c (layout_type): Likewise.
18368 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18369 * tree-cfg.c (verify_gimple_comparison): Likewise.
18370 (verify_gimple_assign_binary): Likewise.
18371 (verify_gimple_assign_ternary): Likewise.
18372 (verify_gimple_assign_single): Likewise.
18373 * tree-pretty-print.c (dump_generic_node): Likewise.
18374 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18375 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18376 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18377 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18378 (vect_shift_permute_load_chain): Likewise.
18379 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18380 (expand_vector_condition, optimize_vector_constructor): Likewise.
18381 (lower_vec_perm, get_compute_type): Likewise.
18382 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18383 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18384 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18385 (vect_recog_mask_conversion_pattern): Likewise.
18386 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18387 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18388 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18389 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18390 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18391 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18392 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18393 (supportable_widening_operation): Likewise.
18394 (supportable_narrowing_operation): Likewise.
18395 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18397 * varasm.c (output_constant): Likewise.
18399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18400 Alan Hayward <alan.hayward@arm.com>
18401 David Sherwood <david.sherwood@arm.com>
18403 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18404 so that both the length == 3 and length != 3 cases set up their
18405 own permute vectors. Add comments explaining why we know the
18406 number of elements is constant.
18407 (vect_permute_load_chain): Likewise.
18409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18410 Alan Hayward <alan.hayward@arm.com>
18411 David Sherwood <david.sherwood@arm.com>
18413 * machmode.h (mode_nunits): Change from unsigned char to
18415 (ONLY_FIXED_SIZE_MODES): New macro.
18416 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18417 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18418 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18420 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18421 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18422 or if measurement_type is not polynomial.
18423 * genmodes.c (ZERO_COEFFS): New macro.
18424 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18426 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18427 Use ZERO_COEFFS when emitting initializers.
18428 * data-streamer.h (bp_pack_poly_value): New function.
18429 (bp_unpack_poly_value): Likewise.
18430 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18431 for GET_MODE_NUNITS.
18432 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18433 for GET_MODE_NUNITS.
18434 * tree.c (make_vector_type): Remove temporary shim and make
18435 the real function take the number of units as a poly_uint64
18436 rather than an int.
18437 (build_vector_type_for_mode): Handle polynomial nunits.
18438 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18439 * emit-rtl.c (const_vec_series_p_1): Likewise.
18440 (gen_rtx_CONST_VECTOR): Likewise.
18441 * fold-const.c (test_vec_duplicate_folding): Likewise.
18442 * genrecog.c (validate_pattern): Likewise.
18443 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18444 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18445 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18446 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18447 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18448 * rtlanal.c (subreg_get_info): Likewise.
18449 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18450 (vect_grouped_load_supported): Likewise.
18451 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18452 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18453 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18454 (simplify_const_unary_operation, simplify_binary_operation_1)
18455 (simplify_const_binary_operation, simplify_ternary_operation)
18456 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18457 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18458 instead of CONST_VECTOR_NUNITS.
18459 * varasm.c (output_constant_pool_2): Likewise.
18460 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18461 explicit-encoded elements in the XVEC for variable-length vectors.
18463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18465 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18468 Alan Hayward <alan.hayward@arm.com>
18469 David Sherwood <david.sherwood@arm.com>
18471 * coretypes.h (fixed_size_mode): Declare.
18472 (fixed_size_mode_pod): New typedef.
18473 * builtins.h (target_builtins::x_apply_args_mode)
18474 (target_builtins::x_apply_result_mode): Change type to
18475 fixed_size_mode_pod.
18476 * builtins.c (apply_args_size, apply_result_size, result_vector)
18477 (expand_builtin_apply_args_1, expand_builtin_apply)
18478 (expand_builtin_return): Update accordingly.
18480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18482 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18483 * cselib.c (cselib_hash_rtx): Likewise.
18484 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18485 CONST_VECTOR encoding.
18487 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18488 Jeff Law <law@redhat.com>
18491 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18492 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18493 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18494 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18497 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18498 explicitly probe *sp in a noreturn function if there were any callee
18499 register saves or frame pointer is needed.
18501 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18504 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18505 BLKmode for ternary, binary or unary expressions.
18508 * var-tracking.c (delete_vta_debug_insn): New inline function.
18509 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18510 insns from get_insns () to NULL instead of each bb separately.
18511 Use delete_vta_debug_insn. No longer static.
18512 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18513 delete_vta_debug_insns callers.
18514 * rtl.h (delete_vta_debug_insns): Declare.
18515 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18516 instead of variable_tracking_main.
18518 2018-01-03 Martin Sebor <msebor@redhat.com>
18520 PR tree-optimization/83603
18521 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18522 arguments past the endof the argument list in functions declared
18523 without a prototype.
18524 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18525 Avoid checking when arguments are null.
18527 2018-01-03 Martin Sebor <msebor@redhat.com>
18530 * doc/extend.texi (attribute const): Fix a typo.
18531 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18532 issuing -Wsuggest-attribute for void functions.
18534 2018-01-03 Martin Sebor <msebor@redhat.com>
18536 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18537 offset_int::from instead of wide_int::to_shwi.
18538 (maybe_diag_overlap): Remove assertion.
18539 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18540 * gimple-ssa-sprintf.c (format_directive): Same.
18541 (parse_directive): Same.
18542 (sprintf_dom_walker::compute_format_length): Same.
18543 (try_substitute_return_value): Same.
18545 2018-01-03 Jeff Law <law@redhat.com>
18547 PR middle-end/83654
18548 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18549 non-constant residual for zero at runtime and avoid probing in
18550 that case. Reorganize code for trailing problem to mirror handling
18553 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18555 PR tree-optimization/83501
18556 * tree-ssa-strlen.c (get_string_cst): New.
18557 (handle_char_store): Call get_string_cst.
18559 2018-01-03 Martin Liska <mliska@suse.cz>
18561 PR tree-optimization/83593
18562 * tree-ssa-strlen.c: Include tree-cfg.h.
18563 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18564 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18565 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18567 (strlen_dom_walker::before_dom_children): Call
18568 gimple_purge_dead_eh_edges. Dump tranformation with details
18570 (strlen_dom_walker::before_dom_children): Update call by adding
18571 new argument cleanup_eh.
18572 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18574 2018-01-03 Martin Liska <mliska@suse.cz>
18577 * cif-code.def (VARIADIC_THUNK): New enum value.
18578 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18581 2018-01-03 Jan Beulich <jbeulich@suse.com>
18583 * sse.md (mov<mode>_internal): Tighten condition for when to use
18584 vmovdqu<ssescalarsize> for TI and OI modes.
18586 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18588 Update copyright years.
18590 2018-01-03 Martin Liska <mliska@suse.cz>
18593 * ipa-visibility.c (function_and_variable_visibility): Skip
18594 functions with noipa attribure.
18596 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18598 * gcc.c (process_command): Update copyright notice dates.
18599 * gcov-dump.c (print_version): Ditto.
18600 * gcov.c (print_version): Ditto.
18601 * gcov-tool.c (print_version): Ditto.
18602 * gengtype.c (create_file): Ditto.
18603 * doc/cpp.texi: Bump @copying's copyright year.
18604 * doc/cppinternals.texi: Ditto.
18605 * doc/gcc.texi: Ditto.
18606 * doc/gccint.texi: Ditto.
18607 * doc/gcov.texi: Ditto.
18608 * doc/install.texi: Ditto.
18609 * doc/invoke.texi: Ditto.
18611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18613 * vector-builder.h (vector_builder::m_full_nelts): Change from
18614 unsigned int to poly_uint64.
18615 (vector_builder::full_nelts): Update prototype accordingly.
18616 (vector_builder::new_vector): Likewise.
18617 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18618 (vector_builder::operator ==): Likewise.
18619 (vector_builder::finalize): Likewise.
18620 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18621 Take the number of elements as a poly_uint64 rather than an
18623 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18624 from unsigned int to poly_uint64.
18625 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18626 (vec_perm_indices::new_vector): Likewise.
18627 (vec_perm_indices::length): Likewise.
18628 (vec_perm_indices::nelts_per_input): Likewise.
18629 (vec_perm_indices::input_nelts): Likewise.
18630 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18631 number of elements per input as a poly_uint64 rather than an
18632 unsigned int. Use the original encoding for variable-length
18633 vectors, rather than clamping each individual element.
18634 For the second and subsequent elements in each pattern,
18635 clamp the step and base before clamping their sum.
18636 (vec_perm_indices::series_p): Handle polynomial element counts.
18637 (vec_perm_indices::all_in_range_p): Likewise.
18638 (vec_perm_indices_to_tree): Likewise.
18639 (vec_perm_indices_to_rtx): Likewise.
18640 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18641 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18642 (tree_vector_builder::new_binary_operation): Handle polynomial
18643 element counts. Return false if we need to know the number
18644 of elements at compile time.
18645 * fold-const.c (fold_vec_perm): Punt if the number of elements
18646 isn't known at compile time.
18648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18650 * vec-perm-indices.h (vec_perm_builder): Change element type
18651 from HOST_WIDE_INT to poly_int64.
18652 (vec_perm_indices::element_type): Update accordingly.
18653 (vec_perm_indices::clamp): Handle polynomial element_types.
18654 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18655 (vec_perm_indices::all_in_range_p): Likewise.
18656 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18658 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18659 polynomial vec_perm_indices element types.
18660 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18661 * fold-const.c (fold_vec_perm): Likewise.
18662 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18663 * tree-vect-generic.c (lower_vec_perm): Likewise.
18664 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18665 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18666 element type to HOST_WIDE_INT.
18668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18669 Alan Hayward <alan.hayward@arm.com>
18670 David Sherwood <david.sherwood@arm.com>
18672 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18673 rather than an int. Use plus_constant.
18674 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18675 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18678 Alan Hayward <alan.hayward@arm.com>
18679 David Sherwood <david.sherwood@arm.com>
18681 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18682 a HOST_WIDE_INT to a poly_int64.
18684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18685 Alan Hayward <alan.hayward@arm.com>
18686 David Sherwood <david.sherwood@arm.com>
18688 * calls.c (load_register_parameters): Cope with polynomial
18689 mode sizes. Require a constant size for BLKmode parameters
18690 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18691 forces a parameter to be padded at the lsb end in order to
18692 fill a complete number of words, require the parameter size
18693 to be ordered wrt UNITS_PER_WORD.
18695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18696 Alan Hayward <alan.hayward@arm.com>
18697 David Sherwood <david.sherwood@arm.com>
18699 * reload1.c (spill_stack_slot_width): Change element type
18700 from unsigned int to poly_uint64_pod.
18701 (alter_reg): Treat mode sizes as polynomial.
18703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18704 Alan Hayward <alan.hayward@arm.com>
18705 David Sherwood <david.sherwood@arm.com>
18707 * reload.c (complex_word_subreg_p): New function.
18708 (reload_inner_reg_of_subreg, push_reload): Use it.
18710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18711 Alan Hayward <alan.hayward@arm.com>
18712 David Sherwood <david.sherwood@arm.com>
18714 * lra-constraints.c (process_alt_operands): Reject matched
18715 operands whose sizes aren't ordered.
18716 (match_reload): Refer to this check here.
18718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18719 Alan Hayward <alan.hayward@arm.com>
18720 David Sherwood <david.sherwood@arm.com>
18722 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18723 that the mode size is in the set {1, 2, 4, 8, 16}.
18725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18726 Alan Hayward <alan.hayward@arm.com>
18727 David Sherwood <david.sherwood@arm.com>
18729 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18730 Use plus_constant instead of gen_rtx_PLUS.
18732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18733 Alan Hayward <alan.hayward@arm.com>
18734 David Sherwood <david.sherwood@arm.com>
18736 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18737 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18738 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18739 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18740 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18741 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18742 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18743 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18744 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18745 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18747 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18748 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18749 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18750 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18751 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18752 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18753 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18754 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18755 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18756 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18758 * expr.c (emit_move_resolve_push): Treat the input and result
18759 of PUSH_ROUNDING as a poly_int64.
18760 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18761 (emit_push_insn): Likewise.
18762 * lra-eliminations.c (mark_not_eliminable): Likewise.
18763 * recog.c (push_operand): Likewise.
18764 * reload1.c (elimination_effects): Likewise.
18765 * rtlanal.c (nonzero_bits1): Likewise.
18766 * calls.c (store_one_arg): Likewise. Require the padding to be
18767 known at compile time.
18769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18770 Alan Hayward <alan.hayward@arm.com>
18771 David Sherwood <david.sherwood@arm.com>
18773 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18774 Use plus_constant instead of gen_rtx_PLUS.
18776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18777 Alan Hayward <alan.hayward@arm.com>
18778 David Sherwood <david.sherwood@arm.com>
18780 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18781 rather than an int.
18783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18784 Alan Hayward <alan.hayward@arm.com>
18785 David Sherwood <david.sherwood@arm.com>
18787 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18788 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18789 via stack temporaries. Treat the mode size as polynomial too.
18791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18792 Alan Hayward <alan.hayward@arm.com>
18793 David Sherwood <david.sherwood@arm.com>
18795 * expr.c (expand_expr_real_2): When handling conversions involving
18796 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18797 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18798 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18807 Alan Hayward <alan.hayward@arm.com>
18808 David Sherwood <david.sherwood@arm.com>
18810 * combine.c (can_change_dest_mode): Handle polynomial
18811 REGMODE_NATURAL_SIZE.
18812 * expmed.c (store_bit_field_1): Likewise.
18813 * expr.c (store_constructor): Likewise.
18814 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18815 and polynomial REGMODE_NATURAL_SIZE.
18816 (gen_lowpart_common): Likewise.
18817 * reginfo.c (record_subregs_of_mode): Likewise.
18818 * rtlanal.c (read_modify_subreg_p): Likewise.
18820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18821 Alan Hayward <alan.hayward@arm.com>
18822 David Sherwood <david.sherwood@arm.com>
18824 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18825 numbers of elements.
18827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18828 Alan Hayward <alan.hayward@arm.com>
18829 David Sherwood <david.sherwood@arm.com>
18831 * match.pd: Cope with polynomial numbers of vector elements.
18833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18834 Alan Hayward <alan.hayward@arm.com>
18835 David Sherwood <david.sherwood@arm.com>
18837 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18838 in a POINTER_PLUS_EXPR.
18840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18841 Alan Hayward <alan.hayward@arm.com>
18842 David Sherwood <david.sherwood@arm.com>
18844 * omp-simd-clone.c (simd_clone_subparts): New function.
18845 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18846 (ipa_simd_modify_function_body): Likewise.
18848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18849 Alan Hayward <alan.hayward@arm.com>
18850 David Sherwood <david.sherwood@arm.com>
18852 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18853 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18854 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18855 (expand_vector_condition, vector_element): Likewise.
18856 (subparts_gt): New function.
18857 (get_compute_type): Use subparts_gt.
18858 (count_type_subparts): Delete.
18859 (expand_vector_operations_1): Use subparts_gt instead of
18860 count_type_subparts.
18862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18863 Alan Hayward <alan.hayward@arm.com>
18864 David Sherwood <david.sherwood@arm.com>
18866 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18867 (vect_compile_time_alias): ...this new function. Do the calculation
18868 on poly_ints rather than trees.
18869 (vect_prune_runtime_alias_test_list): Update call accordingly.
18871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18872 Alan Hayward <alan.hayward@arm.com>
18873 David Sherwood <david.sherwood@arm.com>
18875 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18877 (vect_schedule_slp_instance): Likewise.
18879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18880 Alan Hayward <alan.hayward@arm.com>
18881 David Sherwood <david.sherwood@arm.com>
18883 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18884 constant and extern definitions for variable-length vectors.
18885 (vect_get_constant_vectors): Note that the number of units
18886 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18893 of units as polynomial. Choose between WIDE and NARROW based
18896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18897 Alan Hayward <alan.hayward@arm.com>
18898 David Sherwood <david.sherwood@arm.com>
18900 * tree-vect-stmts.c (simd_clone_subparts): New function.
18901 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18904 Alan Hayward <alan.hayward@arm.com>
18905 David Sherwood <david.sherwood@arm.com>
18907 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18908 vectors as polynomial. Use build_index_vector for
18909 IFN_GOMP_SIMD_LANE.
18911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18912 Alan Hayward <alan.hayward@arm.com>
18913 David Sherwood <david.sherwood@arm.com>
18915 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18916 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18917 for variable-length vectors.
18918 (vectorizable_mask_load_store): Treat the number of units as
18919 polynomial, asserting that it is constant if the condition has
18920 already been enforced.
18921 (vectorizable_store, vectorizable_load): Likewise.
18923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18924 Alan Hayward <alan.hayward@arm.com>
18925 David Sherwood <david.sherwood@arm.com>
18927 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18928 of units as polynomial. Punt if we can't tell at compile time
18929 which vector contains the final result.
18931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18932 Alan Hayward <alan.hayward@arm.com>
18933 David Sherwood <david.sherwood@arm.com>
18935 * tree-vect-loop.c (vectorizable_induction): Treat the number
18936 of units as polynomial. Punt on SLP inductions. Use an integer
18937 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18938 cast of such a series for variable-length floating-point
18941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18942 Alan Hayward <alan.hayward@arm.com>
18943 David Sherwood <david.sherwood@arm.com>
18945 * tree.h (build_index_vector): Declare.
18946 * tree.c (build_index_vector): New function.
18947 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18948 of units as polynomial, forcibly converting it to a constant if
18949 vectorizable_reduction has already enforced the condition.
18950 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18951 to create a {1,2,3,...} vector.
18952 (vectorizable_reduction): Treat the number of units as polynomial.
18953 Choose vectype_in based on the largest scalar element size rather
18954 than the smallest number of units. Enforce the restrictions
18957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18958 Alan Hayward <alan.hayward@arm.com>
18959 David Sherwood <david.sherwood@arm.com>
18961 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18962 number of units as polynomial.
18964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18965 Alan Hayward <alan.hayward@arm.com>
18966 David Sherwood <david.sherwood@arm.com>
18968 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18969 * target.def (autovectorize_vector_sizes): Return the vector sizes
18970 by pointer, using vector_sizes rather than a bitmask.
18971 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18972 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18973 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18975 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18976 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18977 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18978 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18979 * omp-general.c (omp_max_vf): Likewise.
18980 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18981 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18982 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18983 * tree-vect-slp.c (vect_slp_bb): Likewise.
18984 * doc/tm.texi: Regenerate.
18985 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18987 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18988 the vector size as a poly_uint64 rather than an unsigned int.
18989 (current_vector_size): Change from an unsigned int to a poly_uint64.
18990 (get_vectype_for_scalar_type): Update accordingly.
18991 * tree.h (build_truth_vector_type): Take the size and number of
18992 units as a poly_uint64 rather than an unsigned int.
18993 (build_vector_type): Add a temporary overload that takes
18994 the number of units as a poly_uint64 rather than an unsigned int.
18995 * tree.c (make_vector_type): Likewise.
18996 (build_truth_vector_type): Take the number of units as a poly_uint64
18997 rather than an unsigned int.
18999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19000 Alan Hayward <alan.hayward@arm.com>
19001 David Sherwood <david.sherwood@arm.com>
19003 * target.def (get_mask_mode): Take the number of units and length
19004 as poly_uint64s rather than unsigned ints.
19005 * targhooks.h (default_get_mask_mode): Update accordingly.
19006 * targhooks.c (default_get_mask_mode): Likewise.
19007 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19008 * doc/tm.texi: Regenerate.
19010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19011 Alan Hayward <alan.hayward@arm.com>
19012 David Sherwood <david.sherwood@arm.com>
19014 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19015 * omp-general.c (omp_max_vf): Likewise.
19016 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19017 (expand_omp_simd): Handle polynomial safelen.
19018 * omp-low.c (omplow_simd_context): Add a default constructor.
19019 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19020 (lower_rec_simd_input_clauses): Update accordingly.
19021 (lower_rec_input_clauses): Likewise.
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-vectorizer.h (vect_nunits_for_cost): New function.
19028 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19029 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19030 (vect_analyze_slp_cost): Likewise.
19031 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19032 (vect_model_load_cost): Likewise.
19034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19035 Alan Hayward <alan.hayward@arm.com>
19036 David Sherwood <david.sherwood@arm.com>
19038 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19039 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19040 from an unsigned int * to a poly_uint64_pod *.
19041 (calculate_unrolling_factor): New function.
19042 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19045 Alan Hayward <alan.hayward@arm.com>
19046 David Sherwood <david.sherwood@arm.com>
19048 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19049 from an unsigned int to a poly_uint64.
19050 (_loop_vec_info::slp_unrolling_factor): Likewise.
19051 (_loop_vec_info::vectorization_factor): Change from an int
19053 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19054 (vect_get_num_vectors): New function.
19055 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19056 (vect_get_num_copies): Use vect_get_num_vectors.
19057 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19058 to an unsigned int *.
19059 (vect_analyze_data_refs): Change min_vf from an int * to a
19061 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19062 than an unsigned HOST_WIDE_INT.
19063 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19064 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19065 to an unsigned int *.
19066 (vect_analyze_data_ref_dependences): Likewise.
19067 (vect_compute_data_ref_alignment): Handle polynomial vf.
19068 (vect_enhance_data_refs_alignment): Likewise.
19069 (vect_prune_runtime_alias_test_list): Likewise.
19070 (vect_shift_permute_load_chain): Likewise.
19071 (vect_supportable_dr_alignment): Likewise.
19072 (dependence_distance_ge_vf): Take the vectorization factor as a
19073 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19074 (vect_analyze_data_refs): Change min_vf from an int * to a
19076 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19077 vfm1 as a poly_uint64 rather than an int. Make the same change
19078 for the returned bound_scalar.
19079 (vect_gen_vector_loop_niters): Handle polynomial vf.
19080 (vect_do_peeling): Likewise. Update call to
19081 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19082 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19084 * tree-vect-loop.c (vect_determine_vectorization_factor)
19085 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19086 (vect_get_known_peeling_cost): Likewise.
19087 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19088 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19089 (vect_transform_loop): Likewise. Use the lowest possible VF when
19090 updating the upper bounds of the loop.
19091 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19092 rather than an int.
19093 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19094 polynomial unroll factors.
19095 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19096 (vect_make_slp_decision): Likewise.
19097 (vect_supported_load_permutation_p): Likewise, and polynomial
19099 (vect_analyze_slp_cost): Handle polynomial vf.
19100 (vect_slp_analyze_node_operations): Likewise.
19101 (vect_slp_analyze_bb_1): Likewise.
19102 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19103 than an unsigned HOST_WIDE_INT.
19104 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19105 (vectorizable_load): Handle polynomial vf.
19106 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19108 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19111 Alan Hayward <alan.hayward@arm.com>
19112 David Sherwood <david.sherwood@arm.com>
19114 * match.pd: Handle bit operations involving three constants
19115 and try to fold one pair.
19117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19119 * tree-vect-loop-manip.c: Include gimple-fold.h.
19120 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19121 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19122 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19123 Add a path that uses a step of VF instead of 1, but disable it
19125 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19126 and niters_no_overflow parameters. Update calls to
19127 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19128 Create a new SSA name if the latter choses to use a ste other
19129 than zero, and return it via niters_vector_mult_vf_var.
19130 * tree-vect-loop.c (vect_transform_loop): Update calls to
19131 vect_do_peeling, vect_gen_vector_loop_niters and
19132 slpeel_make_loop_iterate_ntimes.
19133 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19134 (vect_gen_vector_loop_niters): Update declarations after above changes.
19136 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19138 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19139 128-bit round to integer instructions.
19140 (ceil<mode>2): Likewise.
19141 (btrunc<mode>2): Likewise.
19142 (round<mode>2): Likewise.
19144 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19146 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19147 unaligned VSX load/store on P8/P9.
19148 (expand_block_clear): Allow the use of unaligned VSX
19149 load/store on P8/P9.
19151 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19153 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19155 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19156 swap associated with both a load and a store.
19158 2018-01-02 Andrew Waterman <andrew@sifive.com>
19160 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19161 * config/riscv/riscv.md (clear_cache): Use it.
19163 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19165 * web.c: Remove out-of-date comment.
19167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19169 * expr.c (fixup_args_size_notes): Check that any existing
19170 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19171 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19172 (emit_single_push_insn): ...here.
19174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19176 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19177 (const_vector_encoded_nelts): New function.
19178 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19179 (const_vector_int_elt, const_vector_elt): Declare.
19180 * emit-rtl.c (const_vector_int_elt_1): New function.
19181 (const_vector_elt): Likewise.
19182 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19183 of CONST_VECTOR_ELT.
19185 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19187 * expr.c: Include rtx-vector-builder.h.
19188 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19189 directly on the tree encoding.
19190 (const_vector_from_tree): Likewise.
19191 * optabs.c: Include rtx-vector-builder.h.
19192 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19193 sequence of "u" values.
19194 * vec-perm-indices.c: Include rtx-vector-builder.h.
19195 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19196 directly on the vec_perm_indices encoding.
19198 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19200 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19201 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19202 * rtx-vector-builder.h: New file.
19203 * rtx-vector-builder.c: Likewise.
19204 * rtl.h (rtx_def::u2): Add a const_vector field.
19205 (CONST_VECTOR_NPATTERNS): New macro.
19206 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19207 (CONST_VECTOR_DUPLICATE_P): Likewise.
19208 (CONST_VECTOR_STEPPED_P): Likewise.
19209 (CONST_VECTOR_ENCODED_ELT): Likewise.
19210 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19211 (unwrap_const_vec_duplicate): Likewise.
19212 (const_vec_series_p): Check for a non-duplicated vector encoding.
19213 Say that the function only returns true for integer vectors.
19214 * emit-rtl.c: Include rtx-vector-builder.h.
19215 (gen_const_vec_duplicate_1): Delete.
19216 (gen_const_vector): Call gen_const_vec_duplicate instead of
19217 gen_const_vec_duplicate_1.
19218 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19219 (gen_const_vec_duplicate): Use rtx_vector_builder.
19220 (gen_const_vec_series): Likewise.
19221 (gen_rtx_CONST_VECTOR): Likewise.
19222 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19223 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19224 Build a new vector rather than modifying a CONST_VECTOR in-place.
19225 (handle_special_swappables): Update call accordingly.
19226 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19227 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19228 Build a new vector rather than modifying a CONST_VECTOR in-place.
19229 (handle_special_swappables): Update call accordingly.
19231 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19233 * simplify-rtx.c (simplify_const_binary_operation): Use
19234 CONST_VECTOR_ELT instead of XVECEXP.
19236 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19238 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19239 the selector elements to be different from the data elements
19240 if the selector is a VECTOR_CST.
19241 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19242 ssizetype for the selector.
19244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19246 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19247 before testing each element individually.
19248 * tree-vect-generic.c (lower_vec_perm): Likewise.
19250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19252 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19253 * selftest-run-tests.c (selftest::run_tests): Call it.
19254 * vector-builder.h (vector_builder::operator ==): New function.
19255 (vector_builder::operator !=): Likewise.
19256 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19257 (vec_perm_indices::all_from_input_p): New function.
19258 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19259 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19260 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19261 instead of reading the VECTOR_CST directly. Detect whether both
19262 vector inputs are the same before constructing the vec_perm_indices,
19263 and update the number of inputs argument accordingly. Use the
19264 utility functions added above. Only construct sel2 if we need to.
19266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19268 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19269 the broadcast of the low byte.
19270 (expand_mult_highpart): Use an explicit encoding for the permutes.
19271 * optabs-query.c (can_mult_highpart_p): Likewise.
19272 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19273 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19274 (vectorizable_bswap): Likewise.
19275 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19276 explicit encoding for the power-of-2 permutes.
19277 (vect_permute_store_chain): Likewise.
19278 (vect_grouped_load_supported): Likewise.
19279 (vect_permute_load_chain): Likewise.
19281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19283 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19284 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19285 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19286 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19287 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19288 (vect_gen_perm_mask_any): Likewise.
19290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19292 * int-vector-builder.h: New file.
19293 * vec-perm-indices.h: Include int-vector-builder.h.
19294 (vec_perm_indices): Redefine as an int_vector_builder.
19295 (auto_vec_perm_indices): Delete.
19296 (vec_perm_builder): Redefine as a stand-alone class.
19297 (vec_perm_indices::vec_perm_indices): New function.
19298 (vec_perm_indices::clamp): Likewise.
19299 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19300 (vec_perm_indices::new_vector): New function.
19301 (vec_perm_indices::new_expanded_vector): Update for new
19302 vec_perm_indices class.
19303 (vec_perm_indices::rotate_inputs): New function.
19304 (vec_perm_indices::all_in_range_p): Operate directly on the
19305 encoded form, without computing elided elements.
19306 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19307 encoding. Update for new vec_perm_indices class.
19308 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19309 the given vec_perm_builder.
19310 (expand_vec_perm_var): Update vec_perm_builder constructor.
19311 (expand_mult_highpart): Use vec_perm_builder instead of
19312 auto_vec_perm_indices.
19313 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19314 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19315 or double series encoding as appropriate.
19316 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19317 vec_perm_indices instead of auto_vec_perm_indices.
19318 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19319 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19320 (vect_permute_store_chain): Likewise.
19321 (vect_grouped_load_supported): Likewise.
19322 (vect_permute_load_chain): Likewise.
19323 (vect_shift_permute_load_chain): Likewise.
19324 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19325 (vect_transform_slp_perm_load): Likewise.
19326 (vect_schedule_slp_instance): Likewise.
19327 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19328 (vectorizable_mask_load_store): Likewise.
19329 (vectorizable_bswap): Likewise.
19330 (vectorizable_store): Likewise.
19331 (vectorizable_load): Likewise.
19332 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19333 vec_perm_indices instead of auto_vec_perm_indices. Use
19334 tree_to_vec_perm_builder to read the vector from a tree.
19335 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19336 vec_perm_builder instead of a vec_perm_indices.
19337 (have_whole_vector_shift): Use vec_perm_builder and
19338 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19339 truncation to calc_vec_perm_mask_for_shift.
19340 (vect_create_epilog_for_reduction): Likewise.
19341 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19342 from auto_vec_perm_indices to vec_perm_indices.
19343 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19344 instead of changing individual elements.
19345 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19346 the vector in d.perm.
19347 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19348 from auto_vec_perm_indices to vec_perm_indices.
19349 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19350 instead of changing individual elements.
19351 (arm_vectorize_vec_perm_const): Use new_vector to install
19352 the vector in d.perm.
19353 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19354 Update vec_perm_builder constructor.
19355 (rs6000_expand_interleave): Likewise.
19356 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19357 (rs6000_expand_interleave): Likewise.
19359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19361 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19362 to qimode could truncate the indices.
19363 * optabs.c (expand_vec_perm_var): Likewise.
19365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19367 * Makefile.in (OBJS): Add vec-perm-indices.o.
19368 * vec-perm-indices.h: New file.
19369 * vec-perm-indices.c: Likewise.
19370 * target.h (vec_perm_indices): Replace with a forward class
19372 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19373 * optabs.h: Include vec-perm-indices.h.
19374 (expand_vec_perm): Delete.
19375 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19376 (expand_vec_perm_const): Declare.
19377 * target.def (vec_perm_const_ok): Replace with...
19378 (vec_perm_const): ...this new hook.
19379 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19380 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19381 * doc/tm.texi: Regenerate.
19382 * optabs.def (vec_perm_const): Delete.
19383 * doc/md.texi (vec_perm_const): Likewise.
19384 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19385 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19386 expand_vec_perm for constant permutation vectors. Assert that
19387 the mode of variable permutation vectors is the integer equivalent
19388 of the mode that is being permuted.
19389 * optabs-query.h (selector_fits_mode_p): Declare.
19390 * optabs-query.c: Include vec-perm-indices.h.
19391 (selector_fits_mode_p): New function.
19392 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19393 is defined, instead of checking whether the vec_perm_const_optab
19394 exists. Use targetm.vectorize.vec_perm_const instead of
19395 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19396 fit in the vector mode before using a variable permute.
19397 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19398 vec_perm_indices instead of an rtx.
19399 (expand_vec_perm): Replace with...
19400 (expand_vec_perm_const): ...this new function. Take the selector
19401 as a vec_perm_indices rather than an rtx. Also take the mode of
19402 the selector. Update call to shift_amt_for_vec_perm_mask.
19403 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19404 Use vec_perm_indices::new_expanded_vector to expand the original
19405 selector into bytes. Check whether the indices fit in the vector
19406 mode before using a variable permute.
19407 (expand_vec_perm_var): Make global.
19408 (expand_mult_highpart): Use expand_vec_perm_const.
19409 * fold-const.c: Includes vec-perm-indices.h.
19410 * tree-ssa-forwprop.c: Likewise.
19411 * tree-vect-data-refs.c: Likewise.
19412 * tree-vect-generic.c: Likewise.
19413 * tree-vect-loop.c: Likewise.
19414 * tree-vect-slp.c: Likewise.
19415 * tree-vect-stmts.c: Likewise.
19416 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19418 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19419 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19420 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19421 (aarch64_vectorize_vec_perm_const): ...this new function.
19422 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19423 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19424 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19425 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19426 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19427 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19428 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19430 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19431 check for NEON modes.
19432 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19433 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19434 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19435 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19437 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19438 the old VEC_PERM_CONST conditions.
19439 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19440 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19441 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19442 (ia64_vectorize_vec_perm_const_ok): Merge into...
19443 (ia64_vectorize_vec_perm_const): ...this new function.
19444 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19445 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19446 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19447 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19448 * config/mips/mips.c (mips_expand_vec_perm_const)
19449 (mips_vectorize_vec_perm_const_ok): Merge into...
19450 (mips_vectorize_vec_perm_const): ...this new function.
19451 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19452 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19453 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19454 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19455 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19456 (rs6000_expand_vec_perm_const): Delete.
19457 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19459 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19460 (altivec_expand_vec_perm_const_le): Take each operand individually.
19461 Operate on constant selectors rather than rtxes.
19462 (altivec_expand_vec_perm_const): Likewise. Update call to
19463 altivec_expand_vec_perm_const_le.
19464 (rs6000_expand_vec_perm_const): Delete.
19465 (rs6000_vectorize_vec_perm_const_ok): Delete.
19466 (rs6000_vectorize_vec_perm_const): New function.
19467 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19468 an element count and rtx array.
19469 (rs6000_expand_extract_even): Update call accordingly.
19470 (rs6000_expand_interleave): Likewise.
19471 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19472 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19473 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19474 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19475 (rs6000_expand_vec_perm_const): Delete.
19476 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19477 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19478 (altivec_expand_vec_perm_const_le): Take each operand individually.
19479 Operate on constant selectors rather than rtxes.
19480 (altivec_expand_vec_perm_const): Likewise. Update call to
19481 altivec_expand_vec_perm_const_le.
19482 (rs6000_expand_vec_perm_const): Delete.
19483 (rs6000_vectorize_vec_perm_const_ok): Delete.
19484 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19485 reference to the SPE evmerge intructions.
19486 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19487 an element count and rtx array.
19488 (rs6000_expand_extract_even): Update call accordingly.
19489 (rs6000_expand_interleave): Likewise.
19490 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19491 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19493 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19495 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19497 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19498 vector mode and that that mode matches the mode of the data
19500 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19501 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19502 directly using expand_vec_perm_1 when forcing selectors into
19504 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19508 * optabs-query.h (can_vec_perm_p): Delete.
19509 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19510 * optabs-query.c (can_vec_perm_p): Split into...
19511 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19512 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19513 particular selector is valid.
19514 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19515 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19516 (vect_grouped_load_supported): Likewise.
19517 (vect_shift_permute_load_chain): Likewise.
19518 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19519 (vect_transform_slp_perm_load): Likewise.
19520 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19521 (vectorizable_bswap): Likewise.
19522 (vect_gen_perm_mask_checked): Likewise.
19523 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19524 implementations of variable permutation vectors into account
19525 when deciding which selector to use.
19526 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19527 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19528 with a false third argument.
19529 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19530 to test whether the constant selector is valid and can_vec_perm_var_p
19531 to test whether a variable selector is valid.
19533 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19535 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19536 * optabs-query.c (can_vec_perm_p): Likewise.
19537 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19538 instead of vec_perm_indices.
19539 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19540 (vect_gen_perm_mask_checked): Likewise,
19541 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19542 (vect_gen_perm_mask_checked): Likewise,
19544 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19546 * optabs-query.h (qimode_for_vec_perm): Declare.
19547 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19548 (qimode_for_vec_perm): ...this new function.
19549 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19551 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19553 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19554 does not have a conditional at the top.
19556 2018-01-02 Richard Biener <rguenther@suse.de>
19558 * ipa-inline.c (big_speedup_p): Fix expression.
19560 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19563 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19566 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19570 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19571 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19572 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19573 cond_taken_branch_cost 3->4.
19575 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19577 PR tree-optimization/83581
19578 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19579 TODO_cleanup_cfg if any changes have been made.
19581 PR middle-end/83608
19582 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19583 convert_modes if target mode has the right side, but different mode
19586 PR middle-end/83609
19587 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19588 last argument when extracting from CONCAT. If either from_real or
19589 from_imag is NULL, use expansion through memory. If result is not
19590 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19591 the parts directly to inner mode, if even that fails, use expansion
19594 PR middle-end/83623
19595 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19596 check for bswap in mode rather than HImode and use that in expand_unop
19599 Copyright (C) 2018 Free Software Foundation, Inc.
19601 Copying and distribution of this file, with or without modification,
19602 are permitted in any medium without royalty provided the copyright
19603 notice and this notice are preserved.